Why Turnover Matters

I have a very mixed opinion on high-turnover strategies as they can have many advantages and disadvantages and this is me thinking aloud about the topic.

Pro

If I see a high-turnover track record - either from another investor or my own strategies - I tend to put more trust into it, all else equal. Why? Imo what counts is not the years and performance alone.

If someone bought 100% NVDA 0.00%↑ in 2005 and held it until today, I congratulate and I am impressed but I cannot learn anything from this track record due to low sample size. Same goes for a backtest. Imo, the number of realized trades matters a lot in order to interpret repeatability and skill. Did the long-term NVDA 0.00%↑ investor hold for 20 years because

a) he shows expertise and conviction

b) he shows fanatic love for the stock

c) he lost his account password

Furthermore, frequent trades and “resets” reduce effects like start date sensibility. In discretionary investing, high turnover is often treated as a sign of low conviction. For me, in quantitative trading, a certain turnover is almost a necessity to build trust in a backtest, process or strategy.

Trust - some may say faith - always is a big component since you never know what the future holds and if past sources of return will remain intact. But turnover might also increase returns in itself by better exploitation of those sources of return. Each mispricing anomly has a certain half-life period.

Short-term signals like news, reversals, revisions, momentum may have a short-term performance impact of 1 month, 1 week or maybe even just 1 day. Longer-term (fundamental) signals might have a more sustainable impact of several months or quarters. But the longer you look out into the future, the harder it becomes to find profitable “predictors”. Every signal becomes noise if I just lag it long enough into the future. Turnover thus is necessary to keep the signal (or factor) exposure as high as possible by rotating into new opportunities.

Of course not every “signal” actually is a sustainable source of return. Thus, I stick to metrics from evidence-based factor categories with some logical background and risk-based or behavioral explanation.

Contra

My main concern with high-turnover strategies is of course the implied cost and underlying slippage assumptions in relation to expected returns. No matter how robust your process, the only two certain things in investing are trading costs (fees, slippage, spreads, market impact) and taxes. Returns are always uncertain.

Now, assume you found an evidence-based multi-factor ranking system which exploits proven short-term signals to create high monthly returns in the microcap space. To exploit the short-term signals to the fullest, you choose a really tight rank tolerance, maybe even with some additional sell rules and weekly or even daily control schedule. This puts you at a 1000% annual turnover. Sounds high but you chose a very conservative slippage assumption of 2% per trade and the higher turnover leads to optimal after-fee returns despite the high costs involved. It’s a long-only strategy which is 100% invested at all times.

Of course you have conviction that the underlying anomalies will persist…

But what if they don’t?

Many people assume that if Value, Momentum, Size, Quality, Sentiment and other factors stop working, these anomalies will do a full 180 degrees reversal and underperform for future decades. But that’s not how it works. If a factor or anomaly dies, it doesn’t turn negative. It instead becomes noise. In a long-only multifactor portfolio, the performance thus wouldn’t turn negative. You would just end up with a noisy market portfolio.

So, no worries, right? If every mispricing goes away, at least I will get market returns.

Well, if it weren’t for your 1000% turnover with 2% slippage per trade, you would.

With your current strategy, at every downturn you will have to wonder:

Is that it? Is it over? Should I pivot? And how long should I wait to confirm my assumption.

And at this point you didn’t even consider if your slippage and liquidity assumptions are right and sustainable? What if the market dries up? Will you be able to trade in and out of your positions even in 2009 or 2020? And if you pause in these market phases, are your positions suitable to be held 1 or 2 months longer or will they melt away because your short-term system mainly picks non-profitable diluters with large debt?

Even if no crash occurs and no anomaly vanishes: The higher the turnover, the shorter the average holding period, the larger the impact of the exact entry and exit. Microcaps often jump 10, 20 or even 50% pre-market or within early trading hours. What if your backtest relies on the exact fast entries and exits after news or other events which in reality are not feasible?

Last but not least if you have to deal with accounting, bookkeeping, taxes on your own, every % extra turnover can mean a lot more work and complexity for everyone involved (if not automated), so make sure you have everything in place beforehand.

Conclusion

As for everything, there is a sweet spot somewhere in the middle between 1-stock NeverSell and 1000-Stock high-frequency tick trading. I adjust turnover purely by tweaking my rank tolerance for existing positions. In most cases that means that I buy the top 10-20 names according to my rankings and hold them until they leave e.g. the top 3-5% of stocks in the universe by the same ranking.

For me personally, the sweet spot lately has been 200-300%, as it balances the following aspects:

1. It is enough turnover to refresh most fundamental and technical signals without too much noise leaking into the decision-making process. Of course, there will be situations where you buy a ticker just to sell it 2 weeks later to again buy it back 1 week after that. In this medium turnover range, the return still mainly come from letting winners ride for 4-8 quarters and cutting losers quickly after a few weeks (with some lucky sprinkles of random short-term movers on top boosting overall performance).

2. The turnover is low enough to keep me sane and to limit the importance of exact entries and exits. I don’t want to be obliged to look at my portfolio daily in order to catch new signals as fast as possible. I even want to have the possibility to say: “If I fall into a coma today and wake up 1 year from now, I will be fine”. A portfolio built from a strategy that works on a 200% turnover basis (6-month average holding period) usually has some fundamental “backbone”. It mostly contains real businesses which of course might underperform if left alone for too long but at least they won’t implode overnight.

3. If the brown stuff hits the fan and all anomalies fade, the AI takes over and markets become 100% efficient overnight in a big Kumbaya, I will be fine. As mentioned above, in this scenario most signals should become noise, only market beta and other true risk-based factors remain and my portfolio would most likely become a noisy version of the FTSE All-World Index. Assuming 7-8% real return and 2% slippage per trade at 200% turnover, I would most likely end up in a volatile sideways movement. Terrible, but not catastrophic.

I hope that helps someone.