Forex Correlation Matrix - QuantShare Trading Software

Former investment bank FX trader: Risk management part II

Former investment bank FX trader: Risk management part II
Firstly, thanks for the overwhelming comments and feedback. Genuinely really appreciated. I am pleased 500+ of you find it useful.
If you didn't read the first post you can do so here: risk management part I. You'll need to do so in order to make sense of the topic.
As ever please comment/reply below with questions or feedback and I'll do my best to get back to you.
Part II
  • Letting stops breathe
  • When to change a stop
  • Entering and exiting winning positions
  • Risk:reward ratios
  • Risk-adjusted returns

Letting stops breathe

We talked earlier about giving a position enough room to breathe so it is not stopped out in day-to-day noise.
Let’s consider the chart below and imagine you had a trailing stop. It would be super painful to miss out on the wider move just because you left a stop that was too tight.

Imagine being long and stopped out on a meaningless retracement ... ouch!
One simple technique is simply to look at your chosen chart - let’s say daily bars. And then look at previous trends and use the measuring tool. Those generally look something like this and then you just click and drag to measure.
For example if we wanted to bet on a downtrend on the chart above we might look at the biggest retracement on the previous uptrend. That max drawdown was about 100 pips or just under 1%. So you’d want your stop to be able to withstand at least that.
If market conditions have changed - for example if CVIX has risen - and daily ranges are now higher you should incorporate that. If you know a big event is coming up you might think about that, too. The human brain is a remarkable tool and the power of the eye-ball method is not to be dismissed. This is how most discretionary traders do it.
There are also more analytical approaches.
Some look at the Average True Range (ATR). This attempts to capture the volatility of a pair, typically averaged over a number of sessions. It looks at three separate measures and takes the largest reading. Think of this as a moving average of how much a pair moves.
For example, below shows the daily move in EURUSD was around 60 pips before spiking to 140 pips in March. Conditions were clearly far more volatile in March. Accordingly, you would need to leave your stop further away in March and take a correspondingly smaller position size.

ATR is available on pretty much all charting systems
Professional traders tend to use standard deviation as a measure of volatility instead of ATR. There are advantages and disadvantages to both. Averages are useful but can be misleading when regimes switch (see above chart).
Once you have chosen a measure of volatility, stop distance can then be back-tested and optimised. For example does 2x ATR work best or 5x ATR for a given style and time horizon?
Discretionary traders may still eye-ball the ATR or standard deviation to get a feeling for how it has changed over time and what ‘normal’ feels like for a chosen study period - daily, weekly, monthly etc.

Reasons to change a stop

As a general rule you should be disciplined and not change your stops. Remember - losers average losers. This is really hard at first and we’re going to look at that in more detail later.
There are some good reasons to modify stops but they are rare.
One reason is if another risk management process demands you stop trading and close positions. We’ll look at this later. In that case just close out your positions at market and take the loss/gains as they are.
Another is event risk. If you have some big upcoming data like Non Farm Payrolls that you know can move the market +/- 150 pips and you have no edge going into the release then many traders will take off or scale down their positions. They’ll go back into the positions when the data is out and the market has quietened down after fifteen minutes or so. This is a matter of some debate - many traders consider it a coin toss and argue you win some and lose some and it all averages out.
Trailing stops can also be used to ‘lock in’ profits. We looked at those before. As the trade moves in your favour (say up if you are long) the stop loss ratchets with it. This means you may well end up ‘stopping out’ at a profit - as per the below example.

The mighty trailing stop loss order
It is perfectly reasonable to have your stop loss move in the direction of PNL. This is not exposing you to more risk than you originally were comfortable with. It is taking less and less risk as the trade moves in your favour. Trend-followers in particular love trailing stops.
One final question traders ask is what they should do if they get stopped out but still like the trade. Should they try the same trade again a day later for the same reasons? Nope. Look for a different trade rather than getting emotionally wed to the original idea.
Let’s say a particular stock looked cheap based on valuation metrics yesterday, you bought, it went down and you got stopped out. Well, it is going to look even better on those same metrics today. Maybe the market just doesn’t respect value at the moment and is driven by momentum. Wait it out.
Otherwise, why even have a stop in the first place?

Entering and exiting winning positions

Take profits are the opposite of stop losses. They are also resting orders, left with the broker, to automatically close your position if it reaches a certain price.
Imagine I’m long EURUSD at 1.1250. If it hits a previous high of 1.1400 (150 pips higher) I will leave a sell order to take profit and close the position.
The rookie mistake on take profits is to take profit too early. One should start from the assumption that you will win on no more than half of your trades. Therefore you will need to ensure that you win more on the ones that work than you lose on those that don’t.

Sad to say but incredibly common: retail traders often take profits way too early
This is going to be the exact opposite of what your emotions want you to do. We are going to look at that in the Psychology of Trading chapter.
Remember: let winners run. Just like stops you need to know in advance the level where you will close out at a profit. Then let the trade happen. Don’t override yourself and let emotions force you to take a small profit. A classic mistake to avoid.
The trader puts on a trade and it almost stops out before rebounding. As soon as it is slightly in the money they spook and cut out, instead of letting it run to their original take profit. Do not do this.

Entering positions with limit orders

That covers exiting a position but how about getting into one?
Take profits can also be left speculatively to enter a position. Sometimes referred to as “bids” (buy orders) or “offers” (sell orders). Imagine the price is 1.1250 and the recent low is 1.1205.
You might wish to leave a bid around 1.2010 to enter a long position, if the market reaches that price. This way you don’t need to sit at the computer and wait.
Again, typically traders will use tech analysis to identify attractive levels. Again - other traders will cluster with your orders. Just like the stop loss we need to bake that in.
So this time if we know everyone is going to buy around the recent low of 1.1205 we might leave the take profit bit a little bit above there at 1.1210 to ensure it gets done. Sure it costs 5 more pips but how mad would you be if the low was 1.1207 and then it rallied a hundred points and you didn’t have the trade on?!
There are two more methods that traders often use for entering a position.
Scaling in is one such technique. Let’s imagine that you think we are in a long-term bulltrend for AUDUSD but experiencing a brief retracement. You want to take a total position of 500,000 AUD and don’t have a strong view on the current price action.
You might therefore leave a series of five bids of 100,000. As the price moves lower each one gets hit. The nice thing about scaling in is it reduces pressure on you to pick the perfect level. Of course the risk is that not all your orders get hit before the price moves higher and you have to trade at-market.
Pyramiding is the second technique. Pyramiding is for take profits what a trailing stop loss is to regular stops. It is especially common for momentum traders.

Pyramiding into a position means buying more as it goes in your favour
Again let’s imagine we’re bullish AUDUSD and want to take a position of 500,000 AUD.
Here we add 100,000 when our first signal is reached. Then we add subsequent clips of 100,000 when the trade moves in our favour. We are waiting for confirmation that the move is correct.
Obviously this is quite nice as we humans love trading when it goes in our direction. However, the drawback is obvious: we haven’t had the full amount of risk on from the start of the trend.
You can see the attractions and drawbacks of both approaches. It is best to experiment and choose techniques that work for your own personal psychology as these will be the easiest for you to stick with and build a disciplined process around.

Risk:reward and win ratios

Be extremely skeptical of people who claim to win on 80% of trades. Most traders will win on roughly 50% of trades and lose on 50% of trades. This is why risk management is so important!
Once you start keeping a trading journal you’ll be able to see how the win/loss ratio looks for you. Until then, assume you’re typical and that every other trade will lose money.
If that is the case then you need to be sure you make more on the wins than you lose on the losses. You can see the effect of this below.

A combination of win % and risk:reward ratio determine if you are profitable
A typical rule of thumb is that a ratio of 1:3 works well for most traders.
That is, if you are prepared to risk 100 pips on your stop you should be setting a take profit at a level that would return you 300 pips.
One needn’t be religious about these numbers - 11 pips and 28 pips would be perfectly fine - but they are a guideline.
Again - you should still use technical analysis to find meaningful chart levels for both the stop and take profit. Don’t just blindly take your stop distance and do 3x the pips on the other side as your take profit. Use the ratio to set approximate targets and then look for a relevant resistance or support level in that kind of region.

Risk-adjusted returns

Not all returns are equal. Suppose you are examining the track record of two traders. Now, both have produced a return of 14% over the year. Not bad!
The first trader, however, made hundreds of small bets throughout the year and his cumulative PNL looked like the left image below.
The second trader made just one bet — he sold CADJPY at the start of the year — and his PNL looked like the right image below with lots of large drawdowns and volatility.
Would you rather have the first trading record or the second?
If you were investing money and betting on who would do well next year which would you choose? Of course all sensible people would choose the first trader. Yet if you look only at returns one cannot distinguish between the two. Both are up 14% at that point in time. This is where the Sharpe ratio helps .
A high Sharpe ratio indicates that a portfolio has better risk-adjusted performance. One cannot sensibly compare returns without considering the risk taken to earn that return.
If I can earn 80% of the return of another investor at only 50% of the risk then a rational investor should simply leverage me at 2x and enjoy 160% of the return at the same level of risk.
This is very important in the context of Execution Advisor algorithms (EAs) that are popular in the retail community. You must evaluate historic performance by its risk-adjusted return — not just the nominal return. Incidentally look at the Sharpe ratio of ones that have been live for a year or more ...
Otherwise an EA developer could produce two EAs: the first simply buys at 1000:1 leverage on January 1st ; and the second sells in the same manner. At the end of the year, one of them will be discarded and the other will look incredible. Its risk-adjusted return, however, would be abysmal and the odds of repeated success are similarly poor.

