Reversion to the Mean

Reversion to mean cannot be stand alone method, it must be complemented by other methods like trend following or reversal method.

 

Reversion to the mean is as old as the hills. All institutional money uses the 200 & 50 MA'sand when price moves too far from these it's most likely to return. Intraday these are too slow for aggressive daytrading and scaplers have developed their own mid points.

 

I look at Range(H-L), Volume and Volatility. Those are areas where you may find mean reversion "works".

 

I've only found mean-reverting strats to be robust when fading moves on govt reports. An ATR-weighted Keltner works pretty well for intraday.

 

This is the example I see most often when refering to mean reversion, but thank you for the example.

Yohoo, I have read many times over about the 50, and 200 ma's. And I would have to agree with they are way to slow for agressive day trading. Could you elaborate more on daytrading setups, or similars?

Arnie, thankds for the info. This is a slightly different of a response then I have been seeing. Could you give a visual example, or, again, give a little more in depth.


-------------------------------------------------------------------------


I appreciate all of the respones. I didnt think It would go this far.

I wanted to also ask if maybe lescor, SammySOESa, or mschey. From what I have read, It seems those three have alot of experience and or do trade mean reversion strategies. Do u think you three could add anything?


Thanks again all,


-secXces

 

I don't know if this is what is being talked about in this thread....but I thought I would show how I use the SD_Mean.

This chart is a 10 tick ER. Depending on the speed of the Market....I often Scalp on 10 or 20 tick charts. Most of the time....I try for 2 ticks....on 4 contracts.

Rather than using SD on the Price....I use it on my Indicator_Oscillator.

The OB_OS lines that you see....adjust automatically....as the market is moving throughout the day.

OB_OS = 1_D
Extreme OB_OS = 1.5_D
Wild Extreme OB_OS = 2_D

While I won't go into details on my Scalping method in this thread....when the Indicator crosses the 1.5_D or 2_D....I have a very high win rate....Scalping the opposite direction.

I find this much better than using lines on the Candles or Bars.

I saved this chart as a Gif....doesn't look as good here....as it does in Paint or in TradeStation.

Good trading to all.

VSTscalper

 

You might take a peek here:
http://www.gummy-stuff.org/RTM.htm
or even here:
http://www.gummy-stuff.org/coin-tossing.htm
or maybe here:
http://www.gummy-stuff.org/mean-regression-2.htm

 

great analysis, gummy,

I was trying to find a chart of a coin tossing simulation in one of recently read books (Van Tharp, Trend following, Fortunes formula ??), but could not find it.

There a simulation of a true coin tossed milions of time and plotted showed that after n tosses it did not revert to the mean. In other words the number of heads and tails were way off the mean the more tosses were performed.

do you happened to have this simulation? I´d be very interested in this.

paul

 

Not sure if I understood the question but no matter how large the number of tosses you expect to find the number of Heads around the mean +- some standard deviations. Standard deviation grows with sqrt(n) so after 1000000 tosses typically the number of heads will be around 500 thousand +- a thousand.

If you are referring to the last time the random walk crossed the mean (switched from a profit to loss or from a loss to a profit) then this is governed by the arcsine law i.e. you can expect to stay most of the time on the green or on the red instead of keep crossing the mean and stay evenly on red and green. This is typical behavior of a random process.

regards

 

thanks for the reply.

The study I was looking for just showed graphically the increases in amplitude around the mean as the sample increases. I believe (but am not quite sure) this was explained as nonrandom sequences within randomness, i.e. the probability of having, say, 50 heads in a row, becomes probable at least every so often if the sample is just large enough.

And thereafter you start again (as always) with a 50% chance.

This was noted as an integral part of why trend followers succeed even though prices follow random distributions.

The graph I am talking about practically showed that you may in essence never return to the mean as the sample increases.

paul