Michael/Male/26-30. Lives in United States/Pennsylvania/Wexford/Christopher Wren, speaks English. Spends 20% of daytime online. Uses a Fast (128k-512k) connection. And likes baseball /politics.
This is my blogchalk:
United States, Pennsylvania, Wexford, Christopher Wren, English, Michael, Male, 26-30, baseball , politics.

Tuesday, April 17, 2007

Glove Week!, Part II: Major Statistical Systems 

Yesterday I talked a little about the evolution in thinking about defense that brought us up to the modern era. Sabremetrics has really exploded over the last four years since the publishing of Moneyball in 2003, which stunned the baseball establishment with a provocative assault on the conventional wisdom that few had evidentially been paying attention to. While things like Runs Created and Defense Independent Pitching (DIPS) have pre-dated the Moneyball era, the numerous sabremetric fielding statistics that have come down are all largely post-Moneyball.

The focus on fielding is largely a product of the great baseball minds moving on from hitting and pitching and finally giving this unknown part of the game its due. Bill James pondered in ’05 on what he’d write about if he reissued the old baseball abstracts, Bill James pondered in ’05 on what he’d write about if he reissued the old baseball abstracts, he answered: “Baserunning and fielding. I know that I've spent more time worrying about fielding in my career than I ever have about hitting, but that's because we started out so far behind and that is still true. We're still way behind on fielding and baserunning. We ought to do better.”

James has focused more of his attention in the recent past to those two fields of study. Base-running information has been in the last two Bill James Handbooks (see, 2006 Handbook; 2007 Handbook), but the big focus is on fielding stats: Bill James recently reworked his old Range Factor into Relative Range Factor, an improved version of his old stat, and heavily participate in The Fielding Bible, a publication issued from Baseball Information Solutions (BIS) and John Dewan, a friend and colleague of James for years. Fielding is a difficult subject, which I think Bill James summed up best by likening hitting to a solid, pitching to a liquid and fielding to a gas, stating that fielding was, like a gas, “formless and hard to see”. As I noted in Part I, Branch Rickey threw up his hands in frustration with fielding stats in the 1950’s when he ran the Brooklyn Dodgers. It was an area of the game that he felt he lacked the tools to attempt to quantify. Now we do, thanks to computers, have some tools to tackle what Alan Schwarz referred to in The Numbers Game as the “holy grail” of baseball statistics … Fielding.

Here is an overview of the systems used …

I. Range Factor (RF). The first stat that really got us focused on fielding. Proposed in the ’77 Baseball Abstract, RF takes the number of successful plays that a player makes (Putouts and Assists), multiplies them by nine and divides by innings played thus giving you the number of plays a player makes in a nine-inning game, much like how ERA gives you the runs a pitcher would surrender in a nine-inning game:

((Putouts + Assists) * 9) / IP

The obvious flaw with Range Factor is that pitching greatly impacts your numbers. Are you a shortstop? If you play on a team with a lot of groundball-oriented pitchers, you are going to have some nice stats. If you play on a team of fly ball-oriented pitchers, you are going to come out behind. Richie Ashburn posted nice Range Factor numbers during the 1950’s (usually a full play better per game than the league average), but he played on teams that featured Robin Roberts, a pitcher who allowed a lot of fly balls. Ashburn was a great fielder, but that must be taken into account.

James created RF and let it hang out there for years without updating it … until now.

II. Defense Efficiency Ratio (DER). Two years after he gave us Range Factor, James proposed Defense Efficiency Ratio. DER simply put, measures how often the fielders behind a pitcher convert balls put into play into outs:

(Batters Faced – (Hits + Walks + Hit By Pitch + Strikeouts)) / (Batters Faced – (Home Runs, Walks + Hit By Pitch + Strikeouts))

The advantage DER has over RF is obvious: it is more of a pitcher-neutral stat. A shortstop playing behind a fly ball pitcher will watch enviously as the balls go sailing into the gloves of his friends in the outfield. DER rates the whole team. This is a way of measuring team defense that is much better than double plays turned or errors or team fielding percentage.

Naturally, no system is perfect. Pitchers impact DER because the types of the balls that they allow to be put into play vary. Hard-hit balls that drop in for line-drives are virtually uncatchable, and thus, not fairly attributable to the fielders. The other thing is that DER evaluates the whole team, how the team’s defensive alignment works, not individuals like RF.

