I have a minimum average daily dollar volume threshold excluding the most illiquide names. Then I calculate with a flat 2% slippage per trade (equivalent to 4% spreads). It really comes down to your average position size and thus market impact if you are too big. I never had much problems in US, Europe is much worse (for some Microcaps here, you can have a 5% spread + 5% market impact by investing mere $10k). US is even less problematic for me as I often get midpoint execution via IBKR.
With this info of course you have to weight your high-frequency signals (e.g. mom) low enough to keep turnover manageable. In my case I target 200% annual turnover.
In reality over the past years, I would say I do not lose more than 3-4% on fees+slippage/spreads per year.
To get a feelingfot the spreads/liquidity, have a look at my current holdings if you like.
That is interesting that you're getting midpoint execution on IBKR for US.
-Are you using an execution algo for US that you can't use for Europe?
-Also, are you crossing the spread for Europe?
I suspect that you're more optimistic about the expected alpha of systematic multi factor than me if you're willing to pay 3-4% on trading costs. :)
One exercise I found useful was to assume annualised alpha and trading costs for each factor, and then optimise factor weights for each factor by maximising SR/IR. When you apply trading costs directly to each factor rather than to the multifactor portfolio, momentum doesn't look quite so good anymore.
Actually I use limit orders I place near midpoint. In the US, I often get instant execution even in microcaps (nonOTC). The full mechanism behind it is not fully clear to me but I'll take it. Nonetheless, I always calculate with 2ish% cost per trade (4% total per closed position).
I fully agree that standalone Revision or Momentum Factors do not survive trading costs. I also see that in my backtests. But within multi-factor, they are "tamed" if used only as a kicker and if your core trading system is of low-turnover nature.
Of course I won't give the exact recipe but in the past I really only used a composite of 1y Sharpe Ratio, 12m Total Return and 6m Total Return Ranks. I got away from using short-term reversal signals or skips (too noisy in concentrated long-only with multi-month holding period)
On P123 I added just some other ways to measure 6-12m Mom to diversify the signal. E.g.
- industry mom,
- up down ratios (volume on up days devided by total volume, gives you a nice info about e.g. "frog-in-the-pan" dynamics and newsflow),
- a smoothed mom based on short-term SMAs instead of single price points,
- quarterly returns (separately) to get info about smoothness or acceleration...
I trade Australian micro and small cap equities using a multi factor model, but determined trading costs are too high to include momentum.
I'd be surprised if this were not the case for US equities too. How do you estimate trading costs?
I have a minimum average daily dollar volume threshold excluding the most illiquide names. Then I calculate with a flat 2% slippage per trade (equivalent to 4% spreads). It really comes down to your average position size and thus market impact if you are too big. I never had much problems in US, Europe is much worse (for some Microcaps here, you can have a 5% spread + 5% market impact by investing mere $10k). US is even less problematic for me as I often get midpoint execution via IBKR.
With this info of course you have to weight your high-frequency signals (e.g. mom) low enough to keep turnover manageable. In my case I target 200% annual turnover.
In reality over the past years, I would say I do not lose more than 3-4% on fees+slippage/spreads per year.
To get a feelingfot the spreads/liquidity, have a look at my current holdings if you like.
Hope that helped
Thanks for your reply.
That is interesting that you're getting midpoint execution on IBKR for US.
-Are you using an execution algo for US that you can't use for Europe?
-Also, are you crossing the spread for Europe?
I suspect that you're more optimistic about the expected alpha of systematic multi factor than me if you're willing to pay 3-4% on trading costs. :)
One exercise I found useful was to assume annualised alpha and trading costs for each factor, and then optimise factor weights for each factor by maximising SR/IR. When you apply trading costs directly to each factor rather than to the multifactor portfolio, momentum doesn't look quite so good anymore.
Actually I use limit orders I place near midpoint. In the US, I often get instant execution even in microcaps (nonOTC). The full mechanism behind it is not fully clear to me but I'll take it. Nonetheless, I always calculate with 2ish% cost per trade (4% total per closed position).
I fully agree that standalone Revision or Momentum Factors do not survive trading costs. I also see that in my backtests. But within multi-factor, they are "tamed" if used only as a kicker and if your core trading system is of low-turnover nature.
Great read thanks.
Great post. Can you spill the secret sauce in your momentum composite? 😁
Not much sauce to spill there 😅
Of course I won't give the exact recipe but in the past I really only used a composite of 1y Sharpe Ratio, 12m Total Return and 6m Total Return Ranks. I got away from using short-term reversal signals or skips (too noisy in concentrated long-only with multi-month holding period)
On P123 I added just some other ways to measure 6-12m Mom to diversify the signal. E.g.
- industry mom,
- up down ratios (volume on up days devided by total volume, gives you a nice info about e.g. "frog-in-the-pan" dynamics and newsflow),
- a smoothed mom based on short-term SMAs instead of single price points,
- quarterly returns (separately) to get info about smoothness or acceleration...