FanPost

Statistics and under/overperforming teams

Ok, I was thinking about this stuff for some reason and needed to type it out to think it through.  This doesn’t really have anything to do with the Mariners per say but is kinda related to record predictors.

Ok, I have 3 different teams.  Team A, B and C all score 810 runs so they average 5 runs per game.  They also allow 810 runs so they average 5 runs per game.  They all should average a record of 81-81 in the classical approach of predicting records.  I don’t disagree with this.  Over a large sample the mean records should be about 0.500.

This post is more about deviation from the mean.  Can you build a team that will consistently play to its potential?  Can you build a team that relies on luck but would have a better shot at having a perfect season?  Can you make a team that gives up the as many runs as it scores and still dominate?

Team A has very inconsistent pitchers.  They either pitch a shutout or give up 10 runs.  Their offense is very consistent and always scores 5 runs.  They will always have the record of 81-81.

Team B has very inconsistent pitching and hitting and randomly scores and allows runs.  They can have any record.  They will usually average a 0.500 record but they could have a perfect season.  The chance that a team like this has a perfect season is (1/2)^162 ~ 1.71x10^-49.

Team C has one starter that gives up 20 runs a game and their offense scores runs with the same frequency no matter who pitches.  This team should consistently outperform A and B because they should win most of the games where this horrible pitcher does not pitch.  If this bad starter pitches every 5th game and pitches in 32 games he gives up 640 runs.  This means that Team C allows 170 runs in the 130 other games played.  This team will probably win almost all of these games and have a record around 130-32.

This doesn’t really tell us anything about how well the Mariners should do this year.  It is more of an interesting problem for somebody to ponder and pull something out of if there is something.  Maybe there is a better predictor of record based on how polarized the talent is of the starting pitching since the offense should produce the same amount of runs on average in every game.  This means that if you have a team that you know will have RA=RS, then you should try to get some good pitchers and a horrible pitcher so you can win the most games.  I don’t think this really helps us learn anything about making a better team but I’m going to post this anyway because I already typed it up.