Sharpe ratio

The Sharpe ratio works like this:
  • It takes the average returns of your strategy;
  • It deducts from these the risk-free rate of return i.e. the rate anyone could have got by investing in US government bonds with very little risk;
  • It then divides this total return by its own volatility - the more smooth the return the higher and better the Sharpe, the more volatile the lower and worse the Sharpe.
For example, say the return last year was 15% with a volatility of 10% and US bonds are trading at 2%. That gives (15-2)/10 or a Sharpe ratio of 1.3. As a rule of thumb a Sharpe ratio of above 0.5 would be considered decent for a discretionary retail trader. Above 1 is excellent.
You don’t really need to know how to calculate Sharpe ratios. Good trading software will do this for you. It will either be available in the system by default or you can add a plug-in.

VAR

VAR is another useful measure to help with drawdowns. It stands for Value at Risk. Normally people will use 99% VAR (conservative) or 95% VAR (aggressive). Let’s say you’re long EURUSD and using 95% VAR. The system will look at the historic movement of EURUSD. It might spit out a number of -1.2%.

A 5% VAR of -1.2% tells you you should expect to lose 1.2% on 5% of days, whilst 95% of days should be better than that
This means it is expected that on 5 days out of 100 (hence the 95%) the portfolio will lose 1.2% or more. This can help you manage your capital by taking appropriately sized positions. Typically you would look at VAR across your portfolio of trades rather than trade by trade.
Sharpe ratios and VAR don’t give you the whole picture, though. Legendary fund manager, Howard Marks of Oaktree, notes that, while tools like VAR and Sharpe ratios are helpful and absolutely necessary, the best investors will also overlay their own judgment.
Investors can calculate risk metrics like VaR and Sharpe ratios (we use them at Oaktree; they’re the best tools we have), but they shouldn’t put too much faith in them. The bottom line for me is that risk management should be the responsibility of every participant in the investment process, applying experience, judgment and knowledge of the underlying investments.Howard Marks of Oaktree Capital
What he’s saying is don’t misplace your common sense. Do use these tools as they are helpful. However, you cannot fully rely on them. Both assume a normal distribution of returns. Whereas in real life you get “black swans” - events that should supposedly happen only once every thousand years but which actually seem to happen fairly often.
These outlier events are often referred to as “tail risk”. Don’t make the mistake of saying “well, the model said…” - overlay what the model is telling you with your own common sense and good judgment.

Coming up in part III

Available here
Squeezes and other risks
Market positioning
Bet correlation
Crap trades, timeouts and monthly limits

***
Disclaimer:This content is not investment advice and you should not place any reliance on it. The views expressed are the author's own and should not be attributed to any other person, including their employer.
submitted by getmrmarket to Forex [link] [comments]

Trading economic news

The majority of this sub is focused on technical analysis. I regularly ridicule such "tea leaf readers" and advocate for trading based on fundamentals and economic news instead, so I figured I should take the time to write up something on how exactly you can trade economic news releases.
This post is long as balls so I won't be upset if you get bored and go back to your drooping dick patterns or whatever.

How economic news is released

First, it helps to know how economic news is compiled and released. Let's take Initial Jobless Claims, the number of initial claims for unemployment benefits around the United States from Sunday through Saturday. Initial in this context means the first claim for benefits made by an individual during a particular stretch of unemployment. The Initial Jobless Claims figure appears in the Department of Labor's Unemployment Insurance Weekly Claims Report, which compiles information from all of the per-state departments that report to the DOL during the week. A typical number is between 100k and 250k and it can vary quite significantly week-to-week.
The Unemployment Insurance Weekly Claims Report contains data that lags 5 days behind. For example, the Report issued on Thursday March 26th 2020 contained data about the week ending on Saturday March 21st 2020.
In the days leading up to the Report, financial companies will survey economists and run complicated mathematical models to forecast the upcoming Initial Jobless Claims figure. The results of surveyed experts is called the "consensus"; specific companies, experts, and websites will also provide their own forecasts. Different companies will release different consensuses. Usually they are pretty close (within 2-3k), but for last week's record-high Initial Jobless Claims the reported consensuses varied by up to 1M! In other words, there was essentially no consensus.
The Unemployment Insurance Weekly Claims Report is released each Thursday morning at exactly 8:30 AM ET. (On Thanksgiving the Report is released on Wednesday instead.) Media representatives gather at the Frances Perkins Building in Washington DC and are admitted to the "lockup" at 8:00 AM ET. In order to be admitted to the lockup you have to be a credentialed member of a media organization that has signed the DOL lockup agreement. The lockup room is small so there is a limited number of spots.
No phones are allowed. Reporters bring their laptops and connect to a local network; there is a master switch on the wall that prevents/enables Internet connectivity on this network. Once the doors are closed the Unemployment Insurance Weekly Claims Report is distributed, with a heading that announces it is "embargoed" (not to be released) prior to 8:30 AM. Reporters type up their analyses of the report, including extracting key figures like Initial Jobless Claims. They load their write-ups into their companies' software, which prepares to send it out as soon as Internet is enabled. At 8:30 AM the DOL representative in the room flips the wall switch and all of the laptops are connected to the Internet, releasing their write-ups to their companies and on to their companies' partners.
Many of those media companies have externally accessible APIs for distributing news. Media aggregators and squawk services (like RanSquawk and TradeTheNews) subscribe to all of these different APIs and then redistribute the key economic figures from the Report to their own subscribers within one second after Internet is enabled in the DOL lockup.
Some squawk services are text-based while others are audio-based. FinancialJuice.com provides a free audio squawk service; internally they have a paid subscription to a professional squawk service and they simply read out the latest headlines to their own listeners, subsidized by ads on the site. I've been using it for 4 months now and have been pretty happy. It usually lags behind the official release times by 1-2 seconds and occasionally they verbally flub the numbers or stutter and have to repeat, but you can't beat the price!
Important - I’m not affiliated with FinancialJuice and I’m not advocating that you use them over any other squawk. If you use them and they misspeak a number and you lose all your money don’t blame me. If anybody has any other free alternatives please share them!

How the news affects forex markets

Institutional forex traders subscribe to these squawk services and use custom software to consume the emerging data programmatically and then automatically initiate trades based on the perceived change to the fundamentals that the figures represent.
It's important to note that every institution will have "priced in" their own forecasted figures well in advance of an actual news release. Forecasts and consensuses all come out at different times in the days leading up to a news release, so by the time the news drops everybody is really only looking for an unexpected result. You can't really know what any given institution expects the value to be, but unless someone has inside information you can pretty much assume that the market has collectively priced in the experts' consensus. When the news comes out, institutions will trade based on the difference between the actual and their forecast.
Sometimes the news reflects a real change to the fundamentals with an economic effect that will change the demand for a currency, like an interest rate decision. However, in the case of the Initial Jobless Claims figure, which is a backwards-looking metric, trading is really just self-fulfilling speculation that market participants will buy dollars when unemployment is low and sell dollars when unemployment is high. Generally speaking, news that reflects a real economic shift has a bigger effect than news that only matters to speculators.
Massive and extremely fast news-based trades happen within tenths of a second on the ECNs on which institutional traders are participants. Over the next few seconds the resulting price changes trickle down to retail traders. Some economic news, like Non Farm Payroll Employment, has an effect that can last minutes to hours as "slow money" follows behind on the trend created by the "fast money". Other news, like Initial Jobless Claims, has a short impact that trails off within a couple minutes and is subsequently dwarfed by the usual pseudorandom movements in the market.
The bigger the difference between actual and consensus, the bigger the effect on any given currency pair. Since economic news releases generally relate to a single currency, the biggest and most easily predicted effects are seen on pairs where one currency is directly effected and the other is not affected at all. Personally I trade USD/JPY because the time difference between the US and Japan ensures that no news will be coming out of Japan at the same time that economic news is being released in the US.
Before deciding to trade any particular news release you should measure the historical correlation between the release (specifically, the difference between actual and consensus) and the resulting short-term change in the currency pair. Historical data for various news releases (along with historical consensus data) is readily available. You can pay to get it exported into Excel or whatever, or you can scroll through it for free on websites like TradingEconomics.com.
Let's look at two examples: Initial Jobless Claims and Non Farm Payroll Employment (NFP). I collected historical consensuses and actuals for these releases from January 2018 through the present, measured the "surprise" difference for each, and then correlated that to short-term changes in USD/JPY at the time of release using 5 second candles.
I omitted any releases that occurred simultaneously as another major release. For example, occasionally the monthly Initial Jobless Claims comes out at the exact same time as the monthly Balance of Trade figure, which is a more significant economic indicator and can be expected to dwarf the effect of the Unemployment Insurance Weekly Claims Report.
USD/JPY correlation with Initial Jobless Claims (2018 - present)
USD/JPY correlation with Non Farm Payrolls (2018 - present)
The horizontal axes on these charts is the duration (in seconds) after the news release over which correlation was calculated. The vertical axis is the Pearson correlation coefficient: +1 means that the change in USD/JPY over that duration was perfectly linearly correlated to the "surprise" in the releases; -1 means that the change in USD/JPY was perfectly linearly correlated but in the opposite direction, and 0 means that there is no correlation at all.
For Initial Jobless Claims you can see that for the first 30 seconds USD/JPY is strongly negatively correlated with the difference between consensus and actual jobless claims. That is, fewer-than-forecast jobless claims (fewer newly unemployed people than expected) strengthens the dollar and greater-than-forecast jobless claims (more newly unemployed people than expected) weakens the dollar. Correlation then trails off and changes to a moderate/weak positive correlation. I interpret this as algorithms "buying the dip" and vice versa, but I don't know for sure. From this chart it appears that you could profit by opening a trade for 15 seconds (duration with strongest correlation) that is long USD/JPY when Initial Jobless Claims is lower than the consensus and short USD/JPY when Initial Jobless Claims is higher than expected.
The chart for Non Farm Payroll looks very different. Correlation is positive (higher-than-expected payrolls strengthen the dollar and lower-than-expected payrolls weaken the dollar) and peaks at around 45 seconds, then slowly decreases as time goes on. This implies that price changes due to NFP are quite significant relative to background noise and "stick" even as normal fluctuations pick back up.
I wanted to show an example of what the USD/JPY S5 chart looks like when an "uncontested" (no other major simultaneously news release) Initial Jobless Claims and NFP drops, but unfortunately my broker's charts only go back a week. (I can pull historical data going back years through the API but to make it into a pretty chart would be a bit of work.) If anybody can get a 5-second chart of USD/JPY at March 19, 2020, UTC 12:30 and/or at February 7, 2020, UTC 13:30 let me know and I'll add it here.

