Intraday mean reversion vs momentum

gaj · Post by **gaj** » Sat May 12, 2018 7:57 am

Long time lurker first time poster. Would like to hear your thoughts on this.

Most short-term (intraday) strategies that I know of are some variant of mean reversion. Examples: market making, pairs trading / stat arb, index arb. The general framework is, look at some indicator (spread between two instruments, price vs moving average, bid-ask vs weighted mid, etc), buy when it goes down, sell when it goes up. You may have other bells and whistles, but the core thesis of these trades is mean reversion of the indicator.

1. Is there any intraday momentum based strategy? I've heard an argument that intraday trends are too small to give any edge after costs. On the other hand, at an ultra high frequency horizon, there may be momentum due to microstructure effects. I guess this is essentially order flow anticipation. Are there any exploitable effects at a longer horizon (seconds to minutes)?

2. Can all short-term strategies be seen as either mean reversion or momentum? A lot of strategies can essentially be reduced to a single one-dimensional indicator. In this case you only have to choice: buy low sell high (mean reversion), or buy high sell low (momentum). Things might get more tricky when you have multiple indicators that can't be combined easily. Is there any strategy that trades on mean reversion on indicator A and momentum on indicator B? Or something more complex than that?

sharpe_machine · Post by **sharpe_machine** » Thu Jun 28, 2018 12:34 am

1. Of course, there are. I personally traded such a strategy on some EMEA market. Holding periods ranged from approx. 1 sec to several minutes. Do not trade it anymore due to low scalability.

2. > Is there any strategy that trades on mean reversion on indicator A and momentum on indicator B? Or something more complex than that?

Just "yes".

gaj · Post by **gaj** » Thu Jun 28, 2018 2:23 am

@sharpe_machine

Is your strategy like A) the price of this instrument has gone up in the last N minutes, so it's likely to go up more in the next minute, or B) other correlated instruments have gone up, so this instrument should also go up.

I have heard people doing B and calling it momentum strategy, although it's actually a mean reversion strategy (the ratio between correlated instruments is mean reverting). Strategy A is what I'm interested in. It is well-documented that the effect exists in longer periods, but does it work in intraday timeframe?

tradeking · Post by **tradeking** » Thu Jun 28, 2018 4:43 am

> Strategy A is what I'm interested in. It is well-documented that the effect exists in longer periods, but does it work in intraday timeframe?

Why don't you test it out yourself?

gaj · Post by **gaj** » Thu Jun 28, 2018 4:55 am

I did. I see slight mean reversion in most cases but I suspect it's partly due to microstructure effects.

sharpe_machine · Post by **sharpe_machine** » Thu Jun 28, 2018 11:08 am

It is very much like "A". On its prime time while it traded symbol "SYM" it did not use any info about other markets.

ronin · Post by **ronin** » Thu Jun 28, 2018 3:31 pm

@gaj,

Your B is a momentum strategy. It's only mean reversion if you hedge it with some of those related instruments that went up.

You can tell by the risk profile. If it is positively skewed and doesn't benefit from diversification, it's momentum. Negative skew and benefits from diversification, it's mean reversion.

HitmanH · Post by **HitmanH** » Fri Jun 29, 2018 2:17 am

put an email in your profile please gaj...

gaj · Post by **gaj** » Fri Jun 29, 2018 2:27 am

EspressoLover · Post by **EspressoLover** » Mon Jul 02, 2018 10:44 am

> mean reversion. Examples: market making, pairs trading / stat arb, index arb.

At the risk of getting into a debate over semantics... I don't think these things are all necessarily inherently mean-reversion. Let's say I'm market making but only joining the side with the largest orders. Or let's say I'm trading based on a pairs signal. But I only enter the side that hasn't moved, because I think the active side is leading and the stagnant side will catch up.

My general intuition is that mean-reversion strategies tend to work because they're providing liquidity in some sense. That is they're smoothing the natural noise-driven imbalance in the order flow. Momentum strategies tend to work because they're enhancing the process of price discovery. Often because some order flow is having insufficient market impact relative to its information content.