Trading strategies selection and portfolio construction

[SamH]

Hello,

I managed to produce a "portfolio" of 90 strategies decent enough to trade them live on different ETFs on daily timeframe with entries and exits at the next bar open. Strategies trade in major ETFs, sector ETFs, commodities ETFs, ... There can be several different long and short strategies per one ETF. Strategies range from mean reversion, patterns, breakout, pullbacks, ... Some trade 20 or more times per year (have high CAR but also high drawdown), some only 4 or 5 times per year in special market conditions for a particular ETF (high win rate and SR, PF).

Currently my capital allocation between strategies is naive equal weight based on max concurrent open positions of the portfolio in history (49 open positions at the same bar was maximum since the year 2000). See chart:

Distribution of open positions is the following:

So my capital is mostly idle:

I want to increase my capital utilisation.
One option is of course adding new strategies that trade when my current ones are idle. This is hard and takes time. And also comes with a diminished return since adding a strategy with an equal weight lowers the exposure of all other strategies so that net effect is less CAR...
Brute force portfolio optimisation is not something I am fond of since strategies are so diverse and besides doing equal weight I do not see another valid approach. I hope I am mistaken. Because just playing around and randomly limiting the number of open positions improves CAR/MDD...

What other strategy selection/portfolio construction approach should one follow?

 

[SimpleMeLike]

What other strategy selection/portfolio construction approach should one follow?

Good Morning SamH,

1. Make sure your strategy selection/portfolio construction beat the SP 500 Index year after year for the past 18 years.

2. Make sure your strategy selection/portfolio construction return XX% year after year for the past 18 years.

If #1 and #2 is accomplished, do nothing, proceed!

 

[Specterx]

90 strategies?

As an initial observation, this feels like you're over-fitting - which is to say you have 40 or 50 strategies which are curve-fitted to a particular symbol and moment in time.

My suggestion is to:
1. Emphasize strategies which show an edge across multiple symbols and larger spans of time (though with a higher weighting on more recent history).
2. Emphasize strategies which are complementary as to market conditions traded. If the general market is in a strong bull trend, it's inefficient to have only a third of your capital deployed because you have one-third of the "position slots" effectively reserved for mean-reversion strategies and another third reserved for pullback trades.

What happens if you limit to a max of 10-15 positions?

 

[SamH]

Thanks.

What happens if you limit to a max of 10-15 positions?

This is what happens when I move from 56 (added some more strategies recently) to 40 to 35 to 30 to 20 to 10 max open positions (left to right in descending order of max open positions). Open in some wide screen with full resolution.

 

[d08]

Get instrument correlations and adjust the position sizes accordingly.