Methodology PECOTA

1 methodology

1.1 comparable players
1.2 peripheral statistics
1.3 probability distributions
1.4 team effort

methodology

silver described inspiration approach follows:

the basic idea behind pecota fusion of 2 different things – [bill] james s work on similarity scores , gary huckabay s work on vlad, [baseball prospectus s] previous projection system, tried assign players number of different career paths. think gary used thirteen or fifteen separate career paths, , pecota doing carrying logical extreme, there separate career path every player in major league history. comparability scores mechanism picks , chooses among career paths.

comparable players

pecota relies on fitting given player s past performance statistics performance of comparable major league ballplayers means of similarity scores. described in baseball prospectus website s glossary:

pecota compares each player against database of 20,000 major league batter seasons since world war ii. in addition, draws upon database of 15,000 translated minor league seasons (1997–2006) players spent of previous season in minor leagues. ... pecota considers 4 broad categories of attributes in determining player s comparability:

1. production metrics – such batting average, isolated power, , unintentional walk rate hitters, or strikeout rate , groundball rate pitchers.

2. usage metrics, including career length , plate appearances or innings pitched.

3. phenotypic attributes, including handedness, height, weight, career length (for major leaguers), , minor league level (for prospects).

4. fielding position (for hitters) or starting/relief role (for pitchers). ... in cases, database large enough provide meaningfully large set of appropriate comparables. when isn t, program designed cheat expanding tolerance dissimilar players until reasonable sample size reached.

pecota uses nearest neighbor analysis match individual player set of other players similar him. although drawing on underlying concept of bill james similarity scores, pecota calculates these scores in distinct way leads different set of comparables james method. furthermore, silver describes following distinct feature:

the pecota similarity scores based on looking @ three-year window of pitcher’s performance. thus, might @ pitcher did ages 35–37, , compare against similar age 35–37 performances, after adjusting parks, league effects, , whole host of other things. different similarity scores might see @ baseball-reference.com or in other places, attempt evaluate totality of player’s career given age.

once set of comparables determined each player, future performance forecast based on historical performance of comparables . example, 26-year-old s forecast performance in coming season based on how comparable major league 26-year-olds performed in subsequent season.

separate sets of predictions developed hitters , pitchers.

peripheral statistics

pecota relies lot on use of peripheral statistics forecast given player s future performance. example, drawing on insights coming out of use of defense-independent pitching statistics, pecota forecasts pitcher s future performance in given area using information past performance in other areas. baseball analyst , journalist alan schwarz writes, silver ... designed sophisticated variance algorithm has examined every big-league pitcher s statistics since 1946 determine numbers best forecast effectiveness, earned run average. findings counterintuitive fans. when try predict future e.r.a. s past e.r.a. s, re making mistake, silver said. silver found predictive statistics, considerable margin, pitcher s strikeout rate , walk rate. home runs allowed, lefty-righty breakdowns , other data tell less pitcher s future .

probability distributions

instead of focusing on making point estimates of player s future performance (such batting average, home runs, , strike-outs), pecota relies on historical performance of given player s comparables produce probability distribution of given player s predicted performance during next 5 years. alan schwarz has emphasized feature of pecota: separates pecota gaggle of projection systems outsiders have developed on many decades how recognizes, flaunts, uncertainty of predicting player s skills. rather generate 1 line of expected statistics, pecota presents 7 – optimistic, pessimistic – each own confidence level. system resembles forecasting of hurricane paths: players can go in many directions, preparing 1 foolish . silver has written,

this procedure requires become comfortable probabilistic thinking. while majority of players of type may progress way – say, peak – there exceptions. moreover, comparable players may not perform in accordance true level of ability. appear exceed in given season, , other times fall short, because of sample size problems described earlier.

pecota accounts these sorts of factors creating not single forecast point, other systems do, rather range of possible outcomes player expect achieve @ different levels of probability. instead of telling s going rain, tell there s 80% chance of rain, because 80% of time these atmospheric conditions have emerged on tuesday, has rained on wednesday.

surely, approach more complicated standard method of applying age adjustment based on average course of development of players throughout history. however, leaps , bounds more representative of reality, , more accurate boot.

team effort

although silver creator of pecota, producing pecota forecasts team effort: might pecota guy, team effort, silver has said of bp staff. it. s baby, takes village run pecota . example, pecota draws on clay davenport s translations (the so-called davenport translations or dt s) of minor league , international baseball statistics estimate major league equivalent performance of each player. in way, pecota able make projections more 1,600 players each year, including many players little or no prior major league experience.

the 2009 preseason forecasts last ones silver took primary responsibility. in march 2009, silver announced pecota s extremely complex , laborious set of database manipulations , calculations moving different platform. although baseball prospectus had been owner of pecota since silver sold them in 2003 – , silver stewarded , took responsibility forecasts – henceforth pecota forecasts generated baseball prospectus team, clay davenport in charge of effort, , later, through 2013 season, colin wyers heading both production , improvements in pecota.