Backtesting

So without too much effort we determined that (1) USD/JPY is strongly negatively correlated with the Initial Jobless Claims figure for the first 15 seconds after the release of the Unemployment Insurance Weekly Claims Report (when no other major news is being released) and also that (2) USD/JPY is strongly positively correlated with the Non Farms Payroll figure for the first 45 seconds after the release of the Employment Situation report.
Before you can assume you can profit off the news you have to backtest and consider three important parameters.
Entry speed: How quickly can you realistically enter the trade? The correlation performed above was measured from the exact moment the news was released, but realistically if you've got your finger on the trigger and your ear to the squawk it will take a few seconds to hit "Buy" or "Sell" and confirm. If 90% of the price move happens in the first second you're SOL. For back-testing purposes I assume a 5 second delay. In practice I use custom software that opens a trade with one click, and I can reliably enter a trade within 2-3 seconds after the news drops, using the FinancialJuice free squawk.
Minimum surprise: Should you trade every release or can you do better by only trading those with a big enough "surprise" factor? Backtesting will tell you whether being more selective is better long-term or not.
Hold time: The optimal time to hold the trade is not necessarily the same as the time of maximum correlation. That's a good starting point but it's not necessarily the best number. Backtesting each possible hold time will let you find the best one.
The spread: When you're only holding a position open for 30 seconds, the spread will kill you. The correlations performed above used the midpoint price, but in reality you have to buy at the ask and sell at the bid. Brokers aren't stupid and the moment volume on the ECN jumps they will widen the spread for their retail customers. The only way to determine if the news-driven price movements reliably overcome the spread is to backtest.
Stops: Personally I don't use stops, neither take-profit nor stop-loss, since I'm automatically closing the trade after a fixed (and very short) amount of time. Additionally, brokers have a minimum stop distance; the profits from scalping the news are so slim that even the nearest stops they allow will generally not get triggered.
I backtested trading these two news releases (since 2018), using a 5 second entry delay, real historical spreads, and no stops, cycling through different "surprise" thresholds and hold times to find the combination that returns the highest net profit. It's important to maximize net profit, not expected value per trade, so you don't over-optimize and reduce the total number of trades taken to one single profitable trade. If you want to get fancy you can set up a custom metric that combines number of trades, expected value, and drawdown into a single score to be maximized.
For the Initial Jobless Claims figure I found that the best combination is to hold trades open for 25 seconds (that is, open at 5 seconds elapsed and hold until 30 seconds elapsed) and only trade when the difference between consensus and actual is 7k or higher. That leads to 30 trades taken since 2018 and an expected return of... drumroll please... -0.0093 yen per unit per trade.
Yep, that's a loss of approx. $8.63 per lot.
Disappointing right? That's the spread and that's why you have to backtest. Even though the release of the Unemployment Insurance Weekly Claims Report has a strong correlation with movement in USD/JPY, it's simply not something that a retail trader can profit from.
Let's turn to the NFP. There I found that the best combination is to hold trades open for 75 seconds (that is, open at 5 seconds elapsed and hold until 80 seconds elapsed) and trade every single NFP (no minimum "surprise" threshold). That leads to 20 trades taken since 2018 and an expected return of... drumroll please... +0.1306 yen per unit per trade.
That's a profit of approx. $121.25 per lot. Not bad for 75 seconds of work! That's a +6% ROI at 50x leverage.

Make it real

If you want to do this for realsies, you need to run these numbers for all of the major economic news releases. Markit Manufacturing PMI, Factory Orders MoM, Trade Balance, PPI MoM, Export and Import Prices, Michigan Consumer Sentiment, Retail Sales MoM, Industrial Production MoM, you get the idea. You keep a list of all of the releases you want to trade, when they are released, and the ideal hold time and "surprise" threshold. A few minutes before the prescribed release time you open up your broker's software, turn on your squawk, maybe jot a few notes about consensuses and model forecasts, and get your finger on the button. At the moment you hear the release you open the trade in the correct direction, hold it (without looking at the chart!) for the required amount of time, then close it and go on with your day.
Some benefits of trading this way: * Most major economic releases come out at either 8:30 AM ET or 10:00 AM ET, and then you're done for the day. * It's easily backtestable. You can look back at the numbers and see exactly what to expect your return to be. * It's fun! Packing your trading into 30 seconds and knowing that institutions are moving billions of dollars around as fast as they can based on the exact same news you just read is thrilling. * You can wow your friends by saying things like "The St. Louis Fed had some interesting remarks on consumer spending in the latest Beige Book." * No crayons involved.
Some downsides: * It's tricky to be fast enough without writing custom software. Some broker software is very slow and requires multiple dialog boxes before a position is opened, which won't cut it. * The profits are very slim, you're not going to impress your instagram followers to join your expensive trade copying service with your 30-second twice-weekly trades. * Any friends you might wow with your boring-ass economic talking points are themselves the most boring people in the world.
I hope you enjoyed this long as fuck post and you give trading economic news a try!
submitted by thicc_dads_club to Forex [link] [comments]

I like my numbers how I like my men

Hard.
Following up on my post about trading economic news, here are some hard numbers about news releases. If you can't be bothered to read any further, just know that a high S value means there is money to be made by trading the release of that economic metric.
I collected historical data (consensus and actual) for each of these economic metrics from January 2018 to the present. These are mostly monthly metrics; one is quarterly, one is biweekly, and one is weekly. I did this manually from an economic news aggregation website because I'm too cheap to pay for exported data. I also noted whether each release was the primary (or only) news being released at that precise moment or whether there were other important economic metrics being released at the same time.
Then I wrote software to fetch 2 minutes of USD/JPY price data (in 5 second candles) starting at the time of each release and correlated each candle to the "surprise" in the metric (the difference between the consensus and the actual). That produced data for each metric that looks like this and give you an idea how reliably the news predicts the short-term price move.
Next the software went back through the price history and measured the "coordinated movement" of the currency pair during each candle. By that I mean it measured the size of each candle in pips and then set the sign to be positive if its moving in the same direction as the first candle or negative otherwise, and averaged all those numbers together. These charts look like this and give you an idea how dramatically the price moves in response to the news.
Then, for each candle, the absolute value of the correlation is multiplied against the coordinated movement, creating a "tradeability" score. The idea is that a strong (positive or negative) correlation and a strong coordinated movement yields a high score, and either a weak correlation or a small price movement yields a low score. These charts look like this.
Finally, for each news release I measured the maximum value of the correlation (R) and the maximum value of the tradeability score (S). Since I don't have access to anything with finer granularity than 5 seconds, it's no surprise that candle 0 had the maximum correlation and maximum tradeability for each metric.
I did this process twice for every metric: first over all releases and second over only those releases that were the primary or only release happening at that moment. I kept the results from the one that returned the better net maximum tradeability score. The net maximum tradeability score is just the product of the maximum tradeability score and the number of releases being considered, either all of them or some smaller number. (This helps avoid bias in the results for metrics that are often released at the same time as other metrics.)
In the results linked at the top of this post, the metrics are ordered by decreasing maximum tradeability score (S) with maximum correlation (R) also shown. Remember, R = +1 means perfect positive linear correlation, R = -1 means perfect negative linear correlation, and R = 0 means no correlation. S values less than 3 are essentially untradeable due to the spread.
A couple notes:
It takes me about 15 minutes to scrape 2 years of monthly data by hand and type it up, and then the software takes about 10 seconds to run per metric. If there are regular forecasted economic releases that you'd like to see correlated and scored, let me know! I can do other currency pairs and other country's economic news as well, it's just a matter of data collection.
submitted by thicc_dads_club to Forex [link] [comments]

