The result is quite remarkable. There is a very strong correlation (with correlation coefficient 0.94) between the ability of one team to beat another and how big its standard deviation is relative to the other team’s, assuming nearly identical RPG allowed. The team with the larger standard deviation (i.e., the team with the more inconsistent defense) tended to win more games.

For each 8% a team’s standard deviation is larger than the other team’s standard deviation the win ratio increases by about 4%. For the teams at the extremes of the graph above, this would correspond to an extra 5 wins or losses per season, if they played all their games against the same team, even though the two teams have the same RPG.

It is a separate question whether consistency is a characteristic of the team, or just a random fluctuation that is not repeatable. The data from 2003 to 2008 has some teams who have a consistently high or low standard deviation, but many have a range of standard deviations, and a larger sample is needed to make any definitive conclusions.

In any event, the next time you wish a pitcher was more consistent, what you really mean is that he should just reduce the number of bad games, since being more consistent without improving your ERA is in fact detrimental to your team.