How do you gain, and measure, your trading systems statistical edge?

[Llolegeagua2012fomex]

Well... not heard much on this, will be interesting... it's normally the hunt for the holy grail...

[kmasnoyamsk]

Well done Davidee, good to see a few clear uncluttered BS complimentary believing. Perhaps not a holy grail guess121, just profitable systems. They usually exist. Just not sold as EAs that is all.

I'll be interested to devote a couple of hours see if I get similar results backtested. In reply to the original question, I use 2 things:
1) On entire trade (using same procedure of exit as entry), workout maximum drawdown: maximum profit possible. Typically I can get around 0.7:1 MD:.
Two ) a 1:1 ratio in multiple values - from around 50 to a couple of hundred pips, eg 100p TP and 100p SL. System should be profitable on aggregate in multiple values. Clearly I'm working on shorter timescales than you.

Interestingly though, just as max profit possible gt; max drawdown, doesn't always result in a profitable system as determined by 2); sometimes it only puts you on the ideal side of a commerce and the TPs become a different situation.

[oxmialb]

There is something missing from that calculation: the biggest drawdown in pips or better, in percentage of the account if a trader would risk 1% on every trade. Anytime I look at a profitable system is not the reward that I am interesting in however the drawdown.
I prefer the calculation in percentage because the volatility is not continuous and 100 pips today for EURUSD aren't the same 100 pips one year past.
These facts make the exact calculation of the border very difficult. Rather than some number which is moot I would prefer to find the profit curve.

LE. Triphop, it seems that we share the same ideas.

[kmasnoyamsk]

I favor the calculation in percentage since the volatility is not continuous and 100 pips now for EURUSD aren't the same 100 pips annually ago. Very true Sauron; these are the steps to see if what I am looking at is statistically significant before I put it to some system. Then bring on the profit stinks...
I do it this way since I've screwed up calculations too frequently and to attempt to avoid curve fitting, I like to strip it down and work out if the inputs I'm using are significant or not until system modelling. Ie is it an edge or a load of cobblers.

[camvany]

I use this http://en.wikipedia.org/wiki/Student#39;s_t-test#Independent_one-sample_t-test.
If you hunt though a number of my posts, I bang on about it someplace else if you are interested.

[salva1984]

There is something missing from that calculation: the largest drawdown in pips or greater, in percentage of the account when a trader would risk 1% on every transaction. Anytime I look in a profitable system isn't the reward that I'm interesting in but the drawdown.
I favor the calculation in percentage because the volatility isn't constant and 100 pips today for EURUSD are not the exact same 100 pips one year ago.
These facts make the exact calculation of the border very hard. Rather than some number that's meaningless I would prefer to see... we've got a winner.

Which are the profits month .

More importantly... what was that the draw down during that interval?

m

[Giuseppeypunpk]

1 Attachment(s) Thank you for all of the replies.

This is not a comprehensive trading system, instead it's a method I'm using to quantify the edge of entries and exits.

To be prosperous in a zero sum game such as the Forex market you need to get an advantage substantial enough to give you a profit following the expenses of trading. What I'm asking is -

1. What's your edge?
And 2. How do you know it an advantage? How large an advantage is it? And how do you quantify it?

And as this board is about sharing info I've included a link to my website to tell you about the way I quantify my advantage. This article was not supposed to form a whole trading system...

But I'll answer the query and attach the full outcomes.

I've tested this on the EUR/USD currency pair just - when I'm creating a system I just use one currency pair and just ever test it on several currency pairs once I've finished since this helps me avoid subconsciously curve fitting. I really get a truer image to the system's value after I test it out on additional pairs not included in the practice of earning it, and that I just do so once I've finished.

About the EUR/USD, using data from MetaTrader from the start of 2000 to the end of 2009, the biggest draw is 2864 pips along with the biggest winning steak is 7974 pips. This comes came from trading the subsequent system free of stop losses -

1. After the price closes lower than it's ever closed in 80 days go short, once the price closes greater than it's ever closed in 80 days ahead.

2. Keep all trades on for just 20 days.

That is before I try and squeeze another advantage out of this exit, which is really a lot tougher to perform.

Much like additional trend following systems, the huge losses usually occur when the trend is ending. After the market is making higher highs or lower lows we are adding rankings, and the longer the trend continues the more profitable the system. But the trend can not continue forever so the trades we create last, prior to the trend ends, will usually be winners. Giving back some of your profits is an essential part of trend following as you are basically holding until your system receives a sign that the market continues to grow and eventually gone against you - and you do not get this sign until after the event. Incidentally, the biggest losing streak came following the biggest winning series...

