The Death of Short-Term Mean-Reversion?
A little background for the uninitiated…
Since the late 1990’s, the stock market has been very prone to “short-term mean-reversion”. After a short period of gains it tends to pullback, and after a short period of losses it tends to bounce a bit. Short-term traders like us have been able to produce strong profits playing these very quick gyrations in the markets.
A simple example of trading short-term mean-reversion might be RSI(2), which measures how overbought/oversold the market is in very recent history. In the graph below I’ve assumed a trader ratcheted up long exposure to the S&P 500 as RSI(2) became more oversold, and short exposure as it became more overbought, vs buy & hold in grey.
Geek notes: (a) I’m trying not to get too geeked out here, but for more info, this is the same scaling approach we use in the monthly State of Short-term Mean-Reversion report, and (b) these results are frictionless but could be more or less reproduced using actively-traded mutual funds, our weapon of choice.
[logarithmically-scaled, growth of $10,000]
Obviously over the last decade, short-term mean-reversion has done a spectacular job pulling long/short gains out of the stock market, but note how in the last half 2009 the approach went flat; when the market has gone up, it’s tended to keep going up without pause (and vice-versa).
We lean heavily on short-term mean-reversion in our own trading, so it’s of the highest concern to me whether this breakdown in short-term MR is a temporary or permanent change (we track the issue monthly in our State of Short-term Mean-Reversion report).
I’ve always maintained my theory (at this moment) that this is a temporary aberration and that the market will return to being dominated by short-term mean-reversion once this bullrun fizzles.
Not everyone agrees with me.
Here’s one example justifying why I hold the theory I hold…
In the graph below I’ve shown the quarterly “daily return vs volatility” of the simple proxy strategy shown above.
Daily return vs volatility is the (geometric) average daily return of the strategy over that quarter divided by the standard deviation of daily returns. I’m using it (instead of just returns) so that I don’t unfairly reward/penalize the strategy for performance during periods of high volatility (like the financial meltdown).
If that didn’t make sense, don’t fret: high numbers = good performance over that quarter, low numbers = bad.
Clearly short-term MR has taken some lumps in late-2009/10, but they have NOT been unprecedented compared to other lumps it’s taken since the turn of the century.
Here’s another look, this one annual instead of quarterly to further smooth the data…
Same story.
YES, short-term mean-reversion is sick right now, but NO, I don’t think it’s time to abandon ship. We’ve been here multiple times this decade and the market has always returned to the new “normal”.
Will it return this time? There’s no way of knowing for sure, but until I see the market doing something out of character, I’m staying the course. Success in mechanical trading depends on not blowing with the winds of “noise”, but playing the big robust historical observations…even when they’re a bit under the weather.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Stock Market Mechanics | 15 Comments
I’m a mean reversion guy in the direction of the trend; in the long run this strategy works, but the 20-30% of the time when the market trends strong like we have for a while, you just cant compete. Again, only temporary set back. Since inception I’m still kicking the piss out of my benchmark.
RE to quant.this: you bring up a good point I didn’t mention in this post. This study (like the State of Short-term MR report) is just isolating short-term mean-reversion, regardless of the intermediate or long-term state of the market. I agree however that a real strategy needs to be considering all of those things simultaneously. Although our performance has been pretty flat as of late, trading just short-term MR regardless of the broader state of the market would have led to some pretty atrocious losses. michael
Michael: as usual, nice analysis. But I would argue that even if you’re right and MR is gonna roar back (big if, but certainly possible, and maybe even probable), YK would still benefit from a more diversified approach that didn’t depend so heavily on MR. That way the strategy could automatically gradually move away from MR and towards (something else) when MR is weakening, and come back when MR is strengthening, much as it adjusts among components now.
Of course, coming up with the (something else) is a non-trivial exercise, but that’s why you get paid the big $$. :)
what we do is combine trend following with mean reversion. So for the past year or so we have been long and adding leverage on a short term basis as the market dips. When the market was going down, we were basically short and adding leverage and selling rallies.
