The total zone data presented here was designed and calculated by Sean Smith (BaseballProjection.com). Sean has produced, in our opinion the finest defensive metrics around utilizing play-by-play data. These measurements compare favorably with direct observation techniques such as Ultimate Zone Rating and Plus/Minus. Since these use play-by-play data, we can apply them to seasons all the way back to the dawn of the retrosheet era. Below is Sean's explanation of what is included on the site.
There are three methods used to calculate the range portion of TotalZone, and depend on how much data is available from the Retrosheet files. The first is used when there is limited information on who fielded each hit, and whether the hits were grounders, flies, line drives, or popups.
For most games, I have information on which fielder makes each out, and the batted ball type. Without information on the hits, I have to make an estimate. I look at each batter's career rates of outs by position. For example, if 30% of a batter's outs are hit to shortstop, then every time that batter gets a hit the shortstop is charged 0.3 hits. Repeat for every position. I look at batting against righthanded and lefthanded pitching separately, as switch hitters will have very different ball in play distributions depending on which side of the plate they hit from. I sum the fractional hits for every fielder, combine with plays made and errors, and get a totalzone. This is then park adjusted, and converted to runs. This method is used for all seasons before 1989, and for the dark years of 2000 to 2002.
The second method is used when hits are coded with a batted ball type and we know who fielded each. The responsibility for ground ball singles hit to left field is split between the third baseman and shortstop, for center field it is between the shortstop and second baseman, and for right field, the second and first basemen. Groundball extra base hits are charged to the first or third baseman. Outfielders are charged with line drive and fly ball hits that they field. The responsibility split between infielders was originally 50/50, but has been refined based on more detailed analysis. Singles to left are charged 60/40 to third and short, to center it is 52/48 between short and second, and to right it is 55/45 first base/third base.
Players are compared to league average for specific parameters, such as batted ball type (for outfielders, line drives are much harder to field), pitcher and batter handedness, and for infielders, whether or not there is a runner on first. Results are park adjusted and converted to runs. This method has been used for years 2003 to 2008.
Finally, for years 1989 to 1999 Retrosheet includes the project scoresheet zones. This allows more detailed assignment of hits, and when comparing to the league average, I compare to the league average for that particular zone to account for difficulty.
Runs estimates are constant using run values from recent seasons. While this may overestimate the value of a hit saved in seasons with a low offensive context, it does allow us to compare fielders from different eras.
Outfield arm, Infielder double play, and catcher data are the same for all seasons.
For outfield arms, I look at an outfielder's rate of holding, allowing advancement, and throwing out baserunners compared to league average for several situations. The situations include runners tagging from bases, advancing first to third or second to home on singles, and first to home on doubles. It also accounts for miscellaneous assists (batter out at first, or trying to advance) and advancement errors.
Infield double plays look at DP rate when there is a runner on first, less than two out, and a groundball that is fielded by an infielder. Rates are compared to league averages and converted to runs.
Catcher data looks at stolen bases allowed, caught stealing, errors, pickoffs, passed balls, and wild pitches. I split the data by pitcher handedness (otherwise a catcher will look better if he catches more lefthanders than normal.) Once again, everything is compared to league averages and converted to runs.
Total defense would then be the sum of the above calculations. Note that these values are relative to average rather than relative to a replacement level.