Win Rate and Drawdown

megamuel

Experienced member
Messages
1,098
Likes
164
If you were testing a trading system with a 1:1 R:R (for the sake of this discussion), what is the minimum win rate you would consider before trading it live? Also what is the maximum drawdown you would be comfortable with? What other statistics do you take in to consideration when testing a system and which are the most important?

By the way, I've been doing a lot of back testing lately and because I'm using multiple custom indicators it is very slow. If anyone wants a tip on how to speed up backtesting, here it is - Make sure your broker platform is open, press ctrl+alt+del and go to your task manager, scroll down and find 'terminal.exe', right click and go to 'Set Priority' and select 'high' and hey presto! Your processor should allow more....stuff.... to Metatrader! You can select 'real-time' but it can make your system unstable and everything else run majorly slow. I'm guessing you can do the same for other platforms also.
 
If you were testing a trading system with a 1:1 R:R (for the sake of this discussion), what is the minimum win rate you would consider before trading it live?

think for me, it depends on number of trades per month and account size etc
 
How about taking into account the forward testing statistics, more than the backtesting stats, before considering it good enough to go live?

Consider this. Whenever I have tried to backtest a system on a set of data, I could easily find a system with a win rate of 60% for 1:1 Risk:Reward. I could find with a little more difficulty, a win rate of 70% because of curve fitting. Depending on the data, I may even be able to find a win rate of 80% (although this would be a fair bit tougher), yet it would still probably be a curve fitted losing system.
 
I agree backtesting has its flaws and is not completely reliable but I think it is a good starting point before forward testing. You can look at it another way if you like - what is the minimum you would consider before forward testing?

Sam.
 
If you think you've got something, forward test it (without money). Why wait? Just demo and see. Backtested results, have little bearing on the forward tested results (unless you've used techniques to eliminate curve fitting etc)
 
You have to be careful even with forward testing without real money. The only TRUE way to test a system is with real money. I have spent thousands of dollars to find the right ones, but it's worth it once you get it going correctly.

For a 1:1 ratio I would want at least 75% win ratio. Test at least 25 live trades before you decide if it will work or not. If it's a scalper system then at least 45.
 
With a 60% win rate and 5% risk (theoretically 20% risk be would be optimal) assuming no moving SL and reinvesting profits, after 60 trades you can expect account to have grown by 69%
 
Last edited:
to earn a wage from?, or to not withdraw profits and grow the account, over what time period?




I guess expected payoff = (probabilithy of win)*reward - (probability of loss)*risk

How about an answer for both - to make a wage from and to compound until there is enough to start taking a wage from.

I got 15.99 for expected payoff, anyone know how that compares?

As for curve fitting, I guess that would suggest I tweak the parameters for optimal results. In fact, the more tweaking I've done the worse the results have got! The original settings seems best.

Sam.
 
Ah so is 15.99 my average profit per trade? That's pretty poor. Testing was done starting at £6k in this example...
 
This is easier!
 

Attachments

  • backtest.jpg
    backtest.jpg
    96.6 KB · Views: 279
"expected payoff" is the amount you can expect to win (or lose if it's negative) PER TRADE. That takes into account wins and losses over the entire range of testing. You should be able to multiply the expected payoff and the total trades and it will equal the total net profit.

In your example, if you complete 5 trades per month with an expected payoff of 15.99 the the testing results show your monthly theoretical expected profit is 15.99 x 5 = 79.95

As you have stated this is a starting point. Don't use back testing results as entirely reliable.

Peter
 
Hotch, that is the average consecutive loss. It had a max of 6 consecutive losers. I will share the EA at some point I just want to refine it and test it a bit more first. Thanks for the link, I'll look into that when I get back from work.

Pete, thanks for that description - makes sense now. Obviously that isn't a great expected pay off but the idea is to do some testing on other pairs/instruments and hopefully it will have similar results on them. If so, I'll forward test on maybe 5+ different pairs/instruments and see how that goes.

Sam.
 
Yeah, I twigged on second look. Good luck with it, personally I'd suggest testing it on a tiny account, I always miss something which means it's fine in backtesting and then when I want to go live I need to change a bunch of things.
 
With a 60% win rate and 5% risk (theoretically 20% risk be would be optimal) assuming no moving SL and reinvesting profits, after 60 trades you can expect account to have grown by 0.69%

i meant 69%


is the strategy backtester automatic?
you demoing it first?
 
Top