III. Fielding Runs (FR). Developed by Pete Palmer as the fielding component of his Total Player Rating (TPR), Fielding Runs attempts to look at a player’s fielding stats and assign a number stating what their contribution to the team was with their glove. I’d refer people to The Numbers Game, specifically to Chapter Eight, where the author talks a little about Palmer and his work. Fielding Runs is fairly complicated. First calculate the league average for players in that position:

.20 * (Putouts + (2 * Assists) – Errors + Double Plays) / League Putouts – League Strikeouts)

Then … once you have the league averages, plug in the individual player’s stats …

.20 * (Putouts + (2 * Assists) – Errors + Double Plays) – ((Team Putouts – Team Strikeouts) * (Player Average * percentage of playing time)

Not a perfect system mind you. Even the author of The Numbers Game, Alan Schwarz, noted the improbability of Glenn Hubbard setting an all-time record for 62 Fielding Runs in 1985 (see, Pages 167-168). I noticed several comments - specifically the sections dealing with Granny Hamner and Bill Buckner - from Bill James assailing Fielding Runs in his Historical Baseball Abstract. (See, Part IV of this series.) Critics have cited problems with Fielding Runs such as the fact that Palmer awards a double value to assists over putouts based on his belief that it takes “more fielding skill” to make an assist over a standard putout. Critics have also noted that Fielding Runs doesn’t attempt to be pitcher-neutral. i.e., Outfielders will rate well on teams that have fly ball pitchers, while infielders look like Ozzie Smith when they play behind guys like Tim Hudson. It is just not an accurate system.

IV. Zone Rating (ZR). Developed by Dick Cramer, Zone Rating analyzes how often balls are hit towards a fielder’s location and how often those balls were converted into outs when hit into their zone. The data of how Zone Ratings are calculated is pretty tough to decipher. How accurate is it? Tough to say. Theoretically, ZR deals with an issue that RF and FR don’t: pitcher-neutrality, however it is something where someone is making a judgment about whether or not someone ought to have made a play on a ball hit near them and thus we run dangerously close to what Bill James spoke of when he talked about fielding stats being a record of opinions. (For more on Zone Rating, click here.)

Note: Zone Rating has been redone by John Dewan to clear up many of the problems associated with it and The Hardball Times is now posting the ZR numbers on their website.

V. Ultimate Zone Rating (UZR). Mitchel Litchman developed UZR by taking ZR and converting it into runs saved or runs allowed. Again, there is no public data on UZR (and in any case, Litchman has stopped publishing UZR last I checked), so it is difficult to sit down and objectively see the numbers, however, UZR does try and correct for ZR’s flaws by taking park factors, a pitcher’s groundball and fly ball ratio and other factors into account. (For more on UZR, click here.)

VI. Probablistic Model of Range (PMR). Developed by David Pinto of Baseball Musings, PMR takes balls put into play and analyzes their probability of being caught. Then PMR takes what each player successfully does and combines them. If a player’s efficiency in the field (their DER) is greater than what they were expected to do under the system, then they have a positive PMR rating. If it is the reverse, then so be it. (Click here for a more complete discussion of PMR.)

VII. Defense Regression Analysis (DRA). Developed by Michael Humphreys, DRA tries to analyze players utilizing traditional defensive statistics, unlike PMR or UZR, which are zone-based systems that look at what players are doing through a modern lens. (Click here for a more complete discussion of DRA.)

VIII. Fielding Win Shares (FWS). Bill James introduced Fielding Win Shares as part of his Win Shares system in 2002. The system is complicated, so I am going to refer anyone interested in seeing how James calculates Fielding Win Shares to The New Historical Baseball Abstract, specifically, pages 350-358. The system figures heavily in James player ratings for the Abstract.

IX. Relative Range Factor (RRF). In 2006 Bill James revisited Range Factor by deciding to smooth out the issues that plagued it, namely pitcher’s groundball / fly ball ratios, left-handed / right-handed issues with pitchers, etc. Contained in The Fielding Bible (specifically, pages 199-209), James works on all of those problems and even comes up with a nifty idea of multiplying the number with the league average so that a number higher than 1.000 is better than average and .999 and lower is below-average.

X. Plus / Minus. Developed by John Dewan, Plus / Minus is the cornerstone of The Fielding Bible, Dewan’s 2006 book that revolutionized fielding stats in the blogging community, though it generated barely a ripple in the mainstream media. Simply up, Plus / Minus takes the plays that a fielder makes and washes it through a computer system. In the example Dewan includes on page 9 of The Fielding Bible, Dewan notes that a ground ball hit on vector 17 to a shortstop is successfully fielded 26% of the time, so if the shortstop makes the play, he is credited +.74 (1.00 minus .26 for the expectation that the play is made). If he misses it, it is -.26, which seems like a fair result. Fail to make a low-percentage play and absorb a little loss, make a low-percentage play and get a lot of credit. It is insanely complicated and not capable of being dissected by the lay fan.

Those generally are overviews of the systems talked about. Naturally I am forgetting a few, but those are the biggies. Now we’ll move on a discuss the Phillies a little.


I just wanted to thank you for your review of "Eight Men Out"! I checked out a copy from the library after reading your piece and couldn't put the book down!

What struck me the most was how the scandal ruined the lives of Buck Weaver, Shoeless Joe, and Kid Gleason among others. Yet, the gamblers and shady characters who manipulated the Sox never faced any serious consequences for their actions.

A lot of people seem to look at baseball's past with rose colored glasses. They are repulsed by the current steroid scandal and Barry Bonds. Yet, baseball has always had its fair share of controversy as the Black Sox scandal so painfully demonstrates.

Thanks Again!
Post a Comment

This page is powered by Blogger. Isn't yours?