I test robustness by changing the parameters that are used in my system. The smaller the impact of these changes on the performance the better. It proves that the basis of the system is good and stable, and confirms that the parameters are not by accident optimized for the tested period. By changing the parameters you "de-optimize" the system, you decrease the possibility that your system works well only for that period.
Help With Evaulating Intra-Day Trading Systems
- Thread starter SimpleMeLike
- Start date
I test robustness by changing the parameters that are used in my system. The smaller the impact of these changes on the performance the better. It proves that the basis of the system is good and stable, and confirms that the parameters are not by accident optimized for the tested period. By changing the parameters you "de-optimize" the system, you decrease the possibility that your system works well only for that period.
What works on one instrument, OFTEN does not work on another, or at least in the exact same way.
SOES Bandits was NASDAQ
NYSE Fishing Orders were NYSE
Trend following in the Turtle Trader days worked much better with commodities than stocks
Etc.. and so on...
However, the general guidelines of what to look for are often similar:
Trends
Mean Reversion
Scalping
Arbitrage
Informational Advantage
Executional Advantage
Cost of Playing Advantage
Regulatory/ Exchange / Counterparty Inefficiency
Etc.. and so on....
___
Also:
The Candy Store Often Closes
Edges end. The market becomes too hip. Loopholes close... especially for purely manual traders.
___
And:
Positive Expectation (edge)
Risk Management
Enough Volume
--are pretty much key ideas everywhere.
Last edited:
Try to evaluate/trade/time/gauge/predict only the obvious major intra-day moves.
I personally like to call them 'inflection points'. and ignore all the other relatively meaningless gyrations along the way.
(scalping or automating a system to capture crumbs is a waste of time)
All trading books are hilarious and meaningless, as far as I'm concerned. A great majority of them are written by squares, for squares.
Make Trading Great Again 2018...High-Five`, ET extraterrestrials
I personally like to call them 'inflection points'. and ignore all the other relatively meaningless gyrations along the way.
(scalping or automating a system to capture crumbs is a waste of time)
All trading books are hilarious and meaningless, as far as I'm concerned. A great majority of them are written by squares, for squares.
Make Trading Great Again 2018...High-Five`, ET extraterrestrials
Besides everything already mentioned, another thing is getting used to looking at equity curves. Yours looks terrible and combined with a profit factor of 1.10, you shouldn't trade that strategy (appears random). A profit factor of 1.5 or better for systems that trade frequently and over 2 for anything less frequent is a good standard.
Max drawdown is important as is drawdown averaged over months/years. It describes the risk because as already mentioned if you hit a 80% drawdown from the moment you start trading, you might as well give up because now you need a 400% return. There's also an emotional aspect to it because I seriously doubt anyone can remain very calm operating in a disastrous drawdown. You don't want to be in an emotionally strained situation because that will make you add to the mistakes already made.
The problem is that different instruments behave differently.
There is no such thing as ONE SIZE FITS ALL.
eg corn hardly move at all, and NQ movement could be quite violent.
even for gold, its behaviour/personality/characteristics change over time.
So I wonder how your program cater to all these differences.
There is no such thing as ONE SIZE FITS ALL.
eg corn hardly move at all, and NQ movement could be quite violent.
even for gold, its behaviour/personality/characteristics change over time.
So I wonder how your program cater to all these differences.
The right way to think about it is to understand how the reward/risk ratio is expressed. If it is expressed in units per time, then the multiplier should be sqrt(time). If it is expressed in units per trade, then the multiplier should be sqrt(trades). Hope this makes sense.
Thanks maxinger. It does not cater to any differences I just add all the instruments just to see which ones will make some money.
Better to be harsh but honest rather than sweet and deceitful, I suppose.
Your best strategy (the ES) produces a profit of $11 per trade. Assuming the ES is the futures contract, this amounts to a profit of less than 1 tick per trade. Did you account for slippage and commissions? If not, the strategy looks unattractive (a consistent loser, in fact). If yes, it looks borderline unattractive.