Sample size?

[LiamH]

I've been messing about with the FXCM Strategy trader platform and have come up with a strategy that seems to do well on 50 & 250 tick bars. The problem is, there is only limited tick data available so I can only test over the last 18 days.

The 250 tick strat places 125 trades and shows a 30% profit with 10% drawdown and the 50 tick strat places 650 trades with 15% profit and 10% drawdown.

I prefer the 50 tick at the moment because I have a larger sample size than for the 250 but 650 trades at 0.5% risk to make 15% doesn't sound too good does it? Also, 15% profit with 10% drawdown along the way can't be good either can it?

I don't really know what's good and bad when coding strategies and my knowledge\ability with the coding language is massively limited at the moment so I can't do all that I want. Also, what kind of difference can I expect to see in these figures if I were to go live with one of the strategies? The backtesting has been done on the bid\ask data.

Cheers

 

[Adamus]

650 trades in 18 days? That's pretty much High Frequency Trading you're getting into there. You don't say what you're trading but you want to make sure that your profits cover the costs of trading - commission, the bid/ask spread you cross, the slippage.

The figures you're talking about are probably limited by the short length of your test period. If you had a 5 year test period and you had a 10% drawdown and made 15%, it's tradeable, but it might not be what you want for your account size. And over five years that 10% might actually climb to 20, 25% etc and also last for months in drawdown.

Basically this is all stuff you can find reading forums here and reading a good book on mechanical systems. I'm sure the stuff I've just said must have been said a hundred times before here.

 

[LiamH]

650 trades in 18 days? That's pretty much High Frequency Trading you're getting into there. You don't say what you're trading but you want to make sure that your profits cover the costs of trading - commission, the bid/ask spread you cross, the slippage.

The figures you're talking about are probably limited by the short length of your test period. If you had a 5 year test period and you had a 10% drawdown and made 15%, it's tradeable, but it might not be what you want for your account size. And over five years that 10% might actually climb to 20, 25% etc and also last for months in drawdown.

Basically this is all stuff you can find reading forums here and reading a good book on mechanical systems. I'm sure the stuff I've just said must have been said a hundred times before here.

Cheers Adamus - I don't suppose there are any good sources of information on backtesting you would care to point me to? I have no idea where to start really..

I'm a complete noob to this and don't really know anything about the pitfalls of backtesting a mechanical system but have lots of idea's I'd like to try... Obviously these would be completely useless if my backtesting results are away with the fairies!!

I'm running tests on EUR\USD and prefer tick data as it produces much more reliable signals than timed bars. My account is small so slippage should be really minimal - I wouldn't honestly know what to put in as a realistic figure though... A pip per trade, 2 pips? I really don't know. Any advice you can give would be much appreciated!

 

[Adamus]

It's all here on the forum. The messages go back to 2001 or so and there's a lot more that can fill the gaps for you here than I can. Use the search.

The slippage on the euro is generally small but I've had 15 points slippage on the GBP/USD last month and that's not meant to be much different. The commission and crossing the spread will probably cost a pip and a half. Add another pip to be safe, so 2.5 pips slippage and commission.

 

[LiamH]

The commission and crossing the spread will probably cost a pip and a half. Add another pip to be safe, so 2.5 pips slippage and commission.

OK, well that's completely anhialated every strat I've tried! Back to the drawing board for me! The problem is when the market ranges I get a lot of losers and even with 1 pip of slippage per trade it's complete devastation. Oh dear.

 

[Adamus]

Your commission of course will depend on your broker, I just gave you mine.

 

[LiamH]

Your commission of course will depend on your broker, I just gave you mine.

In the end it wouldn't make much difference - Spread was already taken into account on the back test but with so many trades being placed even 1 pip slippage per trade says goodnight to the account. I need to come up with something else or forget scalping strategies!

 

[LiamH]

Would like to add, Adamus - Your snippet of advice has more than likely saved me an account - Granted it would have only been a micro but I was revved up to go live today with $100 to see how it got on at $0.10 per pip.

So yeah, cheers for that. Shows how little I knew\know.

 

[Adamus]

Would like to add, Adamus - Your snippet of advice has more than likely saved me an account - Granted it would have only been a micro but I was revved up to go live today with $100 to see how it got on at $0.10 per pip.

So yeah, cheers for that. Shows how little I knew\know.

Doesn't your broker let you trade in a simulated account? Program traders generally run their systems in simulation for a while to flush all the bugs out of the system - and get a bit of experience of what happens when the pips really hit the fan.

 

[LiamH]

Doesn't your broker let you trade in a simulated account? Program traders generally run their systems in simulation for a while to flush all the bugs out of the system - and get a bit of experience of what happens when the pips really hit the fan.

Yeah, I think there is a demo option with FXCM strategy trader but in all honesty I liked the look of the equity curve and didn't think about possible slippage. $100 wouldn't have ended my world had it gone down the pan so was just going to let it loose for $0.10 per pip. Glad I didn't and will definitely test on demo before even considering going live with a mechanical strat in future.

Just hadn't really thought it through properly..