clock menu more-arrow no yes

Filed under:

Are "Balanced" Teams More Successful?

New, 5 comments

(an article very similar to this was published in SABR's Baseball Research Journal)

In this article, I will look at balance in terms of offense and defense. If a team scores 10% more runs than average and allows 10% fewer runs than average, they could be said to be perfectly balanced. Do such teams win more games than teams that are less balanced? For example, if another team scored 15% more runs than average and allowed 5% fewer runs than average, they would obviously be less balanced than the first team. Does the first team win more games due to greater balance, even though they seem to have about the same level of performance as the first team? That's what I am going to look at.

To measure a team's offensive performance, I divided their runs scored per game by the league average. Then that was park adjusted using the park factors from the Sean Lahman database. The 1980 Orioles, for example, scored 4.97 runs per game. That divided by the league average of 4.51 leaves 1.10. But their park factor was 99, meaning that 1% fewer runs were scored in their park than average. So the 1.10 was divided by .99 to get 1.114, which is then multiplied by 100 to get 111.4, meaning the Orioles were 11.4% better than average in scoring. I performed similar calculations for runs allowed. In that case, the Orioles got 111.77, meaning they gave up 11.77% fewer runs than average (I'm following the convention that Pete Palmer uses, so above 100 means the team was better than average at preventing runs). Let's call the runs scored measure "OFF" for offense and the runs allowed measure "DEF" for defense

To measure balance, I found the difference between their OFF and DEF and then found the absolute value. Let's call this result "BAL" for balance. The Orioles had a BAL of .374 (slightly different than the numbers above would imply due to rounding). Is this balance factor important or relevant? To test this, I ran a regression in which team winning percentage was the dependent variable and OFF and DEF were the dependent variables. The equation was:

(1) Pct = -.476 + .49*OFF + .482*DEF

(I divided both OFF and DEF by 100 for the regression). DEF is not negative for reasons explained in above. The standard error for 162 games was about 4 wins. I looked at all teams from 1980-2004.

Then I ran the regression with the balance variable added in. Here are the results:

(2) Pct = -.476 + .486*OFF + .488*DEF - .032*BAL

The standard error was still just about 4 wins for 162 games. It did fall by about .02 wins. So adding in a balance factor does not explain winning much better. The BAL variable was statistically significant with a T-value of -2.6. It has the right sign needed if balance is to help winning. As BAL gets larger (teams get less balanced), they win less. But notice that its impact is only about 1/16 of OFF and DEF. Adding BAL also had very little impact on the equation itself, which you can see by comparing equation (2) to equation (1).

How did the most balanced teams do? The table below shows the teams with the lowest 25 BAL scores.

Their average winning percentage is .491. So they did not win any more games than normal. The next table shows the 25 least balanced teams. Their average winning percentage was .497.

The next table shows the top 25 teams in winning percentage from 1980-2004. Their average BAL score was 11.737, while the average for all teams was 10.138. So the best teams are just a little less balanced than normal (remember that zero is perfect balance).

The next table shows the lowest 25 teams in winning percentage. Their average BAL was 10.165. So the worst teams are just about as balanced as anyone else. Lack of balance is not why they lost so much.

I also looked to see if teams that exceeded their "Pythagorean" winning percentage were more balanced than other teams. The "Pythagorean" winning percentage was invented by Bill James and it says that a team should have a winning percentage equal to runs scored squared divided by (runs scored squared + runs squared allowed). The correlation was .0025, meaning that there is no connection.

I also did a study once on whether or not teams with more balanced lineups scored more runs. I found that that they generally did not. The link to that study is below. That link has a link to a study by Keith Woolner. He found that balance did help runs scoring a little.

The San Lahman database