I just finished a backtest on a particular system over 17 pairs, from 2001 onwards.
I think 17 pairs +12 years is more than sufficient for robustness. I can’t test anymore beyond that since I don’t have data for anything else. But yeah, is this acceptable?
The stats…
Sample size = 600 trades
Profit factor = 1.65
Risk:Reward = 1:0.15 <- (yes, the reward is that small. It’s a “scalping” system on the weekly chart)
Years = 2001 to early 2013
Pairs = 17 pairs
While I can see where Mr Gone is coming from in saying backtesting is a waste of time, I believe there are benefits to backtesting. How could you come up with a strategy in the first place unless you had some idea based on past experience that it might work?
That said, if you are pleased with the backtested results, the next step would be to forward test it. How’s that going for you? I’m intrigued that your “scalping” strategy uses weekly charts, or that it would have only placed 600 trades over the past 12 years across 17 pairs.
Hi Jason. I haven’t tried trading it live as I’ve only finished my backtest last week. But I did optimise using 2001-2009 in-sample data, and tested out-of-sample 2010 to early 2013 and it worked out fine.
Because I’ve done bactesting several times and none worked over the long period, I mean puting the result on fire, it is a battlefield out there, do you think theenemy will do something just because a formula??? C’Mon
As far as how many pairs to back test to be honest you only need to backtest the pairs you plan on trading. those are all that matter. Think about it if you plan on adding a pair down the road you can all ways back test it then.
I will be honest though those results are not very great. If you are happy with them then my opinion dont matter.
Can you explain? It’s a mechanical system, and to be honest, I’ve seen the profit factors for other mechanical systems and they’re usually within this ballpark (1.2 to 2).
Backtesting can be very valuable, but you really need to make sure you are doing it correctly. Obtaining high quality data, cleaning or filtering the data and accounting for trading costs/slippage are a few but necessary steps before you can even begin backtesting.
If you are not willing to spend the time and effort to properly backtest your system, my recommendation would be to run it on a demo account for a couple months to see how it performs. If that takes too long, as it might with a weekly strategy, you could start it on an account without a huge amount of capital and very low leverage.
I tested my strategy in a micro-account. Yes, I spend some cash testing it but I needed to see if my strategy worked by trading with it. You dont trade in demo accounts so there was now way to test my strategy and myself.