The (NEW) State of Short-Term Mean-Reversion: July, 2010

01Aug10

This is our monthly health check of short-term mean-reversion in the US market.

Why a health check? Because short-term mean-reversion (by “short-term”, think for ex. RSI(2) or DV(2)) is so important to what we index swing-traders are doing right now because at this moment in history, it’s the most effective directional trade.

 

*** CLICK TO ZOOM ***

I’ve changed up the format of the report, but we’re still using the same two simple strategies to serve as proxies for all short-term MR:

  • The first, daily mean-reversion, assumes that we went long the S&P 500 at the close when the S&P 500 closed down for the day, and short at the close when it closed up.
  • The second, RSI(2) stretch, assumes we took a larger long position the deeper RSI(2) closed towards 0, and a larger short position the deeper it closed towards 100. For example, RSI(2) of 0 = 100% long, 25 = 50% long, 50 = no position, 75 = 50% short, 100 = 100% short, etc.

I’d never suggest trading either strategy as I’ve defined them, but they make a good proxy because this tendency for the market to retrace very recent gains is exactly why all of these short-term indicators work the way they do.

Both strategies weight long and short trades equally. As the graph of each strategy’s performance since 2000 makes clear, for most of the last decade, playing it “symmetrical” has been an effective strategy.


[logarithmically-scaled, growth of $10,000]

Reading the Report

The first table in the report shows results over various periods of time in terms of volatility-adjusted daily return. Vol-adjusted daily return is simply average daily return divided by the standard deviation of daily returns, and serves to normalize results regardless of market volatility.

From my own experience, I’d characterize results above 10% as being very good (and profitable). Negative results mean that particular strategy lost money over that timeframe. To put the numbers in some perspective, I’ve also included buying and holding the S&P 500 index.

The two graphs below that (click to zoom) plot 3 and 12-month vol-adjusted returns so that we can better see whether short-term MR is waxing or waning.

The State of Short-Term Mean-Reversion

Since I’ve taken up so much time already, I’ll keep my comments short.

As we’ve discussed ad infinitum, short-term MR has been weak for about a year now, but as both graphs show, we’re still within ranges that we’ve seen in the past. For the moment I’m expecting short-term MR to again be the play du jour.

I’ll have more detailed comments in next month’s health check.

Happy Trading,
ms

Geek note: returns in this post have NOT been dividend-adjusted.

. . . . .

To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.



7 Responses to “The (NEW) State of Short-Term Mean-Reversion: July, 2010”

  1. 1 telacimr

    Here’s some food for thought that you might be able to run with…. I also trade mean-reversion type systems and have been trying to figure out why the market mean-reverts at all. For a large part of the 20th century (as you know), the market was momentum-driven. Then about twenty years or so ago, the market started to transition to mean-reverting behavior, with this behavior steadily growing stronger in recent years (although as you point out, it’s been weak now for about a year).

    Digging into the academic literature and looking at the data, here is one hypothesis that if correct would explain the transition to mean-reversion *and* the (relative) weakness in mean-reversion this past year:

    For the better part of the 20th century, the market was dominated by individual investors. But as time went on, particularly after 1980, institutions began to dominate (individual investors gave their money to institutions to invest for them rather than investing in the market themselves). Institutions have high rebalancing needs. Some want to track a benchmark, others want to maintain certain proportions of their assets in bonds and equities, all want to inoculate themselves from being sued by being good fiduciaries, etc. So institutions need to trade a lot, but they’re not trading because they’re informed, they’re trading for reasons unrelated to any material knowledge of which direction the market will take. For this reason, institutions are viewed as noise (i.e., informationless) traders. As as a result, market makers will for the most part trade with institutions without requiring any concessionary price levels (as they assume the institutions aren’t trying to game them). However, every now and then (on the order of days) after the market has made a strong move in either direction, the market makers, having provided liquidity for the institutions, will find themselves with heavily slanted inventory–too many longs or too many shorts. Consequently, in order for market makers to be compensated for this inventory risk, the market undergoes short-term reversals to allow the market makers to unload their slanted holdings at favorable price levels (it’s as though they’ve cornered the market on longs or shorts and thus mark up or mark down their holdings when investor demand shifts after the strong market moves).

    OK, if that theory is in the ballpark, what we should observe is that the mean-reversion properties of the market increase as trading volume increases. And if you plot out trading volume and some measure of the market’s mean-reversion (I use the variance ratio), we do see that there’s a strong correlation between the exponential increase in trading volume (across the 20th century) and a secular decrease in the variance ratio (meaning the market is becoming more mean-reverting).

    Further, because the rise of institutions has led to the rise of noise trading, we should also see a secular increase in volatility as the mean-reversion properties of the market increase. And this too can be seen by taking some measure of volatility (I used the rolling 5-year average of the daily absolute returns) and comparing it to the rise in mean-reversion. There has been a secular increase in volatility over time that has corresponded with the secular increase in mean-reversion.

    Now, if you’ll notice, the level of trading volume we’ve had over the last year has been small in comparison to the trading volume in the 2007-2008 (and early 2009) period. Reduced volume means institutions are not trading as much, which would suggest that market makers don’t have to mark up or mark down slanted inventories as often as they did when the market was making violent moves during the Great Recession selloff or when the trading volume was high during the “goldilocks” period of the 2003-2007.

    If the above is true, then mean-reversion will come back in force when institutions ramp up their trading volume again.

    • 2 telacimr

      One further prediction of this hypothesis would be that Asia ex-Japan, which for the majority of its trading history has been momentum-driven, will become progressively more mean-reverting as its liquidity (trading volume) increases as more and more investors partake in those markets. From my own research, this does indeed seem to be the case. The momentum properties of Asia ex-Japan have become tamer in recent years, to the point that despite the strong upward move of the markets since the 2009 low, a momentum-based system (at least the one that I’m monitoring) would only be flattish trading Asia ex-Japan over the past year.

      On the bright side, at some point Asia ex-Japan will be ripe for mean-reversion strategies as well.

  2. 3 Mike

    Very good posts, the FTSE 100 is also showing a change from momentum driven (often retail) characteristics to a noisy mean-reverting process (largely driven by institutions driven by hedging requirements). The change was pronounced from 1997 to 1998, since then mena-reversion seems to be the order of the day.


  1. 1 Basic Anchored Oscillators « CSS Analytics
  2. 2 A Return to Mean Reversion? | System Trading with Woodshedder
  3. 3 Regime Switching System Using Volatility Forecast « Quantum Financier
  4. 4 The Overnight Edge: Drilling Deeper | System Trading with Woodshedder

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s