Strategy backtests well for ES but not for other index futures

C

cunparis

I have ran across this several times so I'm wondering what's going on:

I've written up a strategy and backtested it over the past 3 years for ES and it tests positive. I then try it with NQ and/or YM and find that it does not do as well, sometimes even negative or breakeven.

First thought would be that it's curve fitted and optimized. But I'm also finding this with systems that don't contain any special numbers other the standard indicators (bollinger bands at 2,20, stochastics at 14, macd etc..).

I'm just curious if this is really possible? I'd expect a system to do about the same for similar markets.
 
What type of exits are you using? How specific are they? A lot of times, the curve-fit exists on the exits more so than the entries as very few entry methods provide any edge to begin with. A good entry on ES will typically apply to other index futures contracts, but again, if the exits are curve-fit, then the system will not be truly robust.
 
Quote from TrueStory:

What type of exits are you using? How specific are they? A lot of times, the curve-fit exists on the exits more so than the entries as very few entry methods provide any edge to begin with. A good entry on ES will typically apply to other index futures contracts, but again, if the exits are curve-fit, then the system will not be truly robust.
More...

Well I'm struggling with my exits, so I made another post about that:

http://www.elitetrader.com/vb/showthread.php?s=&threadid=146005

One that I'm using now that seems to work better than others is simply exiting on a crossover of the 20 period MA. I'm sure there are better exits but for now that seems to work ok.

I've been trying very hard not to curve fit. For example I could run the optimizer on the period for the MA for the exit and probably find that 27 is the optimum, but I know in the future 27 probably wouldn't be the optimal value.
 
One thing to watch out for is how you calculate e.g. the size of the stops. Make sure to use a generic calculations, such as ATR, and not percent or a fixed number of points.

NQ and ES are fairly different animals, and I wouldn't be surprised that some systems do fairly differently on them.

I would try to optimize the base parameters for NQ and see if there's another "sweet spot". if that's the case, try to unify the two sets of variables.
 
1 tick on the ES is not the same as 1 tick on the YM.

SO you can't say like "Exit after 5 ticks" because it will be different on each index.

You'd have to somehow convert them into a similar scale, like maybe 20 YM ticks = 2 ES points, or something. I just made that up but compare moves from top to bottom and arrive at a ratio.
 
Thanks for all the responses especially the previous poster who has observed the same thing.

Yesterday I was working on a strategy that had a nice smooth equity curve for ES, NQ, & EMD over a max 3 year period (3 years for ES, I have less data for NQ & EMD)..

I tested it with YM and it was negative. I can't figure out why, it doesn't make any sense to me. I'm still trying to figure it out.

I agree about the tick size, etc. I'm not using any stops on my strategy, at least not yet. So that's not it. In the past I've used the optimizer to test all variations and come up with the best, but I ended up feeling like it was too curve fitted. So with this one I just use a simple exit strategy with MA crossover. So this should work on all the indexes.

Maybe YM just trades differently?
 
Quote from cunparis:

"... I'd expect a system to do about the same for similar markets.
More...

And it should. If it doesn't you're playing the wrong game to begin with.
 
All markets are different.

Does ES vs. Pork Belly difference be a surprise for you?

If not, why wouldn't ES vs. NQ be any different.

It's all relative and a matter of degree.
 
You must log in or register to reply here.
Back
Top