Is 457 trades, traded over 16 years a good sample size? (Algo trading MT5)
1. I think consistent growth in your balance curve is the problem. There has been periods of several years without profit.
2. You should compare this to a normal investment in bond market for example. You would have earned more during 16 years if you invested in bonds.
3. 54% win rate with 1:1 RR ratio can easily lead to loss if there are glitches in real execution or small change in brokers' policy.
4. I dont know but you may have done optimization also and that could also trap you in over fitting issue.
All in all the number of trades in your sample is not that important with this case.
1. I think consistent growth in your balance curve is the problem. There has been periods of several years without profit.
2. You should compare this to a normal investment in bond market for example. You would have earned more during 16 years if you invested in bonds.
3. 54% win rate with 1:1 RR ratio can easily lead to loss if there are glitches in real execution or small change in brokers' policy.
4. I dont know but you may have done optimization also and that could also trap you in over fitting issue.
All in all the number of trades in your sample is not that important with this case.
These are unleveraged trades.
I'm creating a portfolio of algos and I'm purely looking at stability and consistency without using martingale etc. I'm banking on combining this with 20 other pairs using the same strategy as a portfolio, this should beat bond investments etc.
My main question is can you rely on that sample size... I haven't curve fitted it and have only optimized 5 parameters in one pass.
Also, this is an intraday strategy and it's hard to find consistent intraday strategies which are profitable over many decades.
1. Leverage is not an issue I think. You already have 7% drawdown and that is enough to know there is higher risk than bond investment.
2. If you get same results with 20 other pairs and all show some kind of profitability over a decade that is awesome.
3. Also note that running on several pairs will increase drawdown.
1. Leverage is not an issue I think. You already have 7% drawdown and that is enough to know there is higher risk than bond investment.
2. If you get same results with 20 other pairs and all show some kind of profitability over a decade that is awesome.
3. Also note that running on several pairs will increase drawdown.
Yep, I think the 7% drawdown might be offset by other traded pairs in the portfolio - they're uncorrelated from what I can see so they will be in drawdown at different times which should smoothen out the equity curve.
Your input is much appreciated Yashar.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Hello community,
[PLEASE SEE ATTACHED EQUITY CURVE]
I've searched this forum to work out if I should be focused on backtesting a large sample size of trades, or a smaller sample size backtested over a longer timeframe.
It's a case of 'large sample size' vs 'smaller sample size tested over long timeframes/market conditions' debate.
My trading system trades off the 5 minute timeframe and the backtest runs from 2007-2023 (16 years). We have 457 trades during this period, using a fixed lot size of 1.
Trading stats:
I've backtested on 'every tick based on real ticks' and have added the correct commission settings so everything is correctly set.
This system has a 54% win rate with a 1:1 risk to reward ratio. I haven't added any money management, trailing stop losses or gridding/martingale etc. It's a simple system with a fixed stop loss and take profit target.
My questions: