Many
people are familiar with the concept of regression towards the mean in other
facets of life. It's used as a business model, as an investment tool and in
other areas, such as psychology and genetics. It also has plenty of practical
applications for sports bettors, as well, and it helps those who like to place wagers to
be familiar with the concept.
Regression towards the mean is
simply a way of stating that things tend to even out over time, or at least
gravitate towards their norms. If a .500 baseball team at the All-Star break
wins eight games in a row coming out of the break, regression to the mean would
indicate that they are more likely to go 2-10 in their next 12 games than 10-2.
There is a reason they were a .500 team for the first half of the season and
barring any new circumstances, such as a player returning from injury or the
acquisition of a key player through trade, they should play close to .500 for
the remainder of the season.
In the NFL, a team that has a huge
game one week is likely to come back down to earth a bit in its next game. It
may be a case of a letdown, but it is also an example of a team regressing to
the mean.
Players
vs. Teams
Regression to the mean occurs in both
players and teams, but the team aspect is largely dependent on the player
aspect. If a star player is due for a regression to the mean it will have more
of an impact on the team than the team's performance will on an individual
player. If LeBron James has several big games in a row and a regression to the
mean is likely to occur, that will affect his team more than his team being due
for a bad game will affect his individual stat line.
We touched
upon this for baseball pitchers a number of years ago in Baseball
Pitcher Form Reversal, which is an old concept from my mentor Jim Barnes.
Jim looked at a pitcher's season ERA and doubled that number and then
subtracted the pitcher's ERA from his last three starts - a widely available
statistic - and use that number as a rough guideline for how many runs the
pitcher might be expected to allow. The theory is that a pitcher with a season
ERA of 4.50 is a 4.50 for a reason. If he pitches two straight games where he
allows two runs and then throws a shutout, he will have something close to a
1.67 ERA for his last three games. But he isn't a 1.67 ERA pitcher, he's a 4.50
ERA pitcher and in one of his next couple of starts it's likely that he will
show why.
The theory
isn't concrete and has exceptions but is something that everybody who handicaps
the games needs to take into consideration. Many times a team or player that
has put together several outstanding or several dismal efforts in a row can be
counted on for some sort of reversal. If you have bet sports long enough, you
have a pretty good idea of what regression to the mean is, having seen the basic
concept in action on many occasions. Taking a team to reverse its recent play
also ensures that you're likely to get the best of the odds, as a team
who has looked exceptionally good recent will likely be over-priced, while the
dismal-looking team will be slightly under-priced.