Anyone doing automated trading? I really need a piece of advice

Hello everyone!
At the beginning I would like to tell something about how I came here.
I'm 23 years old software developer and I have to admit, that I really like my job, at least for now. It's hard to predict what will happen in next few years, but I think that life is just too short to spend half of it in a 9-5 job living frugal and then spend most of my savings to build a house of my dreams when I'll be about 50 years old and blow my retirement (in my country there is no such thing like 401k, it looks more like a money stealing organization, so I count on myself). This is not how I want to live and it's my main motivation why I started learning to trade.
So I started learning Forex, because of it's liquidity and high leverage. I know it' not the easiest market to undestand, but who said it's going to be easy?
I learned the basics and how it works and I also learned that trading on lower time frames is harder, but it can be much more rewarding. That's what I was looking for.
At the first I was fascinated by price action trading. The concept that I need just a price chart and no indicators sounds cool. But there are some disadvantages:
- There is a really strong human factor. When looking at the chart, someone will say that he sees a buy oportunity and another one that it's clearly short. Chart can be interpreted in many ways when we trade with price action.
- Most volatile periods are when I'm at work, so it's hard for me to practice
I know that price action trading can be very profitable, but I think it just doesn't fit me.
I wondered if I could connect my programming skills with trading and started to learn about algorithmic trading, but there isn't as much resources as in manual trading.
I'm not interested in HFT (High Frequency Trending). It's probably even not possible to make a profitable HFT robot for a normal person without super fast internet and close connection to source. I would like to make a trend catcher for lower time frames that would make just a few trades a day. I learned some mql4, made my first trading robot (not very effective), wrote some indicators to notice some correlations and I'm just playing a bit right now.
But I think that it's time to pause and think about it. Now it's time for my questions (finally).
Basically, there are two main opinions:
  1. To write a profitable robot, you need to know as much as possible about how markets work, how they work together on a very deep level. Learn and read everything and maybe in a few years you will know enough to gain profit from it.
  2. Most profitable robots are as simple as possible. Just stick to technical analysis, risk management, probability etc.

What' your opinion about that? And more importantly, is it worth to give it a shot? Sometimes I think that I just waste my time, because I know I can make a lot of money as a software developer and I should focus to improve myself in that particular thing. But it was never about money. It's all about freedom. I want to travel, I want to drive cars most people can't afford (not to show off, I just love driving) and build a house without spending savings of my life on it.
I will appreciate any feedback :)
submitted by Eksagnem to Daytrading [link] [comments]

Subreddit Stats: cs7646_fall2017 top posts from 2017-08-23 to 2017-12-10 22:43 PDT

Period: 108.98 days
Submissions Comments
Total 999 10425
Rate (per day) 9.17 95.73
Unique Redditors 361 695
Combined Score 4162 17424