As for diversification, you can also think of YK as one of your strategies, which generates nice returns when MR is healthy and protects your capital when it is not (YK(B) has been quite good at protecting capital, at least in my view). If you couple YK with one or more different strategies – say trend following or swing trading on a multi-day time horizon – then you may get nice, smooth results for your portfolio as a whole. Of course, this way in the long run you may sacrifice some return in exchange for lower portfolio volatility.
That headline is a bit dramatic, considering your many statements about non-attachment to the blog. Thanks for the analysis though, very useful. You have obviously disclosed though that you are heavily invested in these strategies, so have you considered that you are just “talking your book” like any other money-manager ?
RE to TK: “talking your book” implies that you somehow have the ability to move the market invested in. In this case, MarketSci somehow moving the major indices (the only things we trade) is an impossibility. All I care about at the end of the day is performance. michael
Retail and professional traders on the sell-side talk their book as well all the time. I guess I was more loooking towards the possibility of presence of confirmation bias within the analysis (http://en.wikipedia.org/wiki/Confirmation_bias)
Just my sceptical 0.02p (while remaining your customer)
I read your analysis and there is something that bothers me a little bit. Essentially, you are dividing the MR trading results by the average volatility, and showing that this period does not seem to be very different from others. In other words, low trading results in a period of low volatility are “good”.
However, in real life trading, what we get are just the “pure” results of MR, not divided by volatility. From this non-adjusted point of view, MR is indeed going through a very bad period.
Thanks for the excellent analysis, as always.
eber
RE to eber: not sure I follow – I think we’d all agree short-term MR is going through a bad period (adjusting returns for volatility isn’t going to change the “sign” of the result, just the magnitude).
Consider for example the huge spike in the first graph above during the volatile markets brought on by the financial crises. At first glance you would say wowzahs, daily mean-reversion was reallly strong. But when viewing those returns relative to daily volatility, we see that it was still well within historical norms (on vol.-steroids).
michael
Thanks for your analysis, the way i look at this dominant trading style is : MR pnl = -autocorrelation * variance (almost by definition). The nice thing with this formula is the hidden dynamics behind. We have observed that autocorrelation turns negative (between -10% to -20%) when vol is high and there is almost a critical volatility. When autocorrelation is low, the vol is usually low and the negative pnl you’ll experience will stay low. thus a positive skew in the daily pnl profile. There are 2 ways to survive a hibernating equity market : a multi frequency model and a diversification with a few more assets classes (although some of them are to avoid). a
RE to Antoine: one small additional thought – daily follow-through and/or mean-reversion has (historically) been a very binary relationship.
I think correlation doesn’t properly capture it because that metric is by it’s nature taking the magnitude of price changes into account (which haven’t historically been as important). Here’s a past post that discusses:
http://marketsci.wordpress.com/2009/06/10/stock-market-follow-through-on-small-days/
michael
.I have one question.For a given stock in a given time how often is occurence of M size move compared to 2M size move.My studies over 20 years conclude that M size move should be 4 times more common than 2M size move—based on square root of time principal for option pricing/odds of move of a given size in the underlying in the BLACK SCHOLES formula. BLACK SCHOLES may not be perfect–but the fact remains that trillion dollars of derivatives are traded each day based on this formula & option market makers make money year after year & laugh all the way to the BANK.Similar to relationship between price of one month option versus price of 4 month option,If it takes one month to get M move, it would take 4 months for 2M move in the same stock in the same time period.This tells me that if human brain is kept out of the equation & in SYSTAMATIC(automated) investing in stocks, take profit be 1/2 the size of stop loss( & we already know that markets are RANDOM) then over thosands of RUNS investor shall make a whole lot of money.Out of four trials one makes one dollar 3 times & loses 2 dollars one time with net profit of one dollar–all one has to do is that when take profit is HIT, cancell the unfilled stop loss order & repeat the process over & over.Investor stays direction neutral at all times with no bias long/short.This is not just a theoretical question–if you spend few hours pondering this question I bet your next 5 generations can make tons of money over the next 500 years,just clicking on the LAP TOP & would never have to look for A JOB.Any criticism would be appreciated.Thank you for your help in advance.
Dr Prem Nath MD cell# 845 641 6778 email indus68@gmail.com
prem nath – this is easy to test – unfortunately, no it is not profitable. Stock prices are not random over the short-term and prices move in streaks. michael