What is Sabermetrics?
Ever wonder just what the statheads mean when they talk about sabermetrics? James Fraser puts you in the know and helps fill in some details of what we hope to accomplish here.
Sabermetrics is loosely defined as the systematic or scientific
study of baseball statistics. What distinguishes sabermetric writing
(Rob Neyer, or
Baseball Prospectus, or
us) from regular
the manner in which statistics are used.
Sportswriting uses statistics to prove a specific point; often using the wrong statistics to prove a point that otherwise makes no sense. Sabermetrics, ideally, starts with no argument, but rather with a point of discussion that can lead to any conclusion. Sabermetrics looks at any aspect of baseball; from eras, seasons, games, and pitches to leagues, teams, and, finally, to players and offense, defense and pitching.
The principles behind sabermetrics are simple: find something about baseball that merits examination, collect as much information as possible, examine all the evidence as objectively as possible and then allow others to criticize it. The final step is key. We, as sabermetricians, must allow others to criticize our work; without openness to criticism we become as close-minded as the sportswriters we try to avoid.
Sabermetrics is about answering questions, baseball questions.
Below, I will introduce some of the methods that you’ll be seeing us use when we discuss the answers to baseball questions.
Answering Questions About Offense
One of the most common applications of sabermetrics is offense evaluation. A team’s offense is comprised of two parts: the ability to get players on base (avoiding outs) and the ability to move those players around the bases. There are several levels of complexity to offensive evaluation methods, but all of them rely on measuring those two facets of offense: OnBase and Advancement.
The most often used statistic to provide a quick assessment of offense is OPS (Onbase Plus Slugging). This simple addition of a statistic to represent OnBase (On-Base Percentage) and Advancement (Slugging Average) ability, has little theoretical backing. However, as a quick judge of offensive ability it works just fine.
The next level of complexity is run based offensive evaluation methods. These measures attempt to quantify a player or team’s runs contributed by taking traditional statistics and combining them in a logical manner. The most basic run measure is Bill James’ Basic Runs Created, which simply combines the two facets of offense in the context of opportunity:
Runs = (OnBase)*(Advancement)/(Opportunity)
In James’ model, “Onbase” was simply hits and walks, “Advancement” was total bases, and “Opportunity” was plate appearances. This model worked well; most of the time it could estimate a team’s runs scored within 25 runs. However, as James tried to increase the accuracy of his model, it began to lose its theoretical basis.
More accurate models have been created using linear regression. These models apply values to each offensive event to quantify its effect on both OnBase and Advancement, while taking Opportunity into account. The favourite here at the Primer is Jim Furtado’s eXtrapolated Runs (XR):
Runs = .5*(singles)+.72*(doubles)+1.04*(triples)+1.44*(home runs)+.34*(hit by pitch+walks-intentional walks)+.25*(intentional walks)+.37*(sacrifice flies)+.04*(sacrifice hits)+.18*(stolen bases)-.32*(caught stealing)-.09*(at bats-hits-strikeouts)-.098*(strikeouts)-.37*(grounded into double plays).
Once a players runs contributed is estimated accurately by XR, the figure needs to be translated into how it helps his team, or would help any team. Pete (Total Baseball) Palmer uses runs above average by essentially figuring whether or not a player produces more runs than average per plate appearance. Others use outs as the baseline, comparing a player’s runs/out versus the league rate. Here at the Baseball Primer, we will be using a method that takes both the amount of opportunities a player had (plate appearances) and the amount of opportunities he added or subtracted for his teammates (outs). The method is eXtrapolated Wins (XW), (For an explanation click on this.link). XW measures a player’s win contribution above what a replacement (or AAA) player likely would have done in his place.
Answering Questions about Pitching
Recent work on pitching has focused on separating pitcher ability from results. While ERA is a far better at evaluating pitching performance than most batting stats are at measuring hitting, in some cases a pitcher’s ERA does not tell us how good (or bad) a pitcher really is.
Several methods such as Bill James’ ERC attempt to compensate by estimating what a pitchers’ ERA should have been based on his hit, walk, homerun and strikeout totals. Recently, fellow Baseball Primer contributor Voros McCracken’s Defense Independent Pitching Stats, which attempts to separate defense from pitching has garnered much attention. I recommend reading his explanation, it is truly one of the more important recent works in sabermetrics.
A great deal of sabermetric attention has focused on pitcher usage. Studies about pitcher use and abuse are difficult to design; however, I’m sure you’ll be reading a lot about it in the upcoming months.
Answering Questions about Defense
There is a lot of work to be done in the area of evaluating defense. Firstly, the line between pitching a defense, despite Voros’ work, is blurry. Secondly, the best data for evaluating defense, Zone Ratings and Defensive Averages, exists for less than 20 years of baseball’s history.
You’ll see a lot of work done with range factor (which is simply plays made per game) and zone rating (outs made divided by opportunities to make outs). The best work on defense is currently being done by Diamond Mind Baseball and on Fanhome.com
Adjusting for Context
One of the most important aspects of sabermetrics is the constant struggle sabermetricians face to accurately represent a player’s ability and value by adjusting for era and park. We watch baseball games today in an extreme offensively oriented era, and in Colorado, we have the opportunity to see the one of the most extreme parks in baseball history. However, a sabermetrician will never be won over by gaudy numbers put up in Coors field, because he looks at the context which the performance occurred in.
How important are park effects? Well, that’s an area for future research, but it is essential to remember that we must view performance through the different lenses that distract us from isolating true ability and value.
Team performance is also studied by sabermetrics. Studies ranging from batting order design to bullpen usage have been undertaken. One of the most important methods used when examining team performance is the Pythagorean Method of estimating Won/Lost Percentage:
Winning Percentage = (Runs Scored)^2/[(Runs Scored)^2+(Runs Allowed)^2]
The goal of sabermetrics is to answer baseball questions. How do we go about doing that? It’s like designing a scientific experiment. We collect data (usually statistics), analyze it and interpret the results. There is a lot of data to be analyzed: team records, player and pitcher records, play-by-play and even pitch-by-pitch data.
Anything can be and is studied: player projection and evaluation, park and era adjustments, defense and pitching, clutch ability, win probability, reliever use patterns, trade, drafting and managerial strategies, and anything else!
The rule for a successful study is simple: ask yourself a question and then try to answer it. We are always interested in looking at sabermetric studies, so drop us a note if you’ve looked at something that you think is worth printing or if you need a hand trying to design a study.
In future articles, you’ll be seeing more about the history of sabermetrics as well some sabermetric content here at the Baseball Primer. We won’t be overloading you with numbers, but we will be answering as many interesting baseball questions as we can. Above all, we will not be afraid to admit we are wrong and take another objective look at any established theory of baseball.
Posted: March 15, 2001 at 06:00 AM | 0 comment(s)
Login to Bookmark