The only biggest losing trade was 1800 pips in the conclusion of this winning series, stop losses usually degrade the operation of a technical trading system, however in this instance perhaps they could have been used to bring an edge to the system also?
https://www.nigeriaforextrading.com/trading-system-and-egies/105-trading-dilemma.html

[oxmialb]

I would use a stoploss = 2 * ATR or longer.
If you don't utilize a SL then you can't actually say how profitable your system is.
Risk for example 2 percent per transaction and correct your lot size based on SL. Doing this you will see that the number of pips has really no significance.

[Giuseppeypunpk]

In case the EUR/USD is at, state 1.3500, then a win of 7974 pips is roughly 59 percent of the account along with a loss of 2864 pips is a loss of about 21% pips with 1:1 leverage.

[salvasafom]

If the EUR/USD is at, say 1.3500, then a win of 7974 pips is roughly 59 percent of their account and a reduction of 2864 pips is a reduction of approximately 21% pips with 1:1 leverage. Can you conduct some kind of analysis to come up with the 80 day and 20 afternoon parameters?

[salvasafom]

Https://www.nigeriaforextrading.com/attachments/15193627331716561145.png

A frequency distribution is the way I decided the 2-Bar high low breakout had an advantage. It's measuring the most amount of pips possible. Draw down is not taken into account.

[Giuseppeypunpk]

Can you run some kind of analysis to think of the 80 day and 20 day parameters?
Not really, no. I'd love to point out again that the tests for the edges I mentioned above aren't intended, by themselves, to form a comprehensive trading platform. The tests were there only to establish or disprove the existence of statistically significant advantage which is big enough to be tradeable.

The edge(s) would be the first building blocks in my personal trading systems. I work on getting an edge long until I look into money management or anything else as without an edge no trading platform will ever be profitable.


Anyway, you asked about why I choose 80 days to get a breakout... What I have discovered is this -
The longer the trading time period the more important the breakouts, but this is only accurate up to around 120 days where it begins becoming so long past it's no longer relevant. The worst time frames would be the ones most commonly employed, for example 20 days is probably worse than 18 days etc. 50 days and 55 days are also fairly bad since they're quite common too. It is only after about 60 - 70 days that the benefits of using a long term breakout become very important. Therefore any breakout interval between approximately 70 days and 120 days is probably a great one to use to get a long term trend following system. I just picked 80 out of the air for the remainder of that report.

What else can I do? I meanI could have analyzed daily between 70 and 120 days on every pair and come to the conclusion that, say, 91 days was greatest for its EUR/USD, 102 days was greatest for its GBP/USD etc.. . But that would be'curve fitting' and just lead to great historic back-tested results that my system would be not able to duplie later on since I had accidentally'fitted' the machine to carefully into the past - at the markets history does repeat itself, but never exactly.

Why 20 days for an exit? 2 main reasons -

Firstly, I couldn't use 2 or 3 days as 80 day channel breakouts are moderate to long term programs.

And second, after around 15 days the edge really begins to work as that tends to be enough time for it to take effect. No matter how the edge tends to lose it's significance after around 30 days. Look at the outcomes of a 20 day time based exit vs a 40 day time based exit - that the losses are a lot bigger but the advantage barely moves accordingly that the losses begin to grow closer to the magnitude of the winnings. So 20 days because anything between 15 and 30 days will most likely do.

I've a system which employs a similar procedure to those in the content in trending markets and another entirely different way in choppy markets; the outcomes are a lot better than the results of the test for an edge. It wins around 70 percent of the time. I'll publish the comprehensive system on this forum and here http://www.myforexdot.org.uk/FreeTechnicalTradingSystems.html once I have the time. But in this thread I was actually just Searching for people to share their edges and their methods for testing the significance of those edges, so thanks to your contribution

[salvasafom]

And thank you.

[bypoxa]

I find this thread very interesting, due to the use of Donchian breakouts. I use them almost exactly as you can....

So after reading your blog, I attempted to replie the evaluation with my Metatrader. I did not get any 30,000 pips profit over 9 years.... It was more like 6,000, based on the 80 day entry if the price closed at a new high or low. The exit has been 20 days.

So I scratched my head, wondering what your chart looks like on your backtests.


Anyhow, I am satisfied with my research of utilizing Donchian successfully (I used a 40 afternoon breakout myself) and my exits are much less complied that what you're constructing. But I applaud your attempts... It was enjoyable to see your study.




I'd love to add my 2 cents, do with it what you will:

The EURUSD has had a powerful trending characteristic as it's beginning. USDJPY, GBPUSD, and USDCHF do not behave the same (well, USDCHF is horribly close) so that you can't anticipate your algorithm to function across multiple pairs. Don't beat yourself up on it if your EURUSD algorithm fails on additional pairs. By way of instance, GBPUSD is simply too choppy, and Donchian breakouts are filled with false signs there.

