Fun with baseball stats

Dr. Giganto

Being the nerd that I am, I get a kick out of playing with Microsoft Excel, so I decided to put some baseball stats into a spreadsheet, and look at some correlations.

Basically, I considered team stats and how they are correlated with the number of runs a team scores.

For those who dont really know what correlation is, or how it's measured, here's a quick tutorial.

Correlation is the measure of the linear relationship between two variables/groups of data. Basically, if you put all the data points into graph form and put a line through it, how close to the line would those points be?

Correlation can range in value from -1 to 1. A correlation of 1 means that each time you increase one of the variables, the other variable will increase by the same amount. Thus, all the data points would lie exactly on a straight line, with an upward slope.

A correlation of -1 means that each increase in one variable will be followed by a decrease in the other variable, and each decrease is the same magnitude.

A correlation of 0 means that there is no linear relationship whatsoever between the two variables. If you looked at a graph of the data points, they would be scattered all over the page.

So, the closer you are to 1, the more highly correlated two sets of data are. The closer you are to 0, the less correlated. The closer you are to -1, the more the two data sets are negatively correlated.

So, first I considered the correlation between batting average and runs scored for all AL teams for the last 4 seasons. The correlation was .704, which is quite strong. That means that a team with a higher batting average is pretty likely to score more runs.

The next stat I looked at was OBP. The correlation was .844. Thus,
an increase in OBP was more likely to lead to an increase in runs scored.

Next was slugging percentage: .879. This was the best so far, until I considered OPS. The correlation between OPS and runs scored was .948!!! That's almost PERFECT correlation. That means that an increase in OPS is almost guaranteed to have a proportionate increase in runs scored.

Next, I looked at some "smallball" strategies, namely stolen bases, and sacrifice bunts.

The correlation between stolen bases and runs scored was -.145, which suggests that the more you steal, the less runs you're expected to score. However, this number is quite close to zero, and a test of the "significance" of this result suggests that the correlation is not materially different than zero. Thus, there is virtually no effect on a team's runs scored by stolen bases.

Sacrfice bunts had a correlation of -.310, thus the more you sacrifice, the less runs you're likely to score!!!

Last, I considered strikeouts. The correlation was .115, which is slightly positive, but was probably not materially different from zero. Thus, a team that strikes out a lot is not really expected to score many more or many less runs. Strikeouts have virtually no effect on runs scored.

Blitzkrieg

Administrator

Interesting stats, Tom. So based on your statistics, the old Billy Beane theories appear to hold true as far as OBP, but it seems like they don't consider Slugging to be as important, seeing as they never really have many players that rank highly in that category. In fact, I believe Frank Thomas had a nice season with them a couple years back and they let him go for next to nothing.

If I'm not mistaken, Barry Bonds OPS was well over 1.000 last year which even at his age, ranked among the league leaders. All baggage aside, I'm quite surprised that nobody has given him a call.

OPS is regarded as maybe the most important stat in baseball, so it's surprising to me that Jimmy Rollins, even though he had a tremendous season, won the NL MVP award when 4 of his own teammates were ahead of him in OPS!

Stolen bases seem to be a relative non-factor in your statistics, but that's where I think some stats do not tell the whole story.. Pitchers say that a stolen base or someone who is a big time base stealer can throw them off their game and ruin an entire inning, thus disrupting their performance later on..

It's opinion but also.. immeasurable.

Freak

Is it?

Blitzkrieg

Administrator

Well, I don't know.. unless we're going to go back and judge how each pitcher performs after giving up a stolen base.. but then, what would we look for? Runs allowed? Hits allowed? Walks? What if he's getting by with a bit of luck but the manager can tell he's about to relinquish the lead, making bad pitches, or what have you and decides to pull him anyway?

Freak

I don't know, sounds a bit like conjecture.

If the correlation between stolen bases and runs amounts to almost none whatsoever, I'll take that as a general rule.

Blitzkrieg

Administrator

I guess I'm wondering if what pitchers say holds any water. Does it affect them in the inning or even later on in the game? Does it throw them off that much, and why are they always overly concerned when a guy like Reyes is on first?

Just one of the many questions I have about the game. lol.

Dr. Giganto

shit...i had a big long post up that I accidentally deleted. Fuck...

anyway, i dug a little deeper, and I considered caught stealings and stolen base success rate.

The correlation between caught stealing and runs was -.335, and tests of significance showed that it is materially different than zero, with like 99% confidence.

Success rate had a correlation of .221, but this is not materially different from zero. I would need to increase the sample size (add more years of data), to make a more definitive conclusion.

However, it appears that not getting caught too often is the main factor driving the relationship between the running game and scoring runs. Teams apparently shouldnt run just for the sake of running, or run just to "put pressure on the defense" if they are getting thrown out a lot. They need to steal at a high percentage to truly make it worth it. Thus, if you dont have the talent on your team to steal at a high success rate, you should probably only run occasionally, to keep the defense honest.

If you NEVER tried to steal a base, the pitcher would pitch from the wind-up instead of the stretch, and the first baseman wouldnt hold the runner on at first, so you would lose the hole ont he right side of the infield. Thus, if you dont have high-success base stealers, teams should only steal occasionally, to make sure defenses account for it

Dr. Giganto

Here's another little analysis I did.

You know the expression "Dont make the first or third out at third base"? Well, I decided to look into it.

I used this run expectancy matrix: Sabremetrics 101: Run Expectancy Matrix, 1999-2002


What that tells you is that for each baserunner/outs-in-the-inning situation, how many runs did a team, on average, score. Data was from 1999-2002.

We'll assume the starting point is runner on 2nd. With nobody out, the run expectancy (RE) for this situation is 1.189. That means that on average, teams scored 1.189 runs between 1999 and 2002 when there was a runner on 2nd and nobody out.

If the runner either steals third or stretches a double into a triple, the RE goes up to 1.482, for a gain of .293 runs (1.482-1.189). But if the runner gets thrown out, the situation becomes no runners on, with 1 out. The RE for this situation is .297, for a loss of .892 runs.

With 1 out, the potential gain of getting to third is .258 (a little less than the 0 out case), but the loss from getting thrown out is -.608, which is much less damaging than getting thrown out with 0 outs.

With two outs, the gain is minimal (.043), and the loss is .344. More importantly, the RE drops all the way to zero, since the runner would have made the last out of the inning.

Thus, it is proven that if you are going to take a risk getting to third, it should be with 1 out, not with zero or two outs