Top Submitters' Top Submissions

  1. 296 points, 24 submissions: tuckerbalch
    1. Project 2 Megathread (optimize_something) (33 points, 475 comments)
    2. project 3 megathread (assess_learners) (27 points, 1130 comments)
    3. For online students: Participation check #2 (23 points, 47 comments)
    4. ML / Data Scientist internship and full time job opportunities (20 points, 36 comments)
    5. Advance information on Project 3 (19 points, 22 comments)
    6. participation check #3 (19 points, 29 comments)
    7. manual_strategy project megathread (17 points, 825 comments)
    8. project 4 megathread (defeat_learners) (15 points, 209 comments)
    9. project 5 megathread (marketsim) (15 points, 484 comments)
    10. QLearning Robot project megathread (12 points, 691 comments)
  2. 278 points, 17 submissions: davebyrd
    1. A little more on Pandas indexing/slicing ([] vs ix vs iloc vs loc) and numpy shapes (37 points, 10 comments)
    2. Project 1 Megathread (assess_portfolio) (34 points, 466 comments)
    3. marketsim grades are up (25 points, 28 comments)
    4. Midterm stats (24 points, 32 comments)
    5. Welcome to CS 7646 MLT! (23 points, 132 comments)
    6. How to interact with TAs, discuss grades, performance, request exceptions... (18 points, 31 comments)
    7. assess_portfolio grades have been released (18 points, 34 comments)
    8. Midterm grades posted to T-Square (15 points, 30 comments)
    9. Removed posts (15 points, 2 comments)
    10. assess_portfolio IMPORTANT README: about sample frequency (13 points, 26 comments)
  3. 118 points, 17 submissions: yokh_cs7646
    1. Exam 2 Information (39 points, 40 comments)
    2. Reformat Assignment Pages? (14 points, 2 comments)
    3. What did the real-life Michael Burry have to say? (13 points, 2 comments)
    4. PSA: Read the Rubric carefully and ahead-of-time (8 points, 15 comments)
    5. How do I know that I'm correct and not just lucky? (7 points, 31 comments)
    6. ML Papers and News (7 points, 5 comments)
    7. What are "question pools"? (6 points, 4 comments)
    8. Explanation of "Regression" (5 points, 5 comments)
    9. GT Github taking FOREVER to push to..? (4 points, 14 comments)
    10. Dead links on the course wiki (3 points, 2 comments)
  4. 67 points, 13 submissions: harshsikka123
    1. To all those struggling, some words of courage! (20 points, 18 comments)
    2. Just got locked out of my apartment, am submitting from a stairwell (19 points, 12 comments)
    3. Thoroughly enjoying the lectures, some of the best I've seen! (13 points, 13 comments)
    4. Just for reference, how long did Assignment 1 take you all to implement? (3 points, 31 comments)
    5. Grade_Learners Taking about 7 seconds on Buffet vs 5 on Local, is this acceptable if all tests are passing? (2 points, 2 comments)
    6. Is anyone running into the Runtime Error, Invalid DISPLAY variable when trying to save the figures as pdfs to the Buffet servers? (2 points, 9 comments)
    7. Still not seeing an ML4T onboarding test on ProctorTrack (2 points, 10 comments)
    8. Any news on when Optimize_Something grades will be released? (1 point, 1 comment)
    9. Baglearner RMSE and leaf size? (1 point, 2 comments)
    10. My results are oh so slightly off, any thoughts? (1 point, 11 comments)
  5. 63 points, 10 submissions: htrajan
    1. Sample test case: missing data (22 points, 36 comments)
    2. Optimize_something test cases (13 points, 22 comments)
    3. Met Burt Malkiel today (6 points, 1 comment)
    4. Heads up: Dataframe.std != np.std (5 points, 5 comments)
    5. optimize_something: graph (5 points, 29 comments)
    6. Schedule still reflecting shortened summer timeframe? (4 points, 3 comments)
    7. Quick clarification about InsaneLearner (3 points, 8 comments)
    8. Test cases using rfr? (3 points, 5 comments)
    9. Input format of rfr (2 points, 1 comment)
    10. [Shameless recruiting post] Wealthfront is hiring! (0 points, 9 comments)
  6. 62 points, 7 submissions: swamijay
    1. defeat_learner test case (34 points, 38 comments)
    2. Project 3 test cases (15 points, 27 comments)
    3. Defeat_Learner - related questions (6 points, 9 comments)
    4. Options risk/reward (2 points, 0 comments)
    5. manual strategy - you must remain in the position for 21 trading days. (2 points, 9 comments)
    6. standardizing values (2 points, 0 comments)
    7. technical indicators - period for moving averages, or anything that looks past n days (1 point, 3 comments)
  7. 61 points, 9 submissions: gatech-raleighite
    1. Protip: Better reddit search (22 points, 9 comments)
    2. Helpful numpy array cheat sheet (16 points, 10 comments)
    3. In your experience Professor, Mr. Byrd, which strategy is "best" for trading ? (12 points, 10 comments)
    4. Industrial strength or mature versions of the assignments ? (4 points, 2 comments)
    5. What is the correct (faster) way of doing this bit of pandas code (updating multiple slice values) (2 points, 10 comments)
    6. What is the correct (pythonesque?) way to select 60% of rows ? (2 points, 11 comments)
    7. How to get adjusted close price for funds not publicly traded (TSP) ? (1 point, 2 comments)
    8. Is there a way to only test one or 2 of the learners using grade_learners.py ? (1 point, 10 comments)
    9. OMS CS Digital Career Seminar Series - Scott Leitstein recording available online? (1 point, 4 comments)
  8. 60 points, 2 submissions: reyallan
    1. [Project Questions] Unit Tests for assess_portfolio assignment (58 points, 52 comments)
    2. Financial data, technical indicators and live trading (2 points, 8 comments)
  9. 59 points, 12 submissions: dyllll
    1. Please upvote helpful posts and other advice. (26 points, 1 comment)
    2. Books to further study in trading with machine learning? (14 points, 9 comments)
    3. Is Q-Learning the best reinforcement learning method for stock trading? (4 points, 4 comments)
    4. Any way to download the lessons? (3 points, 4 comments)
    5. Can a TA please contact me? (2 points, 7 comments)
    6. Is the vectorization code from the youtube video available to us? (2 points, 2 comments)
    7. Position of webcam (2 points, 15 comments)
    8. Question about assignment one (2 points, 5 comments)
    9. Are udacity quizzes recorded? (1 point, 2 comments)
    10. Does normalization of indicators matter in a Q-Learner? (1 point, 7 comments)
  10. 56 points, 2 submissions: jan-laszlo
    1. Proper git workflow (43 points, 19 comments)
    2. Adding you SSH key for password-less access to remote hosts (13 points, 7 comments)
  11. 53 points, 1 submission: agifft3_omscs
    1. [Project Questions] Unit Tests for optimize_something assignment (53 points, 94 comments)
  12. 50 points, 16 submissions: BNielson
    1. Regression Trees (7 points, 9 comments)
    2. Two Interpretations of RFR are leading to two different possible Sharpe Ratios -- Need Instructor clarification ASAP (5 points, 3 comments)
    3. PYTHONPATH=../:. python grade_analysis.py (4 points, 7 comments)
    4. Running on Windows and PyCharm (4 points, 4 comments)
    5. Studying for the midterm: python questions (4 points, 0 comments)
    6. Assess Learners Grader (3 points, 2 comments)
    7. Manual Strategy Grade (3 points, 2 comments)
    8. Rewards in Q Learning (3 points, 3 comments)
    9. SSH/Putty on Windows (3 points, 4 comments)
    10. Slight contradiction on ProctorTrack Exam (3 points, 4 comments)
  13. 49 points, 7 submissions: j0shj0nes
    1. QLearning Robot - Finalized and Released Soon? (18 points, 4 comments)
    2. Flash Boys, HFT, frontrunning... (10 points, 3 comments)
    3. Deprecations / errata (7 points, 5 comments)
    4. Udacity lectures via GT account, versus personal account (6 points, 2 comments)
    5. Python: console-driven development (5 points, 5 comments)
    6. Buffet pandas / numpy versions (2 points, 2 comments)
    7. Quant research on earnings calls (1 point, 0 comments)
  14. 45 points, 11 submissions: Zapurza
    1. Suggestion for Strategy learner mega thread. (14 points, 1 comment)
    2. Which lectures to watch for upcoming project q learning robot? (7 points, 5 comments)
    3. In schedule file, there is no link against 'voting ensemble strategy'? Scheduled for Nov 13-20 week (6 points, 3 comments)
    4. How to add questions to the question bank? I can see there is 2% credit for that. (4 points, 5 comments)
    5. Scratch paper use (3 points, 6 comments)
    6. The big short movie link on you tube says the video is not available in your country. (3 points, 9 comments)
    7. Distance between training data date and future forecast date (2 points, 2 comments)
    8. News affecting stock market and machine learning algorithms (2 points, 4 comments)
    9. pandas import in pydev (2 points, 0 comments)
    10. Assess learner server error (1 point, 2 comments)
  15. 43 points, 23 submissions: chvbs2000
    1. Is the Strategy Learner finalized? (10 points, 3 comments)
    2. Test extra 15 test cases for marketsim (3 points, 12 comments)
    3. Confusion between the term computing "back-in time" and "going forward" (2 points, 1 comment)
    4. How to define "each transaction"? (2 points, 4 comments)
    5. How to filling the assignment into Jupyter Notebook? (2 points, 4 comments)
    6. IOError: File ../data/SPY.csv does not exist (2 points, 4 comments)
    7. Issue in Access to machines at Georgia Tech via MacOS terminal (2 points, 5 comments)
    8. Reading data from Jupyter Notebook (2 points, 3 comments)
    9. benchmark vs manual strategy vs best possible strategy (2 points, 2 comments)
    10. global name 'pd' is not defined (2 points, 4 comments)
  16. 43 points, 15 submissions: shuang379
    1. How to test my code on buffet machine? (10 points, 15 comments)
    2. Can we get the ppt for "Decision Trees"? (8 points, 2 comments)
    3. python question pool question (5 points, 6 comments)
    4. set up problems (3 points, 4 comments)
    5. Do I need another camera for scanning? (2 points, 9 comments)
    6. Is chapter 9 covered by the midterm? (2 points, 2 comments)
    7. Why grade_analysis.py could run even if I rm analysis.py? (2 points, 5 comments)
    8. python question pool No.48 (2 points, 6 comments)
    9. where could we find old versions of the rest projects? (2 points, 2 comments)
    10. where to put ml4t-libraries to install those libraries? (2 points, 1 comment)
  17. 42 points, 14 submissions: larrva
    1. is there a mistake in How-to-learn-a-decision-tree.pdf (7 points, 7 comments)
    2. maximum recursion depth problem (6 points, 10 comments)
    3. [Urgent]Unable to use proctortrack in China (4 points, 21 comments)
    4. manual_strategynumber of indicators to use (3 points, 10 comments)
    5. Assignment 2: Got 63 points. (3 points, 3 comments)
    6. Software installation workshop (3 points, 7 comments)
    7. question regarding functools32 version (3 points, 3 comments)
    8. workshop on Aug 31 (3 points, 8 comments)
    9. Mount remote server to local machine (2 points, 2 comments)
    10. any suggestion on objective function (2 points, 3 comments)
  18. 41 points, 8 submissions: Ran__Ran
    1. Any resource will be available for final exam? (19 points, 6 comments)
    2. Need clarification on size of X, Y in defeat_learners (7 points, 10 comments)
    3. Get the same date format as in example chart (4 points, 3 comments)
    4. Cannot log in GitHub Desktop using GT account? (3 points, 3 comments)
    5. Do we have notes or ppt for Time Series Data? (3 points, 5 comments)
    6. Can we know the commission & market impact for short example? (2 points, 7 comments)
    7. Course schedule export issue (2 points, 15 comments)
    8. Buying/seeking beta v.s. buying/seeking alpha (1 point, 6 comments)
  19. 38 points, 4 submissions: ProudRamblinWreck
    1. Exam 2 Study topics (21 points, 5 comments)
    2. Reddit participation as part of grade? (13 points, 32 comments)
    3. Will birds chirping in the background flag me on Proctortrack? (3 points, 5 comments)
    4. Midterm Study Guide question pools (1 point, 2 comments)
  20. 37 points, 6 submissions: gatechben
    1. Submission page for strategy learner? (14 points, 10 comments)
    2. PSA: The grading script for strategy_learner changed on the 26th (10 points, 9 comments)
    3. Where is util.py supposed to be located? (8 points, 8 comments)
    4. PSA:. The default dates in the assignment 1 template are not the same as the examples on the assignment page. (2 points, 1 comment)
    5. Schedule: Discussion of upcoming trading projects? (2 points, 3 comments)
    6. [defeat_learners] More than one column for X? (1 point, 1 comment)
  21. 37 points, 3 submissions: jgeiger
    1. Please send/announce when changes are made to the project code (23 points, 7 comments)
    2. The Big Short on Netflix for OMSCS students (week of 10/16) (11 points, 6 comments)
    3. Typo(?) for Assess_portfolio wiki page (3 points, 2 comments)
  22. 35 points, 10 submissions: ltian35
    1. selecting row using .ix (8 points, 9 comments)
    2. Will the following 2 topics be included in the final exam(online student)? (7 points, 4 comments)
    3. udacity quiz (7 points, 4 comments)
