posted
A lesson from professor paul of the prominent probiscus...
First things first. Whats wrong with the traditional metrics of Batting Average, Home Runs, Rbi's, Runs, and for pitchers, Wins, ERA, K's, and Walks.
Well, for the easy ones, home runs, rbi's, wins, etc. are counting statistics. They are good for telling you how much of something a player did, but they leave out context. For example, a terrible player can drive in 100 runs if he's on a good team batting cleanup, and gets a lot of plate appearences. In 1993, Reuben Sierra had what baseball prospectus calls the worst 100 rbi season of all time. He drove in 101 runs, but was on base less then 25% of the time, and his slugging percentage was atrocious for a cleanup hitter... in fact, it was atrocious for a back up outfielder.
If you are a major league player, and on a good team, and play everyday, or as a pitcher, throw every fifth day, you are liable to accumulate a decent number of rbi and runs, or a pitcher wins and strikeouts. These numbers don't mean very much.
By putting numbers into context, we can start to read out some of these problems, by looking at rate stats, such as average and ERA. However, these numbers have a different set of problems. Batting average doesn't tell us very much... for example, Alfonso Soriano has hit around .300 for his career, but his on base percentage is a mediocre-at-best .325, while his slugging percentages have been around .500. On the other hand, last year Manny Ramirez hit around .325, had an OBP of over .400, and a slugging percentage over .550. While Manny had slightly more total bases per at bat, he had a significantly higher percentage of times he reached base, despite only a somewhat higher batting average.
Why does the comparison between their batting averages and the other, lesser used, statistics matter? Because OBP and SLG correspond better to the number of runs a team scores. If a team has 9 .300 hitters, they are probably pretty good, yes? Yes, probably... but only if those .300 hitters do something other then hit .300. If all 9 of those .300 hitters have OBP of .325, and SLG of .350, the team is not going to score very many runs. They don't get on base very often, and don't hit for much power, meaning the team has to rely on "creating" runs by stealing bases, using hit-and-run plays, and sacrifice bunts. These tactics work very well for scoring 1-2 runs in a game, but not for scoring 3 or more. At no time in baseball, has one run been enough to expect to win a game. Even the early '60's dodgers, with one of the best pitching staffs in baseball, and an extremely fast team, typically gave up about 3 runs per game, meaning they needed to score 4 to win. In other words, you need to get people on base, and move them around rapidly in order to get wins.
Obviously, for scoring runs, the most important factor is the number of people on base. Without them, you can't score. However, as a strict matter, high slugging percentage correltates more directly with runs. But when you combine the two statistics, you find a very accurate measure of how well a team scores runs. OPS as it is commonly referred to, has a correlation coefficient of about .925. For those familiar with mathematics, this obviously is not a best "best fit" curve, comparing a stat to runs a team scores... however, for teh world of sports, and baseball, its almost as accurate as you can get... unless you are willing to use complex calculations.
In terms of pitching, how about ERA? ERA is actually a pretty good statistic... but it doesn't take into account several things. Unearned runs, which are partially teh pitcher's fault as well, park effects, and team defense.
For example, 2 years ago, Derek Lowe, Pedro Martinez, and Tim Hudson were all in very close competition for the Cy Young Award. Pedro had the best ERA, but he'd given up more unearned runs, meaning that after a player reached base due to an error, Pedro was liable to let him score more often. Derek Lowe had the fewest strike outs, which means he relied on team defense more then the other two pitchers. (A note: By Baseball prospectus' calculations, Derek Lowe was teh best of the three pitchers, but he finished third in the voting).
To understand a pitchers effectiveness, ERA is a pretty good stat. But its better to look at RA. Run average, rather then earned run average. In terms of understanding how dominant a pitcher is, look at his K's per nine innings, and K's per walk. As a side note, it appears that baseball pitchers have a minimal effect on the batting average of hitters they face, on balls put into the field of play. That is, over a five year span, on balls that are hit into fair territory that are not home runs, all major league pitchers will have approximately the same batting average against. This is why strikeout pitchers with low walk rates tend to be very good from year to year, pitchers such as Randy Johnson, Pedro Martinez, Roger Clemens, Javier Vazquez, and Kevin Brown. Nolan Ryan and Bartolo Colon never quite broke into the top teir of pitchers in terms of runs allowed because they were wild... they put extra batters on base. Also, certain pitchers give up more home runs from year to year. This is why Jose Lima is a bad bet to have another 20 win season. He got lucky on batting average against for a season, but over his career, he gives up too many home runs to keep his RA down.
Which brings me to Wins. Pitcher wins do have some bearing on how well the pitcher pitches... but only relative to the team hitting behind him. Last year, Derek Lowe won a lot of games, despite an ERA over 4... his teammates scored 7 runs per nine innings he pitched.
So, what about stealing bases? Based on some highly mathematical forumlas... if you succeed less then about 75% of the time (in today's game, the percentage has fluctuated between 67 and 75) then you COST your team runs over the course of a season. One thing that makes Derek Jeter more valuable then the average .300/.350/.450 hitter with 20 steals, is that he gets caught only once or twice for a season, with 20-25 stolen bases, meaning he actually creates several runs per season with his speed. On the other hand, a player with 25 steals, and 10 times caught stealing, costs his team 4-5 runs per season.
