From Decision Science News:
What of the adage “the best predictor of future performance is past performance”? It seems less true than Sting’s observation “History will teach us nothing“. Let’s continue the investigation.
DSN did a nice analysis on a ton of baseball game out comes to see whether a team who had just won a game was more likely to win the next game. There have been other studies like this involving basketball players “hot streaks.” Similar results revealed… well, it’s a crap shoot shot to shot, game to game.
Now, over the long haul winning records, shot percentages indicate there is some skill involved. But at the micro level it just ain’t true!
Now why do we as fans, observers, interested parties believe in hot streaks, win streaks, etc. etc? is it a side effect of some other useful thing we do in associating events? or is there really some direct value in assuming immediate past performance indicates a similar future performance?
what can we test to figure that out?
the nba hot streak article has some insights….
http://pagingdrgupta.blogs.cnn.com/2010/05/07/in-gambling-brain-explains-attraction-of-near-misses/?hpt=C2
I’m sure this is related…
Howdy Russ – I think there’s a real difference between this and the “hot hand” belief.
When a pro-player makes, say, a free throw, they do so with a probability p, which is greater than .5 (depending on the pro player). Since the player is experienced at free throws, there’s no good reason to believe that his or her probability of getting the next basket should go up or down based on one observation.
However, in this problem, with two randomly chosen ball teams, and without any other information, the chance that one will beat the other is .5. After one team wins, the estimated chance of it winning again against the same team should be updated to greater than .5. So everyone was correct in assuming that p would go up beyond .5, but it just turns out to be the case that one win, while informative, isn’t that informative.
BTW, I’m doing the same analysis in basketball and hockey next.
Agree that when you dig into the details there’s a real difference. However at the most basic level these issues all fall into the same category – for whatever reason humans need to think really hard to really understand the probabilities and often never really understand them – no matter how much data ya give em.
i disagree with your specific difference you called out. with a pro basketball player… if given know other information the free through is .5. it will go in or not. this is not really different than taking two random baseball teams you have no information on. The bball players free throw percentage is perhaps slightly more informative than a baseball teams winning percentage to predict likelihood of hitting any shot or winning a game.
am i wrong?
>am i wrong?
well, if you know something about pro basketball players, and since I said “pro” basketball players, it would be reasonable take as a “prior” the average hit rate for pros, which is about .75.
http://www.nytimes.com/2009/03/04/sports/basketball/04freethrow.html