    4. pdf of lecture (3 points, 4 comments)
    5. print friendly version of the course schedule (3 points, 9 comments)
    6. about learner regression vs classificaiton (2 points, 2 comments)
    7. is there a simple way to verify the correctness of our decision tree (2 points, 4 comments)
    8. about Building an ML-based forex strategy (1 point, 2 comments)
    9. about technical analysis (1 point, 6 comments)
    10. final exam online time period (1 point, 2 comments)
  23. 33 points, 2 submissions: bhrolenok
    1. Assess learners template and grading script is now available in the public repository (24 points, 0 comments)
    2. Tutorial for software setup on Windows (9 points, 35 comments)
  24. 31 points, 4 submissions: johannes_92
    1. Deadline extension? (26 points, 40 comments)
    2. Pandas date indexing issues (2 points, 5 comments)
    3. Why do we subtract 1 from SMA calculation? (2 points, 3 comments)
    4. Unexpected number of calls to query, sum=20 (should be 20), max=20 (should be 1), min=20 (should be 1) -bash: syntax error near unexpected token `(' (1 point, 3 comments)
  25. 30 points, 5 submissions: log_base_pi
    1. The Massive Hedge Fund Betting on AI [Article] (9 points, 1 comment)
    2. Useful Python tips and tricks (8 points, 10 comments)
    3. Video of overview of remaining projects with Tucker Balch (7 points, 1 comment)
    4. Will any material from the lecture by Goldman Sachs be covered on the exam? (5 points, 1 comment)
    5. What will the 2nd half of the course be like? (1 point, 8 comments)
  26. 30 points, 4 submissions: acschwabe
    1. Assignment and Exam Calendar (ICS File) (17 points, 6 comments)
    2. Please OMG give us any options for extra credit (8 points, 12 comments)
    3. Strategy learner question (3 points, 1 comment)
    4. Proctortrack: Do we need to schedule our test time? (2 points, 10 comments)
  27. 29 points, 9 submissions: _ant0n_
    1. Next assignment? (9 points, 6 comments)
    2. Proctortrack Onboarding test? (6 points, 11 comments)
    3. Manual strategy: Allowable positions (3 points, 7 comments)
    4. Anyone watched Black Scholes documentary? (2 points, 16 comments)
    5. Buffet machines hardware (2 points, 6 comments)
    6. Defeat learners: clarification (2 points, 4 comments)
    7. Is 'optimize_something' on the way to class GitHub repo? (2 points, 6 comments)
    8. assess_portfolio(... gen_plot=True) (2 points, 8 comments)
    9. remote job != remote + international? (1 point, 15 comments)
  28. 26 points, 10 submissions: umersaalis
    1. comments.txt (7 points, 6 comments)
    2. Assignment 2: report.pdf (6 points, 30 comments)
    3. Assignment 2: report.pdf sharing & plagiarism (3 points, 12 comments)
    4. Max Recursion Limit (3 points, 10 comments)
    5. Parametric vs Non-Parametric Model (3 points, 13 comments)
    6. Bag Learner Training (1 point, 2 comments)
    7. Decision Tree Issue: (1 point, 2 comments)
    8. Error in Running DTLearner and RTLearner (1 point, 12 comments)
    9. My Results for the four learners. Please check if you guys are getting values somewhat near to these. Exact match may not be there due to randomization. (1 point, 4 comments)
    10. Can we add the assignments and solutions to our public github profile? (0 points, 7 comments)
  29. 26 points, 6 submissions: abiele
    1. Recommended Reading? (13 points, 1 comment)
    2. Number of Indicators Used by Actual Trading Systems (7 points, 6 comments)
    3. Software Install Instructions From TA's Video Not Working (2 points, 2 comments)
    4. Suggest that TA/Instructor Contact Info Should be Added to the Syllabus (2 points, 2 comments)
    5. ML4T Software Setup (1 point, 3 comments)
    6. Where can I find the grading folder? (1 point, 4 comments)
  30. 26 points, 6 submissions: tomatonight
    1. Do we have all the information needed to finish the last project Strategy learner? (15 points, 3 comments)
    2. Does anyone interested in cryptocurrency trading/investing/others? (3 points, 6 comments)
    3. length of portfolio daily return (3 points, 2 comments)
    4. Did Michael Burry, Jamie&Charlie enter the short position too early? (2 points, 4 comments)
    5. where to check participation score (2 points, 1 comment)
    6. Where to collect the midterm exam? (forgot to take it last week) (1 point, 3 comments)
  31. 26 points, 3 submissions: hilo260
    1. Is there a template for optimize_something on GitHub? (14 points, 3 comments)
    2. Marketism project? (8 points, 6 comments)
    3. "Do not change the API" (4 points, 7 comments)
  32. 26 points, 3 submissions: niufen
    1. Windows Server Setup Guide (23 points, 16 comments)
    2. Strategy Learner Adding UserID as Comment (2 points, 2 comments)
    3. Connect to server via Python Error (1 point, 6 comments)
  33. 26 points, 3 submissions: whoyoung99
    1. How much time you spend on Assess Learner? (13 points, 47 comments)
    2. Git clone repository without fork (8 points, 2 comments)
    3. Just for fun (5 points, 1 comment)
  34. 25 points, 8 submissions: SharjeelHanif
    1. When can we discuss defeat learners methods? (10 points, 1 comment)
    2. Are the buffet servers really down? (3 points, 2 comments)
    3. Are the midterm results in proctortrack gone? (3 points, 3 comments)
    4. Will these finance topics be covered on the final? (3 points, 9 comments)
    5. Anyone get set up with Proctortrack? (2 points, 10 comments)
    6. Incentives Quiz Discussion (2-01, Lesson 11.8) (2 points, 3 comments)
    7. Anyone from Houston, TX (1 point, 1 comment)
    8. How can I trace my error back to a line of code? (assess learners) (1 point, 3 comments)
  35. 25 points, 5 submissions: jlamberts3
    1. Conda vs VirtualEnv (7 points, 8 comments)
    2. Cool Portfolio Backtesting Tool (6 points, 6 comments)
    3. Warren Buffett wins $1M bet made a decade ago that the S&P 500 stock index would outperform hedge funds (6 points, 12 comments)
    4. Windows Ubuntu Subsystem Putty Alternative (4 points, 0 comments)
    5. Algorithmic Trading Of Digital Assets (2 points, 0 comments)
  36. 25 points, 4 submissions: suman_paul
    1. Grade statistics (9 points, 3 comments)
    2. Machine Learning book by Mitchell (6 points, 11 comments)
    3. Thank You (6 points, 6 comments)
    4. Assignment1 ready to be cloned? (4 points, 4 comments)
  37. 25 points, 3 submissions: Spareo
    1. Submit Assignments Function (OS X/Linux) (15 points, 6 comments)
    2. Quantsoftware Site down? (8 points, 38 comments)
    3. ML4T_2017Spring folder on Buffet server?? (2 points, 5 comments)
  38. 24 points, 14 submissions: nelsongcg
    1. Is it realistic for us to try to build our own trading bot and profit? (6 points, 21 comments)
    2. Is the risk free rate zero for any country? (3 points, 7 comments)
    3. Models and black swans - discussion (3 points, 0 comments)
    4. Normal distribution assumption for options pricing (2 points, 3 comments)
    5. Technical analysis for cryptocurrency market? (2 points, 4 comments)
    6. A counter argument to models by Nassim Taleb (1 point, 0 comments)
    7. Are we demandas to use the sample for part 1? (1 point, 1 comment)
    8. Benchmark for "trusting" your trading algorithm (1 point, 5 comments)
    9. Don't these two statements on the project description contradict each other? (1 point, 2 comments)
    10. Forgot my TA (1 point, 6 comments)
  39. 24 points, 11 submissions: nurobezede
    1. Best way to obtain survivor bias free stock data (8 points, 1 comment)
    2. Please confirm Midterm is from October 13-16 online with proctortrack. (5 points, 2 comments)
    3. Are these DTlearner Corr values good? (2 points, 6 comments)
    4. Testing gen_data.py (2 points, 3 comments)
    5. BagLearner of Baglearners says 'Object is not callable' (1 point, 8 comments)
    6. DTlearner training RMSE none zero but almost there (1 point, 2 comments)
    7. How to submit analysis using git and confirm it? (1 point, 2 comments)
    8. Passing kwargs to learners in a BagLearner (1 point, 5 comments)
    9. Sampling for bagging tree (1 point, 8 comments)
    10. code failing the 18th test with grade_learners.py (1 point, 6 comments)
  40. 24 points, 4 submissions: AeroZach
    1. questions about how to build a machine learning system that's going to work well in a real market (12 points, 6 comments)
    2. Survivor Bias Free Data (7 points, 5 comments)
    3. Genetic Algorithms for Feature selection (3 points, 5 comments)
    4. How far back can you train? (2 points, 2 comments)
  41. 23 points, 9 submissions: vsrinath6
    1. Participation check #3 - Haven't seen it yet (5 points, 5 comments)
    2. What are the tasks for this week? (5 points, 12 comments)
    3. No projects until after the mid-term? (4 points, 5 comments)
    4. Format / Syllabus for the exams (2 points, 3 comments)
    5. Has there been a Participation check #4? (2 points, 8 comments)
    6. Project 3 not visible on T-Square (2 points, 3 comments)
    7. Assess learners - do we need to check is method implemented for BagLearner? (1 point, 4 comments)
    8. Correct number of days reported in the dataframe (should be the number of trading days between the start date and end date, inclusive). (1 point, 0 comments)
    9. RuntimeError: Invalid DISPLAY variable (1 point, 2 comments)
  42. 23 points, 8 submissions: nick_algorithm
    1. Help with getting Average Daily Return Right (6 points, 7 comments)
    2. Hint for args argument in scipy minimize (5 points, 2 comments)
    3. How do you make money off of highly volatile (high SDDR) stocks? (4 points, 5 comments)
    4. Can We Use Code Obtained from Class To Make Money without Fear of Being Sued (3 points, 6 comments)
    5. Is the Std for Bollinger Bands calculated over the same timespan of the Moving Average? (2 points, 2 comments)
    6. Can't run grade_learners.py but I'm not doing anything different from the last assignment (?) (1 point, 5 comments)
    7. How to determine value at terminal node of tree? (1 point, 1 comment)
    8. Is there a way to get Reddit announcements piped to email (or have a subsequent T-Square announcement published simultaneously) (1 point, 2 comments)
  43. 23 points, 1 submission: gong6
    1. Is manual strategy ready? (23 points, 6 comments)
  44. 21 points, 6 submissions: amchang87
    1. Reason for public reddit? (6 points, 4 comments)
    2. Manual Strategy - 21 day holding Period (4 points, 12 comments)
    3. Sharpe Ratio (4 points, 6 comments)
    4. Manual Strategy - No Position? (3 points, 3 comments)
    5. ML / Manual Trader Performance (2 points, 0 comments)
    6. T-Square Submission Missing? (2 points, 3 comments)
  45. 21 points, 6 submissions: fall2017_ml4t_cs_god
    1. PSA: When typing in code, please use 'formatting help' to see how to make the code read cleaner. (8 points, 2 comments)
    2. Why do Bollinger Bands use 2 standard deviations? (5 points, 20 comments)
    3. How do I log into the [email protected]? (3 points, 1 comment)
    4. Is midterm 2 cumulative? (2 points, 3 comments)
    5. Where can we learn about options? (2 points, 2 comments)
    6. How do you calculate the analysis statistics for bps and manual strategy? (1 point, 1 comment)
  46. 21 points, 5 submissions: Jmitchell83
    1. Manual Strategy Grades (12 points, 9 comments)
    2. two-factor (3 points, 6 comments)
    3. Free to use volume? (2 points, 1 comment)
    4. Is MC1-Project-1 different than assess_portfolio? (2 points, 2 comments)
    5. Online Participation Checks (2 points, 4 comments)
  47. 21 points, 5 submissions: Sergei_B
    1. Do we need to worry about missing data for Asset Portfolio? (14 points, 13 comments)
    2. How do you get data from yahoo in panda? the sample old code is below: (2 points, 3 comments)
    3. How to fix import pandas as pd ImportError: No module named pandas? (2 points, 4 comments)
    4. Python Practice exam Question 48 (2 points, 2 comments)
    5. Mac: "virtualenv : command not found" (1 point, 2 comments)
  48. 21 points, 3 submissions: mharrow3
    1. First time reddit user .. (17 points, 37 comments)
    2. Course errors/types (2 points, 2 comments)
    3. Install course software on macOS using Vagrant .. (2 points, 0 comments)
  49. 20 points, 9 submissions: iceguyvn
    1. Manual strategy implementation for future projects (4 points, 15 comments)
    2. Help with correlation calculation (3 points, 15 comments)
    3. Help! maximum recursion depth exceeded (3 points, 10 comments)
    4. Help: how to index by date? (2 points, 4 comments)
    5. How to attach a 1D array to a 2D array? (2 points, 2 comments)
    6. How to set a single cell in a 2D DataFrame? (2 points, 4 comments)
    7. Next assignment after marketsim? (2 points, 4 comments)
    8. Pythonic way to detect the first row? (1 point, 6 comments)
    9. Questions regarding seed (1 point, 1 comment)
  50. 20 points, 3 submissions: JetsonDavis
    1. Push back assignment 3? (10 points, 14 comments)
    2. Final project (9 points, 3 comments)
    3. Numpy versions (1 point, 2 comments)
  51. 20 points, 2 submissions: pharmerino
    1. assess_portfolio test cases (16 points, 88 comments)
    2. ML4T Assignments (4 points, 6 comments)