So whats the big deal over 4-5 runs? Well, 95% of teams fall into a certain formula for wins.
In a typical year, 1-2 teams fall outside of three wins of their expected win percentage based on that forumla. Lets say you are a very good team, and score 900 runs, while only allowing 700. This gives an expected win percentage of .623, or 101 wins... a very good year. What happens if you score 5 fewer runs? You've lost half a game. Not a big deal. But if, as a TEAM, each of your 9 regular players costs you 4-5 runs a season due to being caught stealing, you are down to 97 wins. 4 Wins lower then if you were smart, and never ran the bases, or maybe 5-6 wins lower then if 2 players stole at a high sucess rate like jeter does.
OVer the course of a season, 4 wins can often be the difference between the playoffs and "just another team." And this is just with the base numbers of 900 runs scored and 700 runs allowed. The number of wins can be very different with a different base set of numbers.
So... how about defense? How do we evaluate defense? The only statistics we have, traditionally, are Errors, fielding percentage, assists, and putouts... 3 of which are counting stats, which are not very useful, and one of which measures something very impresicely. Fielding percentage is based upon errors and total chances. Errors, however, is an inexact science. In an attempt to learn better how defense works, lets look at balls in play.
As a team, you'd hope to sucessfully turn 100% of balls put into the field of play into outs. However, this is unlikely. A very good defense will actually turn 73% or more balls into outs, while a poor one will turn 69% into outs.
On an individual level, you'd like to turn 100% of balls hit into your "zone" into outs. This gives us range factor. How many balls in your zone do you get to per game? An outfielder should handle 2-4 balls per game, and turn them into outs. As it turns out, for a right fielder, a range factor of 3 is pretty good. He turns three balls per game into outs. A shortstop is liable to have a range factor over 4.
But this still doesn't tell us very much. For example, does Derek Jeter play behind a ground ball pitching staff, or a flyball staff? If, as a team, the Yankees give up more ground balls then other teams, we'd expect Derek Jeter to have a greater number of chances per game then, for example, Miguel Tejada, meaning a higher range factor. Thus, range factor is still imprecise, but better then fielding percentage.
THere are other statistical measures, but none that you are likely to encounter as a baseball fan unless you go digging. What you should look for when evaluating a player, essentially, is how many balls he gets to that are hit into his area of responsibility, and then turns into outs. By this measure, Derek Jeter is a terrible shortstop, while Nomar Garciaparra is about average, and A-Rod is maybe 1-2 "davenport runs" per year better then average. Why is Derek Jeter so bad, when he looks so good?
The answer lies in understand the game beyond your own eyes. Derek Jeter does not to get to as many balls as he should, but makes plays at the edge of his range look spectacular.
Finally, ballparks effect how players play there. Fenway Park is a high scoring stadium, but, as Curt Schilling learned this winter before signing with the Red Sox, is not conducive to home runs. All offensive numbers need to be adjusted for the ball park, if we care to compare players. A-Rod benefitted from Texas Stadium, while Soriano was hurt by Yankee Stadium. This matters when evaluating the increases and decreases in team strength from that trade.
I hope you enjoyed that. I enjoyed writing it.
Posts: 4112 | Registered: May 2001
| IP: Logged |
posted
Sometimes, when I'm looking at my Excel spreadsheets, I am moved to tears by the heroism displayed by the Accounts Receivable column.
Posts: 37449 | Registered: May 1999
| IP: Logged |
posted
Paul, you and my fiance would be great friends. He's a huge fan of Rob Neyer and Bill James, and he read chapters from Moneyball to me every night when it came out.
It just makes you wonder how long it'll take the rest of the baseball world to figure it out. I mean, come on. Derek Jeter??
Posts: 2220 | Registered: Jun 1999
| IP: Logged |
posted
This is one reason why bseball is so beautiful. THere is this cerebral component that can fill up the off-season, and statistical analysis is still new enough that every few months, they have new insights published.
For instance, while Paul's pitching overview is true for your stock pitching type, it completely breaks down when trying to analyze knuckleballers, in fact, it tends to devalue them. Like the pitch itself, a knuckleball pitcher is as predictable as weather in March, in New England.
Also, while OPS is a decent correlation to runs scored, it's been found that if you weight OBP a bit more (by a factor of 1.1 or 1.2) the correlations is even closer).
posted
More like, I won't want to watch it while sitting next to some smart-aleck who gets all his stats from sports radio, and thinks Manny [EDIT: Manny Ramirez, that is] had an off year because he only knocked in 100 runs!!!
posted
Statistical analysis is cool and I appreciate both SABR and Paul. In fact, I started watching baseball as a kid because I wanted to understand where all the numbers came from on the back of baseball cards. But baseball is so much more than numbers. Baseball is about the Tigers beating the Yankees, the Yankees being the Yankees, and a full count with two outs and the bases loaded in the bottom of the ninth down by three.
Posts: 1295 | Registered: Jan 2003
| IP: Logged |
Printer-friendly view of this topic
![[Smile]](smile.gif)