[oxmialb]

My evaluations actually showed 39000 pips from 2002 to 2009 with no stoploss. Also 80 and 100 days were the best choises, with fewer and fewer pips for 60, 40 or 20.
EURUSD appears to be the fantastic winner using Donchian, another pairs acting less succesfully. Risking 1% for each 200 pips you could have multiplied 6-7 times your account, not bad at all considering the biggest drawdown would have been 20%. No brainer system, my dog can comprehend it.
I did not anticipate EURUSD are the very trending pair. Nothing guarantees it will remain like that but you can use all 4 majors.

[Giuseppeypunpk]

I find this thread very interesting, due to the use of Donchian breakouts.... Thanks for your reply,

Are you sure you've back-tested exactly the same system I am using rather than a'normal' Donchian breakout? What I mean is everytime the breakout creates a higher high or lower do you add another place? Even when there are different positions that have not closed due to being around for under 20 days?

The reason I ask is I buy approximately 6,000 pips when I input only after when the channel is originally broken, but of course the losses are a lot smaller then also. Does your system'pyramid', including more lots to the open positions during a long-lasting fashion?

I am using metatrader data and I have included the complete results from 2000.01.03 - 2009.12.31 in the reply in post number 8 of this thread in the attached document.

The figures are, the date, the highest closing price in the previous 80 days, the lowest closing price in the previous 80 days, today's closing price, the closing price 20 days later on, GoLongFlag - a 1 if now closed over the highest closing price in the last 80 days to indicate that we place on another long position, GoShortFlag - a 1 if now closed below the lowest closing price in the last 80 day to indicate that we place on another short place, and the result.

The outcomes over this time period are 29020 pips lost and 61016.1

Could you compare my closing prices together with all the closing prices on your information so we can get to the bottom of this.

[Giuseppeypunpk]

My evaluations actually showed 39000 pips from 2002 to 2009 with no stoploss. Also 80 and 100 days were the best choises, with fewer and fewer pips for 60,... That is how I feel about it also, 6 occasions my account more than 7 years is something that I would be quite happy with - especially at 1% for each 200 pips which will be leverage of less than 1:1.

When I first started trading I started out spread betting, aiming for profits of 3 percent a day compounded and other ridiculous figures like that. The end result was that I lost my whole account, put more cash in my account, then just dropped it all again...

I believe one of those lessons traders need to learn is to know the market will and will not give you, rather than to aim for anything greater than that. I like systems in this way because they convince me that 30% a year could be done, and I've learned to be happy with this.

So many traders aren't happy with results like that thoughthey want their $5000 account to become $5,000,000 within 18 months and what are you, but they can't mathematically demone the existence of a statistical edge big enough to create these profits that they are aiming for - that I suspect these traders are where a great deal of Forex'failures' come from.

[lcasqwl]

So many traders are not happy with results like that thoughthey want their $5000 account to become $5,000,000 within 18 months and what are you, but they can not mathematically demone the existence of a statistical edge large enough to create these profits they are aiming for - I guess those traders are where a large amount of Forex'failures' come from.

[lheneo]

19 posts rather than just one mention of risk adjusted return? Pathetic what this place is now...

[lgpkni]

19 articles and not just one mention of risk adjusted return? Pathetic this place has become... https://www.nigeriaforextrading.com/trading-system-and-egies/106-dumping-audjpy-yen-pairs-worth.html

for anybody with an edge and who's also looking for the quality of risk adjusted return calculations.

[Giuseppeypunpk]

https://www.nigeriaforextrading.com/trading-discussion/76-charting-software.html

for anybody with a border and who's also searching for the quality of risk adjusted yield calculations. Thanks, I will use that if/when I attempt an incorporate this advantage into a complete trading system.

[oxmialb]

That formula is overly complied and doesn't make sense before you start trading. So long as I can backtest 10 years of data and I know the maximum drawdown and also the potential win then I will look if the 2 * maxDD and triumph / 2 are still ok for my taste. If I would start to exchange it I would consider 2 * maxDD are the maximum accepted loss.

[kmasnoyamsk]

19 posts and not just one mention of risk adjusted return? Pathetic what this place has become... Ds, that's bs. I don't say it softly because you're one of my favorite posters on this. However, there is a strategy the combination of an inefficiency - a border - and trading rules. If you jump straight into dd and rar before you've even shown if your edge is statistically valid then you'll muddy your stats straight off the bat.
Some advantages could be proven but do not readily fall into a system. Some systems are profitable for a certain amount of time but do not have a border - which means failure is around the corner.

This guy is asking if his border is valid, not if his strategy is worth investing. Very different Imo.

[ClauPevi]

Not really, no. I'd love to point out the evaluations are not intended, by themselves, to create a complete trading system. The evaluations were there only to establish or disprove the existence of statistically significant advantage which is large enough to be tradeable.