Top Commenters

  1. tuckerbalch (2296 points, 1185 comments)
  2. davebyrd (1033 points, 466 comments)
  3. yokh_cs7646 (320 points, 177 comments)
  4. rgraziano3 (266 points, 147 comments)
  5. j0shj0nes (264 points, 148 comments)
  6. i__want__piazza (236 points, 127 comments)
  7. swamijay (227 points, 116 comments)
  8. _ant0n_ (205 points, 149 comments)
  9. ml4tstudent (204 points, 117 comments)
  10. gatechben (179 points, 107 comments)
  11. BNielson (176 points, 108 comments)
  12. jameschanx (176 points, 94 comments)
  13. Artmageddon (167 points, 83 comments)
  14. htrajan (162 points, 81 comments)
  15. boyko11 (154 points, 99 comments)
  16. alyssa_p_hacker (146 points, 80 comments)
  17. log_base_pi (141 points, 80 comments)
  18. Ran__Ran (139 points, 99 comments)
  19. johnsmarion (136 points, 86 comments)
  20. jgorman30_gatech (135 points, 102 comments)
  21. dyllll (125 points, 91 comments)
  22. MikeLachmayr (123 points, 95 comments)
  23. awhoof (113 points, 72 comments)
  24. SharjeelHanif (106 points, 59 comments)
  25. larrva (101 points, 69 comments)
  26. augustinius (100 points, 52 comments)
  27. oimesbcs (99 points, 67 comments)
  28. vansh21k (98 points, 62 comments)
  29. W1redgh0st (97 points, 70 comments)
  30. ybai67 (96 points, 41 comments)
  31. JuanCarlosKuriPinto (95 points, 54 comments)
  32. acschwabe (93 points, 58 comments)
  33. pharmerino (92 points, 47 comments)
  34. jgeiger (91 points, 28 comments)
  35. Zapurza (88 points, 70 comments)
  36. jyoms (87 points, 55 comments)
  37. omscs_zenan (87 points, 44 comments)
  38. nurobezede (85 points, 64 comments)
  39. BelaZhu (83 points, 50 comments)
  40. jason_gt (82 points, 36 comments)
  41. shuang379 (81 points, 64 comments)
  42. ggatech (81 points, 51 comments)
  43. nitinkodial_gatech (78 points, 59 comments)
  44. harshsikka123 (77 points, 55 comments)
  45. bkeenan7 (76 points, 49 comments)
  46. moxyll (76 points, 32 comments)
  47. nelsongcg (75 points, 53 comments)
  48. nickzelei (75 points, 41 comments)
  49. hunter2omscs (74 points, 29 comments)
  50. pointblank41 (73 points, 36 comments)
  51. zheweisun (66 points, 48 comments)
  52. bs_123 (66 points, 36 comments)
  53. storytimeuva (66 points, 36 comments)
  54. sva6 (66 points, 31 comments)
  55. bhrolenok (66 points, 27 comments)
  56. lingkaizuo (63 points, 46 comments)
  57. Marvel_this (62 points, 36 comments)
  58. agifft3_omscs (62 points, 35 comments)
  59. ssung40 (61 points, 47 comments)
  60. amchang87 (61 points, 32 comments)
  61. joshuak_gatech (61 points, 30 comments)
  62. fall2017_ml4t_cs_god (60 points, 50 comments)
  63. ccrouch8 (60 points, 45 comments)
  64. nick_algorithm (60 points, 29 comments)
  65. JetsonDavis (59 points, 35 comments)
  66. yjacket103 (58 points, 36 comments)
  67. hilo260 (58 points, 29 comments)
  68. coolwhip1234 (58 points, 15 comments)
  69. chvbs2000 (57 points, 49 comments)
  70. suman_paul (57 points, 29 comments)
  71. masterm (57 points, 23 comments)
  72. RolfKwakkelaar (55 points, 32 comments)
  73. rpb3 (55 points, 23 comments)
  74. venkatesh8 (54 points, 30 comments)
  75. omscs_avik (53 points, 37 comments)
  76. bman8810 (52 points, 31 comments)
  77. snladak (51 points, 31 comments)
  78. dfihn3 (50 points, 43 comments)
  79. mlcrypto (50 points, 32 comments)
  80. omscs-student (49 points, 26 comments)
  81. NellVega (48 points, 32 comments)
  82. booglespace (48 points, 23 comments)
  83. ccortner3 (48 points, 23 comments)
  84. caa5042 (47 points, 34 comments)
  85. gcalma3 (47 points, 25 comments)
  86. krushnatmore (44 points, 32 comments)
  87. sn_48 (43 points, 22 comments)
  88. thenewprofessional (43 points, 16 comments)
  89. urider (42 points, 33 comments)
  90. gatech-raleighite (42 points, 30 comments)
  91. chrisong2017 (41 points, 26 comments)
  92. ProudRamblinWreck (41 points, 24 comments)
  93. kramey8 (41 points, 24 comments)
  94. coderafk (40 points, 28 comments)
  95. niufen (40 points, 23 comments)
  96. tholladay3 (40 points, 23 comments)
  97. SaberCrunch (40 points, 22 comments)
  98. gnr11 (40 points, 21 comments)
  99. nadav3 (40 points, 18 comments)
  100. gt7431a (40 points, 16 comments)

Top Submissions

  1. [Project Questions] Unit Tests for assess_portfolio assignment by reyallan (58 points, 52 comments)
  2. [Project Questions] Unit Tests for optimize_something assignment by agifft3_omscs (53 points, 94 comments)
  3. Proper git workflow by jan-laszlo (43 points, 19 comments)
  4. Exam 2 Information by yokh_cs7646 (39 points, 40 comments)
  5. A little more on Pandas indexing/slicing ([] vs ix vs iloc vs loc) and numpy shapes by davebyrd (37 points, 10 comments)
  6. Project 1 Megathread (assess_portfolio) by davebyrd (34 points, 466 comments)
  7. defeat_learner test case by swamijay (34 points, 38 comments)
  8. Project 2 Megathread (optimize_something) by tuckerbalch (33 points, 475 comments)
  9. project 3 megathread (assess_learners) by tuckerbalch (27 points, 1130 comments)
  10. Deadline extension? by johannes_92 (26 points, 40 comments)

Top Comments

  1. 34 points: jgeiger's comment in QLearning Robot project megathread
  2. 31 points: coolwhip1234's comment in QLearning Robot project megathread
  3. 30 points: tuckerbalch's comment in Why Professor is usually late for class?
  4. 23 points: davebyrd's comment in Deadline extension?
  5. 20 points: jason_gt's comment in What would be a good quiz question regarding The Big Short?
  6. 19 points: yokh_cs7646's comment in For online students: Participation check #2
  7. 17 points: i__want__piazza's comment in project 3 megathread (assess_learners)
  8. 17 points: nathakhanh2's comment in Project 2 Megathread (optimize_something)
  9. 17 points: pharmerino's comment in Midterm study Megathread
  10. 17 points: tuckerbalch's comment in Midterm grades posted to T-Square
Generated with BBoe's Subreddit Stats (Donate)
submitted by subreddit_stats to subreddit_stats [link] [comments]

