Should Strategy Work Across All Instruments

[Dhalsim]

When i am testing strategies now i am finding some great opportunities for the ES/YM/NQ - the strategies i test work across the board on these 3 markets (which should be expected). However, i find from years of trading these markets discretionary based that they all behave differently -especially the NQ.

Now i have started to test my strategy against a a basket of stocks. I am finding it very very difficult to find strategies which work well on all instruments.

None of my strategies are indicator based.

Do you guys have a similar approach or do you use common sense to deduce that stocks behave very differently to emini indices as do commodities to other futures markets.

If my strategy does not work across all instruments does that mean it is curve fitted to that one market, even if in sample and out of sample testing was done correctly?

Should a strategy on the SPY work equally well on the ES??
Do your strategies conform across all instruments or they tailored to exploit edges on only various instruments?

 

[Dhalsim]

Moreover, would you then devise and separate strategies based on particular sectors e.g. i have a strategy which performs really good on banking stocks and not so good on tech.
Would this be simply curve fitting or is this the way we should devise strategies??

 

[dom993]

I think the "one size fits all" concept doesn't work in trading, especially getting down to the details of a trading strategy.

Each instrument has its own personality, which reflects its dominant players trading habits.

Stock indices (& corresponding ETFs & futures) aggregate their individual components personality, and add to that more trading habits of its own dominant players.

Then, you have to factor the liquidity/volume for each instrument, hedging markets (or lack-of), etc.

And stocks mostly don't trade overnight, where stock index futures do, which creates distinct opportunities in the 1st hour of the day-session.

 

[Sergio77]

My opinion (everyone has one you know...) is that unless a system works well on many instruments of the same class without any fooling around with the variables then it is probably a fit. It may not be the probability is high it is. I see some traders claiming uniqueness of strategies and others, like this one, arguing that a strategy must also work on negatively correlated markets to ensure there is no fit. The choice is ours and the consequences will be reflected on our account balance.

 

[hidethereal]

In my experience, most of the strategies I've developed generally work with most products but I do have to adjust for volatility and then decide if it's worth pursuing based on risk. Sometimes the volatility is so low that it doesn't make the trade worth it and sometimes the volatility is so high that I have to take on too much risk.

I have strategies that only work in slow moving bonds and strategies that only work in fast moving currencies and energy markets.

I've yet to have a strategy that I could blindly apply to every product and it be profitable without making minor adjustments.

You have to consider that some products are more trending and some are more mean reverting. It's difficult to have one strategy that works equally well in both conditions.

 

[Sergio77]

Quote from hidethereal:

I've yet to have a strategy that I could blindly apply to every product and it be profitable without making minor adjustments

I agree it is hard but unless that is the case chances are the strategies are curve fitted. The problem is that when you find out it is often too late.

 

[MrN]

Quote from Sergio77:

I agree it is hard but unless that is the case chances are the strategies are curve fitted. The problem is that when you find out it is often too late.

Curve fitting is not always a bad thing. Certainly, there are foolish ways to do it that reveal nothing but fools gold, but done correctly It can be used to find the strategy with the highest (based on present knowledge) expected profit.

The anti-curve fitting thing kind of reminds me of all the people who say, "correlation does not imply causation" Yes, of course this is true as a generality, but in many contexts correlation does indeed imply causation.

 

[vikana]

Quote from Dhalsim:


If my strategy does not work across all instruments does that mean it is curve fitted to that one market, even if in sample and out of sample testing was done correctly?

If you rethink "curve fitting" and realize it's about understanding how your systems operates in parameter space; there's a lot to be learned from extensive optimization/curve fitting/or-whatever-you-call it.

In general, I trade systems that do well over a large range of SIMILAR instruments (e.g. all stocks). I do expect that an ES system can be made to work as a SPY system, unless there is a market-structure reason for that not being the case.

I do not expect a gold-system to necessarily trade stocks, or forex.

 

[abattia]

Quote from Dhalsim:
... I am finding it very very difficult to find strategies which work well on all instruments...

IMO, here you’re asking “Should [Edge] Work Across All Instruments[?]”.

Ask any group of traders “What’s an edge?” and I doubt you’d get a single, unanimous, unambiguous answer. So I suspect you’ll also get little agreement on the answer to your question …

IMO, an “edge” is how a trader exploits some specific aspect of market structure to obtain a positive expectancy in a trade.

And as far as I know there is nothing that says traders can only exploit aspects of market structure that are common to all traded instruments.

So I don’t see why an edge has to work on all instruments to be called an edge …

 

[Stok]

Here is a system I trade on 7 markets. Same system & parameters on all markets, obviously some do better than others, but net positive on all (over 10 year sample here). When combined, the EC is nice.