the futility of trading system optimisation

There are many who would say that is a holy grail ;-)

Hold on a mo' - I'm talking about the optimisation, not live trading or walkforwards.

If I can't find an optimisation with only small drawdowns, then I've been misleading myself. Of course it could be a curve fitted beautiful system, but I hope I can find something without curve fitting. I need to find something without curve fitting.

Otherwise I'm just going to give up algorithms and start trading fundamentals.
 
Hold on a mo' - I'm talking about the optimisation, not live trading or walkforwards.

If I can't find an optimisation with only small drawdowns, then I've been misleading myself. Of course it could be a curve fitted beautiful system, but I hope I can find something without curve fitting. I need to find something without curve fitting.

Otherwise I'm just going to give up algorithms and start trading fundamentals.

Ok..in a nutshell...what has been/is your current Sharpe ratio and max drawdown?...What do you consider to be reasonable levels for both? What was your consec. winning trades vs consec.losing trades?

Astoria
 
Ok..in a nutshell...what has been/is your current Sharpe ratio and max drawdown?...What do you consider to be reasonable levels for both? What was your consec. winning trades vs consec.losing trades?

Simple questions but I'm having a really hard time answering them because until now generally I just looked at the equity curve and judged it by eye.

Haven't got Sharpe Ratios yet. Plan to program them in at some point. Everyone seems to ask for it all the time.

Max drawdown I'm looking at about 5000 $/€/£ on any particular one lot futures contract on average ignoring volatility.

Total max tolerable drawdown is 50% of equity - $20K, that's my personal escape clause. However I would reject any optimisation I looked at with such a drawdown. I can only assume walk-forwards and then real trading results would be worse.

max winning or losing streak = 6 trades in optimisation

i'm looking for less than 6 months in drawdown, and less than $5K

why do you ask? (n) ? (y) ?
 
Simple questions but I'm having a really hard time answering them because until now generally I just looked at the equity curve and judged it by eye.

Haven't got Sharpe Ratios yet. Plan to program them in at some point. Everyone seems to ask for it all the time.

Max drawdown I'm looking at about 5000 $/€/£ on any particular one lot futures contract on average ignoring volatility.

Total max tolerable drawdown is 50% of equity - $20K, that's my personal escape clause. However I would reject any optimisation I looked at with such a drawdown. I can only assume walk-forwards and then real trading results would be worse.

max winning or losing streak = 6 trades in optimisation

i'm looking for less than 6 months in drawdown, and less than $5K

why do you ask? (n) ? (y) ?

Was trying to get a feel as to how far you have progressed with your venture...not knowing your logic or anything else about system...I would recommend that drawdown not exceed 6%..if possible...12% absolute red flag...Sharpe ratio needs to be around 1.2- 1.5 to be sound...+1.8 and you should be cruising ...provided data is greater than, say 6 months...but the key is 'walk forward'....

Good luck..(y)

Astoria
 
All systems will fail at some point. Drawdown is based on the size of the sample. Larger the sample, larger the drawdown. The key is what you do with the money while you have it. You know that at some point with in the next 5-10 years the account is going to go on a slide so just take out profit at regular increments. That's how you win in the Expert Advisor world. I've got a system that lasts 10 years finally, but the thing is, if I didn't pull any profits along the way, it would have only averaged 10% annually.

I code in logic to stop trading at double my deposit. So, for instance I deposit 10k, system will bounce around and when it hits 20k I'll stop and pull my chips, then repeat. At some point it will have to fail, but hasn't in last 10 years, so I should keep running it until I see smoke.
 
Top