DOWNLOAD FREE Pips Wizard Pro generate $5,000 $8,000 $13,000 daily

Pips Wizard Pro Review – Lots of the reviews created here might undoubtedly be of excellent assist to you within your efforts to remain away from scams! This Pips Wizard Pro might supply you with information on what it is, everything you ought to anticipate as well as the way it works out. Our own website is knowledgeable and skilled critiques and we’ve got acquired all sided reviews on Pips Wizard Pro for you to get any detailed crucial details. Our work is to aid consumers such as you who have confidence in us and regard our responsibility to create a smart selection. Each of the evaluations published here might certainly be of great help for you inside your time to remain away from scams! The secret approach will decrease your working period of time in half and obtain better final results!
Features of Pips Wizard Pro:
*Pips Wizard Pro is the most accurate indicator of the point, is so profitable in every possible aspect. *This unique tool “, an independent broker,” and thus more enjoyable trading experience and now any broker can use your discretion. *A new, unique and amazingly lucrative Pips Wizard Pro with index points, you now have a very simple and user-friendly way to be able to track the movement of any future in the market. It is a miracle to be able to turn your life around! *You do not need to see the need to continue in the market. Now you have your e-mail and / or mobile phone right from the time of trading can get immediate notification. You always know when a new trading signal! *It focuses on the logical stop-loss account – a few points away from the price – but not so strong that reducing your risk and increases the chance of profit and stop loss.
How To Use Pips Wizard Pro
Installing the system is very simple, you will simply place the file into your MetaTrader 4 and start dropping it on the charts. Using it however, isn’t as simple.
One of the most difficult elements when dealing with Karl Dittman products is figuring out how to use it properly. While the sales page makes it seem like you can purchase the software and start profiting immediately, this isn’t truly the case. In order to succeed using a Forex indicator with no specific guidelines like this one traders need to develop their own strategy.
The Pips Wizard Pro has the potential of being a great software, but the trader plays a large role in determining this. Since the system can work on any time frame from M15 to D1 and works with every single pair available there are a plenitude of options. To truly win long-term with a software like this traders will have to determine which pairs and what timeframes work well together over a long sample size.
DOWNLOAD PIPS WIZARD PRO NOW!!!
It is entirely possible to make a profit on Forex trading. However, most successful Forex traders spend an inordinate amount of time studying and following the markets, and usually do not rely upon an automated trade signaling software to provide guidance on which currency pairs are destined to gain value versus which are going do lose value.
This is a common criticism of trade signal programs; markets are simply too complex and volatile to be predicted by current algorithms, no matter claims to their accuracy.
Many other signaling software programs are available over the internet – some of which are free, provided you use a specific broker, while others need to be purchased – and almost all of them offer outlandish (and unsubstantiated) claims of accuracy and success.
Pips Wizard Pro is, sadly, no different. There's no indications that this Forex trade signal algorithm is superior to any other out there.
While impressive screenshots may show highly accurate predictions supposedly made by the software, screenshots can and have been faked, either through running the program in demo mode, or through outright image manipulation.
If Pips Wizard Pro was being offered for sale from a reputable source, this might change our tune, but Karl Dittman is anything but reputable; in fact the truth is that there is no “Karl Dittman.” The name is a pseudonym that has been used for years – a fact that is buried deep in the Disclaimer boilerplate of the Pips Wizard Pro website.
Pips Wizard Pro review is going to be scored one of the highest quality products in this region. The income and profits can certainly be fairly significant, that document precisely how effectively-needed Pips Wizard Pro Pdf file is not to mention simply how much it is actually merely staying needed in this particular program customers. Nevertheless, you can expect entirely cash back guarantee pertaining to this program recommendation legality, there is undoubtedly however no returning rate. If you feel you want an most up-to-date release of Pips Wizard Pro testimonial, it is effortless to get it because of this web site website. We offer excellent deals in addition to quite a lot of benefits to enhance the program. Added to that, our via the internet topic site regularly continues dynamic. From the moment you get access to our correlate site, it is achievable to visit experience reassured, you will find adjacent to practically nothing complicated. Mastering or developing the most effective from get the program critiques is surely an important please for any who website link on top of us. When the terrific quantity of factors included in this particular system is a one that frightens you against seeking Pips Wizard Pro, you actually are totally incorrectly diagnosed frequently. This system legitimate requires joy in the impressive good popularity currently in the marketplace. If to start with it breaks down to fulfill the needs you might have, simply send in a return request for and make it easy for item turning into a story of history.
How Does Pips Wizard Pro Works?
Pips Wizard Pro is one of a kind and once you experience its simple, powerful functionality and amazingly unique features will blow you away. It is the new most powerful technology is being with the unique indicator tool accurately predict all price movements. All you need to sit and wait while Pips Wizard Pro analyzes the markets for you and gives you select signals. It also alerts you when a signal is occurring, so you do not even need to sit in front of a computer in which it does great as an income side.
With Pips Wizard Pro Indicator there is no reason to learn anything. This is another great benefit as this indicator can be used by all traders, novices, and professionals. It does not matter what you know or does not know about the currency. This indicator tool makes accurate predictions of market movement before they even occur. It works on M15-D1 in all currency pairs where you will be able to make correct trading decisions based on the flag signal lines drawn right on the chart. This is a new trade sign and job alerts for all deadlines from M15 to D1. In all Forex pairs. Here’s how the indicator works:
*Step 1: Open a currency pair or stock you would like to trade.
*Step 2: Choose your calendar: M15, M30, 1H, 4H or D1.
*Step 3: Whenever you receive a pop-up alert with buy or sell signals
*Step 4: The exit point and stop loss will be printed in the alert window!
VISIT PIPS WIZARD PRO OFFICIAL WEBSITE
*All you need is Pc with an internet connection and this system for trading. So, without an internet connection, you can’t get the best result at desired time. Pips Wizard Pro Guide
*It doesn’t make any promise to make you rich at overnight, it takes a little time to increase your income level. Pi
submitted by tranlinhphuong1 to PipsWizardProFree [link] [comments]

How to use Currency Correlation CORRECTLY (tools and live ... Best Forex Charting Software Free - YouTube Using the Currency Strength Meter to Master Correlation in ... How to Compare Two Charts at Once (Forex Correlation ... Correlation Matrix & Trader - Pepperstone Forex Correlation Chart Dashboard Using Correlation in Forex Trading by Adam Khoo - YouTube

In diesem Artikel möchte ich einmal näher auf die Korrelationen im Forex eingehen. Diejenigen, die unsere FOREX Daytrading Signale traden, wissen, dass mein Ansatz ca. 40 % Korrelationen beinhaltet und versucht, von den Abhängigkeiten der einzelnen Devisenpaare untereinander zu profitieren. Die restlichen 40 % sind Intraday Fundamentals und ca. 20 % ist Charttechnik. Forex Correlation shows, within some period two currency pairs move in the same direction or opposite direction or random direction. Correlation is a measurement of the relationship between two currency pairs. This Forex correlation indicator will show the entry point & exit point to trade. You can combine other analyses for further clarification. Currency correlation is the process of looking at relationships between currencies. We show you how to find the correlations and how to use the data. Home; Expert Advisors; Best Forex EAs; Forex Brokers; Forex Signals; Forex Trading; Analytical Tools. Charting Software ; No Result. View All Result. Home; Expert Advisors; Best Forex EAs; Forex Brokers; Forex Signals; Forex Trading; Analytical ... Forex Correlation. The following tables represent the correlation between the various parities of the foreign exchange market. The charts give precise details on the correlation between two parities. They show the history and the distribution of the correlation over a given period. The Forex Correlations Table displays relationships in the data from the Open Positions module that you can explore interactively! The correlation is very low and it is colored in Red if its value is lower than -60%. This Forex correlation pivot table is based on the correlation of the last 30 trading days. Of course, you can update the formula and measure the correlation using another period (90 days for example) or in another time frame (example: hourly or 1-minute). The system is based on the correlation table below, and is in fact quite simple to understand and implement. The aim of this thread is to fine-tune the settings and develop a simple EA to automate the trading. As you can see in the correlation table, certain pairs overreact to a counter-correlation price movement of two other correlated pairs, and always do it in the same direction. I´ve ...

[index] [12815] [23512] [23393] [2885] [13354] [7002] [24642] [25956] [21993] [29790]

How to use Currency Correlation CORRECTLY (tools and live ...

Hello traders! Glad to see you here! In this video, I show you how I personally implement Forex Correlation into my trading. I always look for the correlatio... Best Forex Charting Software Free - http://tradewithjagger.com/ This is the charting software I use for all my forex trading analysis. In this webinar, Navin will be go over how the Urban Forex Strength Meter can be used to help you use Correlation with your Forex trading. Correlation can be... Get more tips: http://www.tradingheroes.com/blog/ The economies of certain countries are highly correlated to commodity markets. For example, the Canadian ec... When currency pairs are correlated, it increases the probability of identifying winning forex trading setups. These are essential Forex trading strategies fo... How To Create Your Own Forex Correlation Matrix in Excel - Duration: 15:30. TRADE INC 13,215 views. 15:30. FOREX CORRELATION ANALYSIS METHOD (The One And The Many) - Duration: 16:14. ... Watch this video on how to setup your charts using the DX (Dollar Index Futures Chart) as the anchor for optimal technical analysis to analyze market trading environments. Category Education

http://binary-optiontrade.tocardode.tk