Fallacious terms - "Regression to mean"

[spike]

A term I see used a lot in the poker forums. But I'm sure it's being misused here.

It appears to be used in these forums by people who believe that if you have a good (or bad) run of cards that temporarily change your win rate, then you will in due course, experience a contrary run of cards that will make you regress to your mean win rate.

It's almost as if the people who use this term believe that a player has a fixed mean win rate and that the gods of poker notice when they deviate too far from their assigned mean and force them to "regress" to it.

I notice though that they only ever use this term to describe people on a winning rush who should then regress down to their mean. Players who have a bad run never seem to be able to use the same concept to get back up again.

Spike

[brad]

'then you will in due course, experience a contrary run of cards that will make you regress to your mean win rate.
'

i think it is you who are misusing it.

assume win rate of 300/hr for 100 hours.

then assume 30/hr for 6000 hours.

so for 6000 hours you have 30,000 + 180,000

so winrate is 210/6 or 70/2 or 35/hr.

probably better if i didnt make the good run so good. heh

[pudley4]

Assume you are a winning player (+1BB/hour). Now you go on a hot streak averaging +10BB/hour.

You should not expect to encounter a losing streak that brings you back down to exactly 1BB/hour. What you should expect is to "regress" back to your original win rate of 1BB/hour from here on out

No, No. A winning player avg 1BB/hour over a long period goes on a hot streak averaging 10BB/hour for a short period should expect a win rate of 1BB/hour for the next hour. Similarly if he averages a loss of 1BB/hour for a short hour should expect a win rate of 1BB/hour for the next hour.

Regression to the mean occurs when the two variables have some positive correlation. Here the two variables are two separate time slots (the last hours and the next hour) there is no statistical correlation of the results between these two time periods, hence the best guess for the next hour is the Mean.

Regression to the mean is misused in this context.

[Glenn]

The partial understanding of concepts is often more dangerous than never being exposed to them at all. It is my belief that people without a strong background in statistics and probability should stay away from these concepts unless they first study and understand the basics of the subject itself. Otherwise, just knowing the results is fine. I don't fault authors like Mason and David for broaching these subjects and giving solid, correct analysis of how probability and statistics relate to poker. However, the average player doesn't have the background to understand this information, nor does he have a real need to. Of course, understanding these subjects help serious players a great deal. However, for most recreational players, they do more harm than good. For example, the people who make 100K a year at their jobs, play 3-6 twice a month for fun, and obsess about bankroll management.