The OP has asked for a hand in determining the statistical advantage of systems or more setups since I see.

So, in this case the drawdown isn't a concern, can we see why?

As it is, a traders personal objectives are what will determine how much or to what level drawdown worries him. Like such as Acumen (whose work I respect btw) is looking for his drawdown or earn characters on a monthly basis and perhaps you are on a yearly.

If you are likely to reinvest profits and compound in order to attain maximum terminal wealth at the end of a specified time frame, then drawdown is essentially meaningless. Consider it, the number of heads in a row (or tails) in a 1000 tosses, the number of these streaks? In the conclusion when ur ev, then ur ev. Its the biggest loss which tells us exactly what we can afford to get away betting (optimally, for a few ) without getting our head handed to us. Basically drawdown as a stat is inherently not what traders must concentrate on, 10 losses in a row dosent mean ur process is completed (If you have value any merit to start with that is).

I understand that people will thing that everybody is not out for maximum wealth (and logically so), and I agree, but that is a function of each persons usefulness (which changes a lot), so we cant decide on almost any'one' utility curve. Someone could be a gunslinger.
Hence the systems advantage is different from drawdown. Largest reduction provides the narrative (but no stops so. .) .

The above is just my view.

[oxnwlsanchwzmivemo]

Can you maybe talk about the indior or what ever you use to calculate the border automatically? I really don't think you have completed it by hand or am I wrong?

[IMWIN92]

... I could have analyzed daily between 70 and 120 times on every pair and come to the conclusion that, state, 91 times was greatest for its EUR/USD, 102 times was greatest for its GBP/USD etc.. . But that would be'curve fitting' This statement is true but misses the purpose of why you should do it. If the machine was profitable utilizing 80 times but just half as profittable at 75 and 85 then you have by chance curve fitted. The same holds for the 20 day holding period rule.

The significance of the results to the input parameters is also an important factor in deciding if the there is in fact a strong system . A strong system will do the job reasonably well over a range of values. If it just worked for a few values I'd consider the results to be unrepresentative.

A previous post also made the point about this not working for other pairs. I would also examine it on an identical pair such as GBP/USD becuase if it same parameters did not operate reasonably for GBP it would be an additional suggestion the 80 day worth is a fluke.

[oxmialb]

We tend to use round numbers for simplicity, so I'd be surprised if anybody would begin his test at 83 days.
As the very best number of times is a wonderful instance of curve matching, a better idea is to have a fourth of profit aftter 60, 80, 100 and 120 days.

[Jokflan723]

I really don't think the statistical edge is an immediate constant. I think it changes given market conditions. I think you would have a hard time finding a system which maintains a constant edge over any significant amount of time.

Sometimes your system border may be high where as other times it immediately goes to DD.

Because markets change, so will any statistical edge with it.

Given that tho' I enjoy the conversation. I may have gleaned at least one new idea from it. Also any platform should be made so that you can live with it. If you take for a small 2 or 3 percent a week, but are satisfied with that then you may shave some DD off by not being overly greedy.

Whenever you don't be concerned about the extra 2 or 3 percent more that you might have gotten I think that is if you become a more confident trader.

[Jokflan723]

Ya know I had been thinking about what I get hung up on is that fixed point in time in the future and this over the last day or so.

I think the fixed point in time should not be a fixed point at all. I believe as the need be, that it should adapt. But I am not sure how to go about it just yet.

As a primary example if you have a fixed point in time and the time over laps a news event the outcomes may or may not be accurate as opposed to other less moving time periods.

Also looking back in other analysis I have done in the past will show action as to when gains were hit / closed out vs losses obtained etc.. A few days were more or less sterile with no action at all... Quite simply it varied.

So I think to properly forecast the edge you'd need some type of debate that that would change the predetermined time period as required, thus hoping to maintain the edge at all times.

[Giuseppeypunpk]

I first read about the edge ratio in the Novel Method of The. The advantage the trading system's entry provides is quantified using this technique by measuring how the price behaves within a definite period of time, beginning from the moment the transaction is taken until a fixed point in time in the future.

Basically, when the entrance criteria is fulfilled the amount of pips the market moves in favour of the transaction is split by the amount of pips the market moves against the transaction from the moment the transaction is entered before the fixed point in time in the future. It's this amount that offers the entry's edge ratio.

If discovered that 80-day Donchian channels generated a border ratio of approximately 1.40 to 1 over 20 days and this produced a system that had a ratio of around 2.10 to 1 when it arrived into the machine's pips won to pips lost ratio. This was improved to 2.29 to 1 when used with an exit that also had an advantage.

The complete results are here http://www.myforexdot.org.uk/MeasuringYourTradingSystemsEdge.html

So what is you trading system's advantage and how do you measure it?