Baseball Strategy:Which stats matter?

The traditional statistics most people use to judge a hitter are severely flawed.

If you ever watch an ESPN telecast, you’ll see three numbers on a player’s graphic-overlay when they first come up to the plate: Batting Average, Home Runs, and Runs Batted In. These “Triple Crown” statistics have always played a big part in how a player is judged by fans, managers, and executives alike; they largely determine a player’s playing time, his role on the team, and even how much money he makes. Oh, and they’re not very indicative of how productive a player he is.

First of all, a lot of what we call “batting average” is plain ol’ bad defense. If simply making contact is the hardest thing to do in sports, then “hitting ‘em where they ain’t” is practically a Herculean task. Just think about the variables that go into just getting one hit: you have to put the bat on the ball, and fairly flush, to keep it in play; you have to swing at just the right level to avoid a pop-up or a harmless grounder; you have to weigh where the defenders are before the pitch and try to direct the ball into a gap... Ah, yes, the defenders: you could do everything right on your end of the swing, but an Andruw Jones or an Orlando Hudson could rob you with a play that you couldn’t make in your wildest fantasies... Conversely, you could hit a fly ball that would be an out with most outfielders, but since Marquis Grissom takes a downright bizarre route to the ball, it falls in for a “hit”. This is why batting average has such a high variance from year to year. As much as it pains me to say so, pure luck can make a true “.300 hitter” hit .270, or vice-versa. So why do stat-heads love On-Base Percentage so much? It’s basically a player’s walk rate tacked on to Batting Average. Since walk rates don’t vary much from season to season for most players, OBP’s are far more stable indicators of a player’s skill, because BA’s can fluctuate so wildly from year to year. OBP also correlates better to run-scoring than BA, because it takes into account more information -- hits and walks.

Home runs are fairly important, but without context they lose meaning as well. At which park were they amassed? 20 HR at RFK Stadium or Petco Park may be pretty good; 25 HR at Coors Field may not even be adequate. In how many plate appearances were the home runs hit? In 2005, Jeff Francoeur hit 14 HR in 274 PA’s; Mark Kotsay hit 15 in 627 PA’s... which player was the better power hitter? Well, Francoeur, obviously, but you wouldn’t know from ESPN’s overlay.

That said, RBI’s are the worst. Remember Tony Batista’s 103 RBI two years ago? Yeah, neither does anyone else; he spent the following year with the Fukuoka Hawks of the Japanese League. At least BA and HR are relatively individual stats; RBI is almost totally dependent on your spot in the batting order (a lot of guys put up big RBI numbers simply because of the sheer number of plate appearances they receive), and, even more so, who your teammates are -- specifically, how well the guys batting ahead of you get on base. You could be the best hitter in the world and post low RBI totals because nobody’s setting the table, making RBI perhaps the most overrated, overvalued stat in all of sports.

• If RBI Are Overrated, Runs Scored Are Underrated.

The corollary to the “RBI are overvalued” statement is that Runs Scored by a player are not valued enough. While they also depend heavily on teammates, guys with a lot of RS are getting on base a lot, and they’re not getting caught often on the basepaths -- two things that players must do for an offense to perform well.

Pitcher Win-Loss Records Are Basically Meaningless.

The problem with pitcher W-L records is that they depend so heavily on things that have nothing to do with pitcher himself -- run support, strength of opponent, the defense behind him, even pure luck -- that they should never, ever be used to determine a pitcher’s value. Bottom line, end of story.

Beware Small Sample Sizes!

In statistics (the mathematical discipline, that is), sample size is the term used to describe how much information is being used to arrive at a conclusion. The smaller the sample size, the more likely it is that the outcome is the result of chance; the larger the sample size, the more likely it is that the outcome reflects reality.

Sample size comes up all the time in baseball. Announcers will frequently make comments about how a certain batter “owns” a pitcher because of his good numbers in head-to-head matchups between the two, but the career head-to-head numbers are almost always amassed over a very small number of at-bats -- a statistically insignificant sample size in which luck plays a bigger part than skill.

Early-season “assessments” often rely heavily on minuscule sample sizes. To wit: Beane's Oakland Athletics have been slow starters in recent years, usually causing sportswriters to declare the Moneyball paradigm DOA... But the numbers of April and May still represent a small sample relative to the entire season, and, sure enough, the A's always perform up to expectations over the entire season. The 2005 Orioles were the opposite of this: they started red-hot (read: they were a fluke), and many fans were shocked when they cooled off big-time in June and finished the year with only 74 wins. Again, the problem was sample size; those early performances were not indicative of the team’s true ability, and only over a full season’s worth of games did they perform more like what was expected before the season. Not that that’s of any consolation to Orioles fans. Assuming there are any, that is.

Minor League Stats Are Meaningful.

I don’t know why it was always assumed that they weren’t; I guess GM’s were burned too many times by hot-hitting AAA players who cooled off immediately upon joining the big club. The problem is, minor-league stats are only significant if you consider context (come to think of it, all statistics are only significant if you consider context). Again, ballparks must be considered, as their distorting effects are felt even more heavily at the minor league level than in the majors. In addition, the league that a player plays in must be adjusted for -- the runs-per-game context of the offense-crazed Pioneer League, for instance, is vastly different than that of the pitcher-friendly Florida State League. Finally, a player’s age relative to his competition must be considered; a 25-year-old putting up big numbers at AA may not even be an ML prospect, while a 20-year-old with the same numbers may be a can’t-miss future star. After you account for all of these variables, though, minor league stats can be as good a predictor of future performance in the majors as past ML stats are for players already in the bigs.