The process of evaluating offensive talent has been carried out for over a century, since the beginning of organized baseball. In the early years of baseball, when power was essentially a non-factor, batting average was all you really needed. The batters with the highest batting average were the best hitters in the league. But when guys like Babe Ruth came around and started hitting 400-foot home runs with regularity, batting average was no longer sufficient. Why should a home run count the same as a single?
Several decades and several hundred slide rules later, people have come up with better ways to evaluate how good hitters are at doing their jobs. I just skipped a whole lot of baseball history, but if you want a great review of the history of baseball statistics, I suggest you read The Numbers Game, by Alan Schwarz.
Today, there are dozens of run estimators available– some good, some not so good. In my opinion, the best and most logical of these run estimators is called Weight On Base Average, or wOBA for short. The logical derivation for wOBA can be found here. That article mentions the “run values of each event,” without clearly defining what those values are. Those run values can be found here, by scrolling down to the bottom. These stats will come courtesy of FanGraphs.
Another aspect of offensive talent, albeit less talked about, is base running. I might not always include base running in evaluating players, since it’s usually a very small number, either positive or negative. The fastest players in the league rarely get more than 10 runs out of their speed, and most everybody else falls within a few runs of average. Base running numbers will come courtesy of Baseball Prospectus. I will include base running if the player in question is especially fast or especially slow. Otherwise, it just doesn’t make much of a difference, and any variation away from average could just be statistical noise.
While looking at last year’s offensive numbers can give us a pretty good idea of what a player will to next season, one year of data is subject to huge variation–it is fluky, if you will. For example, look at the career stats of Darin Erstad. From 1999-2001 there was massive variation around what his true talent actually is. One year he was below average, the next year he was an MVP-candidate, only to go back to being below average the following season. Here is where projection systems come in. The best publicly available projection system is that of Chone Smith, which he calls CHONE (Note: his name isn’t actually Chone). This can be found on FanGraphs as well (once he runs the 2009 data). The simplest projection system is Marcel, named after Marcel the Monkey because of its simplicity. Tom Tango is the “care-taker” of this system, but promotes it as the simplest projection system that other systems should be judged against (essentially it’s a baseline for other systems). Marcel takes 3 years of weighted data, regresses to league average, and applies an age adjustment. That’s it. CHONE is a little more complicated, in that it finds historically comparable players for each player it is projecting. A projection is our best estimate of what a player’s true talent is, going forward. And that is what we are really after.
Remember this: Runs are used as a proxy for wins, since in the end all we really care about is winning games. In the current run environment, every 10 runs added is approximately equal to one win. So a player worth 30 runs above average is worth 3 wins above average. Here’s why: “The basic idea is that if you look at all teams in baseball history that have scored 1 more run than they allowed, per game (+/- 0.1, to increase the sample size), you will find that they have a .600 win%. And similarly if they allowed one more run than they score, they will have a .400 win%. That means each additional run leads to 0.100 additional wins, above the .500 mark. And 1 divided by .1 is 10,” (quoted from TangoTiger chat, FanGraphs 12/10/08).