Tuesday, August 29, 2006

Trading By Mean Reversion

A number of good trade ideas can be generated from the concept of the market's average trading price. Rather than trend following, this can be considered trading by mean reversion.

Let's start with the average trading price from the previous day's trade. The volume-weighted average price is posted daily to the Trading Psychology Weblog. A simple alternative is the pivot point defined by the average of the day's high, low, and close price.

Since September, 2002 (N = 998 trading days), 70% of all trading sessions in SPY have revisited their prior day's average trading price (defined by the high/low/close mean). This 70% figure has remained constant since 2004.

It makes sense that an index such as the S&P 500 would retrace many of its moves. It is subject to considerable arbitrage and, as I pointed out in a prior post, has shown poor trending properties. Think of it this way: The Spooz have averaged a daily gain of around .03% for the past several years. The daily range of the index, however, has been around a full percent during this time. Clearly there must be considerable backing and filling: much noise surrounding the market trend.

Since September, 2002, 88% of all trading days have traded above their prior day's close, but only 54% have actually closed higher. 85% of market days trade below their previous day's close, but less than half have closed lower.

Trading by mean reversion becomes a powerful strategy when we realize that the 70% probability of mean reversion expands significantly when markets are losing volume (volatility) and when they are losing momentum. This is one important reason I track such indicators as the NYSE TICK and the volume at the bid/offer in my market updates. When we see buying or selling pressure wane after a market move toward a range extreme, the odds are greatly enhanced of a reversion to the mean of that range.

Note that this is a trading concept that is scalable by time. During 2006, for example, the odds of the afternoon ES market trading back to the average trading price of the morning are about 75%. Once again, when we see waning volume and buying/selling pressure in the morning, those odds go way up. Tracking order flow during short term price ranges can assist the trader in finding mean reversion scalps.

Traders who utilize the Market Profile framework should be quite familiar with the mean reversion trade. When we test edges of the value range and cannot facilitate trade at higher or lower levels, a move back toward the point of control is a high probability trade.

I daresay a disciplined trader could make a living simply trading this pattern.

11 comments:

capmanager said...

You are very right Brett,

That's what the guys in the pit do.

gangsTA

Brett Steenbarger, Ph.D. said...

Good point, gangsTA. Mean reversion on a very short time frame is the basic way that locals trade. That's in the pits, but also on the screen among electronic market makers.

Brett

John Wheatcroft said...

What happens when everyone starts "trading to the mean"? Eventually all volatility gets wrung out and we reach that nirvana of stock markets - a null set - everyone agrees on the price of everything let's all go home.

I don't disagree with your premise I practice it everyday but I'm wondering why, since it is so obvious and has been for a long time, everyone doesn't.

What's the psychology behind this enigma?

Brett Steenbarger, Ph.D. said...

Hi John,

You ask a great question. It seems to be human nature to extrapolate from the immediate past to the immediate future, expecting rises to continuing rising and declines to continue declining. As long as that human nature is at work, it seems as though we'll have profitable opportunities by fading market moves.

Brett

NO DooDahs said...

What most people call "value investing" is really a mean-reverting sentiment trade with a long time horizon.

Brett Steenbarger, Ph.D. said...

Very good point re: value investing. Thanks for the note. The value approach has performed quite well in recent years, as have many mean reversion strategies at shorter time frames.

Brett

D TradeIdeas said...

Dear Brett,

Excellent article. This is the kind of trading style that The Odds Maker is shouting at us. We gave an explanation on our blog here. You bring to the discussion some solid historical data that supports the evidence.

Brett Steenbarger, Ph.D. said...

Thanks, David. I do see a lot of benefit from using the Trade Ideas program to calculate odds of moves during the trading day. This would allow a trader to calculate the odds of a move before the market open (based on historical analysis of the prior day's action) and then *update* those odds based on the most recent action (based on Odds Maker). Saweeet!

Brett

Winace said...

Brett,
Excellent blog, I take this whole concept one step further. I apply the principal of the trade (cost averaging exponentially) and combine it with my own proprietary way of identifying launches and extension limitations from a mean trading range. I did want to expand on one of your commentors questions. The fact if everyone traded in this style, volatility would be neutralized. He is exactly right, but there are a few reasons this does NOT happen. First off, all traditional trading material for educating the trader is exactly what you do NOT want to do. You identified that volatility is required for trading "against the sheep". Whatever is required to inject volatility and confusion into the markets will be done by the powerful entities within the markets. You can tatoo the way to make money consistently on most traders forheads, and they STILL will not listen, that is human nature, it goes against what they have been taught. I trade 99% winning trades, there are two things you need to control to be profitable. These two things are:
1. Money is power, there is a reason it is referred to as "buying Power". Identify how much capital is required to cover the market move you expect, construct an average cost calculator.
2. Market depth. You need to identify the historical percentages of market travel within the issue you are trading. We are making new extremes of market depth now that has never been seen before, I can prove this. Play for the maximum extension then add some for insurance. Take nickels until they hand out dollars.

Brett,
You and I need to talk, I think I have searched the world over for someone that has any idea of what I am talking about ;-)
Winace, Former "Ace" TA analyst, now playing the sure thing! www.channellines.com

The Fixer said...

Hi Brett,
Was looking for mean reversion trading and came across your post.
Your post is relevant for reverting to the last trading days avg. price. what about mean reversion over the longer time period ? For example reversion to the 200 DMA after being say +40% above the 200 DMA.
Need your views on this.
I have done a backtesting study on this on Indian NSE INDIA NIFTY index for the last 18 years. And based on that predicted a 15% fall in NIFTY by this month Nov`09 end. In fact I believe that most global markets have the above similar pattern. see details at my blog http://saanpaurseedi.blogspot.com

Cheers
~ The Fixer

Imrryr said...

Hi Brett,

I have a BA in Economics and a PhD in Cognitive Psychology. Currently I am a do educational datamining and models of human memory. About a month ago I started daytrading.

At first I was amazed the efficient market hypothesis seems to be such a weak and slow effect (i.e. stocks prices move so much that it isn't very efficient). Then I got to thinking about it. Traders all have some latency of decision after hearing a tradeworthy news item. Indeed, this trade latency distribution is probably Weibull shaped and the probability of a trade is likely something like the logistic distribution of a collection of power functions with different decay constants.

So, what this means, is that if you see strong momentum back towards a mean, unless there is other news, that movement is likely to grow strongly at first , have some period of strong movement, and then peter out as the long tail of the Weibull plays out. Obviously this depends on other factors, and it is also true that things like technical indicators can be treated as news by the street. Indeed, the price itself and it's movement can be considered news, and so we can see things like sustained momentum above the mean occasionally.

For instance, it seems that when an otherwise good company gets a bad news day you have 2 groups of traders. 1. Those in the stock currently who dump as quickly as their decision reaction time (some of them coming in with big loses if they didn't have high enough stops in) and 2. Traders outside the stock who see an opportunity. These (technical) traders are driven by the news of a stock below its average, expecting (at least implicitly) a reversion to the mean, they watch the stock and the farther it gets from the mean, the more keen the opportunity becomes. Since all these traders have different risk tolerances, eventually someone comes in as the price drops, and then more as the negative momentum slows, and then finally, a big signal is given as the stock stabilizes. At that time you get a gradual avalanche (Weibull shaped) as the stock starts to move up consistently as the traders move back in.

At this point all these movements seem pretty slow, so I think we are far from the birth of a truly efficient market. I just wish I had the time to make this into a nice predictive math model...

Since I'm new to this, feedback is welcome.

Phil