How to choose the right system for the current market?

[elit]

I have a question I hope we could discus here.

I'm wondering how I should solve the following situation. I have a few systems, one is good when markets are
trending up, the second is good when markets are trending down, and the third is good when markets are mostly going
sideways.

I would like to know how I can decide which system to trade when markets are in one of the three scenarios - up, down or sideways.
By looking at the chart in hindsight, it's easy to point out where it would have paid to use which ever system. By looking
at the equity curves for the systems and comparing the results to what markets were doing at the time clearly tells
what would have paid to focus on during the up, down or sideways markets.

To do it live is much harder.

Maybe a moving average could do the trick - whenever it is pointing up or down, I know which one of my systems are the best choice.
but this would require means to measure how much it is trending up or down. In a sideways market a moving average is still
always pointing north or south too....

how do you measure that? And is it a viable solution? This would probably lead to getting in late, wouldn't it? Do you have any better idea, please share.

I'm a systems guy so I guess I could incorporate some trading rules -

if MA is pointing up so much, then use system 1
if MA is pointing down so much, then use system 2
if MA is not pointing up or down so much, then use system 3

then the rest of the rules for each system.


Hope you have some experience or thoughts about this!

 

[giggollo]

You could trade all your systems simultaneously, and periodically change the capital allocated to each system based on how well that system has been doing over the last period. That way you don't have to figure out whether you're in an up/down market, and you let performance dictate which system is working better right now, rather than making assumptions about which system "should be working right now"

 

[ACM Trader]

Quote from giggollo:

You could trade all your systems simultaneously, and periodically change the capital allocated to each system based on how well that system has been doing over the last period. That way you don't have to figure out whether you're in an up/down market, and you let performance dictate which system is working better right now, rather than making assumptions about which system "should be working right now"

Agree. Almost impossible to anticipate changes in market dynamics. It is better to let the systems trade independently. Capital allocation, money management then is what we can focus on.

 

[elit]

Thanks!

I guess if the systems drawdowns are reasonable, then this is a good way. When each of the systems trades under the "wrong" market conditions, it should not be a big loser. And when it is in the "right" market conditions it should be quite profitable.

How do you then decide where to allocate the money? By looking at how the system is performing at the moment and putting more capital in a system that is performing well, while also looking at what the markets are doing?

 

[giggollo]

If you base capital allocation to systems on just how well the system is doing, you don't have to also worry about how the market is doing. The system performance will already be factoring in the relationship between the market and the system.

 

[elit]

Quote from giggollo:

If you base capital allocation to systems on just how well the system is doing, you don't have to also worry about how the market is doing. The system performance will already be factoring in the relationship between the market and the system.

I guess that's good. Should I take into consideration how well the systems are doing in comparison with how well they should perform according to my backtesting, or just how well they are doing compared to each other?

I guess how well the systems are doing in comparison with my backtesting should give me pointers on when the markets no longer reflect the markets for which the system was created.

 

[giggollo]

If there's a discrepancy between backtest results and actual results for a system, you need to determine why that's happening and fix it. Positive backtest results are not worth much if you can't replicate them in the real world, so that's an issue that needs to be dealt with separately for each system. Once you've sorted that out, there are many ways to allocate capital among systems, too many to list here, and some of those capital-allocation methods will involve comparing systems to each other somehow.

 

[elit]

Quote from giggollo:

If there's a discrepancy between backtest results and actual results for a system, you need to determine why that's happening and fix it. Positive backtest results are not worth much if you can't replicate them in the real world, so that's an issue that needs to be dealt with separately for each system. Once you've sorted that out, there are many ways to allocate capital among systems, too many to list here, and some of those capital-allocation methods will involve comparing systems to each other somehow.

Yes thanks! I realize these are two completely different topics.

Do you think simply allocating most capital to the system that is performing the best at the moment is a good approach. Or in other words allocating the least money to the system taht is performing the worst. I understand this can probably be optimized in a number of ways, but I like to keep things simple and fast.

This is something that is easy to figure out, by looking at the equity curve, isn't it? "System 1 seems to be having a bad week/month... -let's take away a little capital from that system and put it in system 3 that is doing great!"


EDIT:Thanks giggollo and ACM for taking your time answering!

 

[itmediaco]

Quote from elit:

I guess that's good. Should I take into consideration how well the systems are doing in comparison with how well they should perform according to my backtesting, or just how well they are doing compared to each other?

I guess how well the systems are doing in comparison with my backtesting should give me pointers on when the markets no longer reflect the markets for which the system was created.

There have been many improvements to Markowitz, perhaps someone more experienced with portfolio theory can elaborate (there's no point 'diversifying' among highly correlated stocks, if the market tanks...)

with my systems, I simply compare the correlation coefficient of the daily log return of each system... the goal is to reduce volatility of the sum of the equity curves by combining systems with a minimum risk of loss at different profit targets. if the series are cointegrated, I know it's not a random walk, etc.

 

[elit]

Quote from itmediaco:

There have been many improvements to Markowitz, perhaps someone more experienced with portfolio theory can elaborate (there's no point 'diversifying' among highly correlated stocks, if the market tanks...)

with my systems, I simply compare the correlation coefficient of the daily log return of each system... the goal is to reduce volatility of the sum of the equity curves by combining systems with a minimum risk of loss at different profit targets. if the series are cointegrated, I know it's not a random walk, etc.

itmediaco, how many systems are you trading at the same time? Are they long-/short- and sideways-biased, or do each system trade any market?