June 01, 2004
Luck Demonstration
I recently wrote this post about how, after about 45 games, we really couldn't tell if teams were good, bad or unlucky. Just to give you an interactive demonstration, I wrote a script to simulate a 45 game stretch. There are sixteen teams, and each team plays the other 15 three times. In each case, teams have an equal chance of winning. Hit the refesh button to run it over and over and see how, even equally matched, the standings spread out.
Posted by David Pinto at
11:00 PM
|
Standings
|
TrackBack (1)
I like it, 'cause I like the laws of probability, and I'm a former gambler (on sporting events), but I don't quite see how this is really useful. There aren't any evenly matched teams in any league of any sport.
Even the ones considered evenly matched talent-wise, have various things changing the balance - daily mentality (are they on a winning or losing streak perhaps? off the field problems?), injuries, and pressure...to name a few.
Coins don't have any factors to alter the balance, so they actually do rely on the laws of probability.
Altho I should point out that I do believe there is a factor of luck in the games, just not enough to compare it to anything evenly matched.
I find it funny you labeled them Team 0-15, rather than the normal 1-16. That's either the mark of a programmer or a geek.:)
Dave,
You should do this (A) for an entire season, and (B) for a short (i.e. 5- or 7-game) series.
This is great -- thanks for the effort.
Dave,
I bumped into this website by accident, but I'm hooked. You do a great job making any little thing interesting, and you bring new perspective into baseball blogs. Thanks for all your hard work and I loved the program (I'm a KC aka Small Market fan, and I like to see that it's possible that my team can win, to)
Actually, if you were going to show a short series, it would be more illuminating, I think, to demonstrate two unbalanced teams, so that people can get some sense of how often the underdog wins.
As long as I'm suggesting features, you might try a score based approach as well, so people can get a sense of how often the gritty weaker team wins the close games.
If you are interested in calculating (rather than simulating) a short series, see the URL above.
Danil
The great book "Curve Ball" by Jim Albert and friends has some neat analysis along these lines. For instance, they have an analysis of how often the best team will win the World Series, based on chance but also including differences in talent. I forget the percentage, but it's surprisingly low. I agree with Danil that MLB teams don't have equal talent -- so this exercise is misleading that way.
The same thinking should be applied to hitting streaks, clutch hitting, etc.
What does this tell us about the NFL?
