Errors and Winning
February 26, 2005
I received an email from a high school baseball coach today:
I’m a new assistant high school coach and I am trying to relay the importance of not making errors to my infielders. Do you have or know where I can find a graph, chart, or statistic that shows the more errors you make the less likely you are to win?
You know, I tend to get so carried away with advanced baseball statistics that I sometimes forget the basics. And we should always remember the basics, right? So here’s a graph of every team in major league history, illustrating their fielding percentage (errors divided by total chances) and their winning percentage:
It seems to be pretty simple, right? If you make less errors, your fielding percentage goes up. And, as your fielding percentage goes up, you win more games. But take a closer look.
See, the triangles pretty much flatten out once fielding percentage reaches around .860, which implies that if you average less than four or five errors a game, errors don’t impact winning. But that has more to do with baseball history than anything. Those fielding percentages below .900 primarily occurred in the 1800’s, when some teams registered as many as ten errors a game. Those guys really played with a patch of leather on their hands instead of what you’d call a glove. Today, major league teams average less than one error per game. So let’s draw a different graph that corrects for this data problem.
I created an “error index” for each team, which basically compares each team’s errors per game to the average number of errors per game that year. This way, teams are compared to other teams in similar playing conditions. Here’s a bar graph of the index against the average winning percentage:
As the index goes up, the wins go down. Said differently, if you make more errors than your competition, you’re more likely to lose. Teams that made half as many errors as the competition averaged a winning percentage of .600. Teams that made 50% more errors than the competition averaged under .400.
It really is pretty simple after all.
Studes’ findings are right in line with others I have seen—the correlation between W/L % and fld% is much higher than it should be, based on the variance and weight of errors.
If you run his numbers for a team which allows half of the lg avg errors for a .600 W/L%, an error saved is worth about .3 win, or about 3 runs. But an error is really worth only .5 runs.
So, the natural conclusion is that error rate is much more important than commonly thought, probably/possibly because errors have a strong “indicative” significance of good players/fielders.
IOW, perhaps players with good Fld % are better fielders than we think, and that the contrary example of a statue with few errors simply is not common.
So, a GM who wants a “shortcut” method of evaluating players, should look for the best hitters he can find who have good Fld %.
I guess that implies Jeter at SS, but even in that sort of test case, you can do a lot worse than having Jeter at short on your team.
Posted by .(JavaScript must be enabled to view this email address) on 02/26 at 04:41 PM
What if we adjust for number of balls in play? Does that change anything? Or am I misreading the way you did the error index?
Posted by .(JavaScript must be enabled to view this email address) on 02/26 at 10:54 PM
You read it correctly, Tom, but that doesn’t really change anything in the macro sense. I ran the same metric for fielding percentage instead of errors/game and got the same trend.
Posted by
studes on 02/27 at 07:08 AM
Dave, could you rerun, but only over the last 50 years or so?
Echoing what David is saying, what’s being said here is that you get an extra 0.1 wins per game, by saving about 0.4 errors (I’m assuming 0.8 errors per game). That means 1 error = 0.25 wins, or 2.5 runs.
We know that errors minus outs are worth about 0.80 runs. So, why an extra 1.7 runs?
As David said, the errors contains extra information, like just overall bad fielders to begin with.
If you were to run a regression analysis that included both errors and hits per ball in play (assuming that DER is 100% fielding, which it is not), you would get a much different set of numbers.
Posted by .(JavaScript must be enabled to view this email address) on 02/28 at 01:20 PM
The trend is the same, but less extreme for more recent years. From 1970 to now, teams with an error index of 0.5 had a .595 winning percentage; teams that had an error index of 1.5 had a winning percentage of .395.
I certainly agree that errors carry extra information with them, but I’m not sure a high school team should be too concerned about that.
Posted by
studes on 02/28 at 01:49 PM
Btw, Dave, I’m not sure why you didn’t link to your own article, but you might be interested in this article as well:
http://www.baseballprospectus.com/article.php?articleid=648&mode=print&nocache=1109622246
I think it was written a year after Drinen’s…
Posted by .(JavaScript must be enabled to view this email address) on 02/28 at 02:33 PM
It’s not just that high field% players may be better fielders overall; high field% players may also be better hitters. In fact, high field% teams may even have better pitching, because they have more $$$ and/or smarter GMs. The potential for spurious correlations is substantial (it’s OK Studes: high school kids should learn about statistics!).
The other intriguing possibility is reverse causation: perhaps official scorers give more errors (unconsciously) to losing teams?
Posted by Guy on 02/28 at 09:14 PM
Page 1 of 1 pages
Commenting is not available in this channel entry.
