Friday, April 25th is the one-year anniversary of one of the most entertaining ten minutes of video I've ever encountered, an interview between MLB Network's Brian Kenny and Chicago White Sox TV play-by-play announcer Ken Harrelson. What started out as a pleasant conversation quickly devolved into Harrelson decrying the use of numbers in baseball and introducing the most interesting metric of the past 25 years, The Will To Win. I come here not to bash Harrelson or rehash an old argument but instead to remind consumers of sports statistics why we do it.
To state there has been an explosion in both data and its availability over the past 10-15 years is an extreme understatement, and combined with the processing power of computers allows for analysis only dreamed of as recently as 20 years ago. What used to be painstaking research either is already available or can be constructed with some technical skills. With great information comes great responsibility to both gather it correctly and properly interpret and explain it.
Since baseball was invented there has been one goal--to score more runs than the other team, and every effective metric ever devised or conceived attempts to measure how to best accomplish this. Fielding Independent Pitching isn't useful because it provides a new stat with which to bore our friends but because it very effectively shows how well a pitcher controls those variables directly in his control. Weighted On-Base Average better measures offensive production instead of relying on older values like batting average or slugging percent that don't provide the entire context.
If used correctly, statistics add nuance and understanding to what is seen on the field. For example, simply writing Pedro Martinez had an ERA+ of 200 or greater in five different seasons is rather dry--we need to know what ERA+ is, and even more importantly, how often other pitchers had similar seasons. Since 1901, there have been 34 seasons in which pitchers with 20 or more starts had an ERA+ of 200 or greater, which means Martinez had around 15 percent of these seasons. The years were 1997, 1999, 2000, 2002 and 2003--he accomplished this in the heart of the enhanced offensive production from around 1995-2007. The fact he "only" won 219 games in his career gets much more attention than this stunning achievement, which is a far better measure of his effectiveness and dominance.
Baseball stats can help properly evaluate a strategy to see if it really is effective. One of the most enlightening items I've come across is Tom Tango's Run Expectancy Matrix, available at tangotiger.net. This isn't data requiring doctorates in arcane areas of mathematics in order to interpret as much as serious data amalgamation and seeing how often something occurred during a given state. Harrelson discusses bunting, but when the Run Expectancy Matrix is reviewed, we can see the following--assume a runner on first and no outs (numbers are from the second chart):
Percent runner on first with no outs scored--44.1 percent
Percent runner on second with one out scored--41.8 percent
Sacrificing an out to move the runner over decreased the percent of time a run scored. The average number of runs scored also decreased from around .94 to around .72 (the first table). Does this mean the sacrifice bunt should be jettisoned as a strategy? Of course not--there are numerous circumstances where it is appropriate, and using advanced data merely quantifies the consequences and ramifications.
There are two typical responses by those who don't enjoy metrics--they ignore them, or they denigrate them. I don't have a problem with people who choose to ignore them, because not all people have a facility with numbers. To force people to bow at the altar of SIERA or UZR/150 would be similar to telling Brian Kenny how wonderful the pitching win is as a stat. I'm a firm advocate of using whatever a fan is comfortable with to enjoy the game, and a good analyst can take those facts and build on them in a friendly and informative way.
The second approach is to denigrate the metrics or distort what they show, for which there is no excuse. Baseball has constantly evolved as players, parks, conditions and economic circumstances changed, and to insist otherwise is to be a modern-day Luddite, stuck in an old way of understanding. No serious student of baseball metrics uses them to distort what is seen on the field--there may be misapplications or misinterpretations, but this happens in all walks of life. Sabermetrics hasn't and won't replace scouting, evaluation or on-field performance, but instead augment and supplement these functions, and to suggest otherwise is disingenuous at best.
The war is essentially over. Baseball front offices are in varying stages in the implementation and use of advanced metrics and are probably 5-10 years ahead of us in player evaluation. I wouldn't be surprised if measures like FIP and UZR are viewed as old hat as they have moved on to newer, proprietary measures we fans can only dream of. For example, the new MLBAM metrics will provide ways to measure fielding that can't even be imagined until the data is available. The general public won't be seeing it for at least two or three years, and no one knows in what amount. Every major league team will be far ahead of the general public.
In Game 5 of the 2013 ALCS, Detroit Tiger shortstop Jose Iglesias was shifted over near second as David Ortiz was batting in the third inning. This was an educated use of data, since Ortiz has a tendency to pull balls to right field. In this case, the Tigers shift appeared to fail, as Ortiz lifted a popup to short left field. However, Iglesias covered serious ground to catch the ball, one of the most amazing defensive plays I've ever seen, which the play-by-play reported as:
This play is one of the best illustrations of my point. Even though Ortiz had a tendency to pull balls to right, there's no guarantee he'll do it every time--probability doesn't always lead to actuality. The head-in-the-sand crowd would jump on this play as some "failure" of advanced metrics. Others see it as what we'll be able to measure when every play is analyzed and can't wait to see the possibilities. We know Iglesias made a spectacular play--what we really want to know is how often plays like this occur.
I used Troy Tulowitzki's picture to illustrate a very basic point--no baseball metric is the final answer, and in the case of WAR, not everyone can even agree how to calculate it--Baseball-Reference lists these as the best position players, whereas FanGraphs lists these. The number changes but the listed players are similar, suggesting an ability in WAR in to identify productive play. It's not perfect, and no one suggests it is, but merely provides a data-driven benchmark with which to evaluate players--isn't more information preferred to less or none?
I had this brief Twitter exchange with Cubs TV play-by-play broadcaster Len Kasper:
@ScottLindholm No, 1) Sundays are Wrigley 100 days & 2) we will incorporate a lot of the adv. stats into every b'cast. It's time.— Len Kasper (@LenKasper) April 20, 2014
Consider Len's answer very carefully--he didn't say he and broadcast partner Jim Deshaies would beat viewers over the head with advanced metrics but incorporate them naturally and with explanation. Some facsimile of the Run Expectancy Matrix is already used in Cubs broadcasts and it's not unusual to hear Len and JD discuss how much better a pitcher is pitching than his record indicates. The data and the metrics are there, why not use them to inform viewers? As he is on most everything, Len is right--it's time indeed.
Data from Baseball-Reference and FanGraphs
Scott Lindholm lives in Davenport, IA. Follow him on Twitter @ScottLindholm.