### Sunday, September 19, 2004

## Santana vs. Schilling

As this is the hot topic du jour, I thought I'd add a brief comparison of Johan Santana and Curt Schilling using defense-neutral expectancy rates. I used the same methodology in my Odalis Perez post, and it entails using the formula created by JC Romero for batting average on balls in play and the basic runs created formula (((H+BB)(TB))/(AB+BB)) plus a few other derived statistical rates like SLGFactor ((Park-Adjusted Doubles)+(2*Park-Adjusted Triples))/(Hits on balls in play)).

This time I adjusted for park factors on walks, doubles, triples, and home runs, so the results make everything defense-neutral as well as park-neutral with the exception of park effects on strikeouts (for a reminder of why those might be important, though I don't have the data to adjust for them, check out the latest entry of Rich Lederer's Abstracts from the Abstracts; I will note that both Santan and Schilling strike out batters slightly more frequently at home but neither by a wide margin). Using this formula, I came up with an expected run average of 3.89 for Curt Schilling and 2.75 for Johan Santana. Schilling, despite playing in the most extreme hitter's park in the AL, has actually outperformed his component stats in run average by .40 runs/9 IP. Santana, on the other hand, has a 2.82 RA, just slightly above his component stats' projection. Schilling has a FIP (Fielding Independent Pitching, ((13HR + 3BB - 2K)/IP) + 3.2) of 3.31 and Santana has a 2.97 FIP. Those numbers mask that Schilling gives up line drives at a clip well over league average and Santana gives them up at a rate well under league average. This effects not only batting average but Isolated Power, as Santana's given up 18 doubles and 3 triples versus 54 doubles and 2 triples for Schilling, a huge difference even when accounting for park factors (Schilling has allowed 22 walks on the road, Santana only 7).

I thought I'd see what that difference would look like when factored into each pitcher's support-neutral statistics. First, I figured out what their Expected Win-Loss records would be if I multiplied their expected winning percentage by their actual run average divided by component run average and then pretending they'd have the same number of decisions as expected by SNWL. This yielded a 12.5-10.3 record for Schilling and a 14.5-7.9 record for Santana. So while it's true that Santana has been unlucky relative to starters who have had a lot of luck, he hasn't been terribly unlucky (just unlucky enough that he might lose the Cy Young balloting); after all, Santana's offense isn't chopped liver, as 5.51 runs per 9 innings is about a half run over league average. Next, I used the same formula but with the team's expected win-loss with these players, and Santana's lead became to 20.8-10.2 to 16.1-13.9, a pretty huge advantage. Next, I adjusted support-neutral lineup-adjusted value added, and Santana's edge was 7.37 to 4.93.

You might have no idea what I'm talking about, so I'll give you the short version: there is only one statistic that supports Schilling's candidacy. That statistic (wins) is one that no pitcher in the American League can control. Even if a pitcher threw 15 innings a game with no hits and no walks, his team could lose and the pitcher could do nothing about it because he doesn't step up to the plate. Neither Schilling nor Santana has been throwing 15-inning perfect games, but Santana has been about as close as any pitcher in this era while Schilling has been merely good. The one statistic that supports Schilling's candidacy is the one statistic that the pitchers have the least control over, and Santana dominates Schilling across the board in all other statistics.

This time I adjusted for park factors on walks, doubles, triples, and home runs, so the results make everything defense-neutral as well as park-neutral with the exception of park effects on strikeouts (for a reminder of why those might be important, though I don't have the data to adjust for them, check out the latest entry of Rich Lederer's Abstracts from the Abstracts; I will note that both Santan and Schilling strike out batters slightly more frequently at home but neither by a wide margin). Using this formula, I came up with an expected run average of 3.89 for Curt Schilling and 2.75 for Johan Santana. Schilling, despite playing in the most extreme hitter's park in the AL, has actually outperformed his component stats in run average by .40 runs/9 IP. Santana, on the other hand, has a 2.82 RA, just slightly above his component stats' projection. Schilling has a FIP (Fielding Independent Pitching, ((13HR + 3BB - 2K)/IP) + 3.2) of 3.31 and Santana has a 2.97 FIP. Those numbers mask that Schilling gives up line drives at a clip well over league average and Santana gives them up at a rate well under league average. This effects not only batting average but Isolated Power, as Santana's given up 18 doubles and 3 triples versus 54 doubles and 2 triples for Schilling, a huge difference even when accounting for park factors (Schilling has allowed 22 walks on the road, Santana only 7).

I thought I'd see what that difference would look like when factored into each pitcher's support-neutral statistics. First, I figured out what their Expected Win-Loss records would be if I multiplied their expected winning percentage by their actual run average divided by component run average and then pretending they'd have the same number of decisions as expected by SNWL. This yielded a 12.5-10.3 record for Schilling and a 14.5-7.9 record for Santana. So while it's true that Santana has been unlucky relative to starters who have had a lot of luck, he hasn't been terribly unlucky (just unlucky enough that he might lose the Cy Young balloting); after all, Santana's offense isn't chopped liver, as 5.51 runs per 9 innings is about a half run over league average. Next, I used the same formula but with the team's expected win-loss with these players, and Santana's lead became to 20.8-10.2 to 16.1-13.9, a pretty huge advantage. Next, I adjusted support-neutral lineup-adjusted value added, and Santana's edge was 7.37 to 4.93.

You might have no idea what I'm talking about, so I'll give you the short version: there is only one statistic that supports Schilling's candidacy. That statistic (wins) is one that no pitcher in the American League can control. Even if a pitcher threw 15 innings a game with no hits and no walks, his team could lose and the pitcher could do nothing about it because he doesn't step up to the plate. Neither Schilling nor Santana has been throwing 15-inning perfect games, but Santana has been about as close as any pitcher in this era while Schilling has been merely good. The one statistic that supports Schilling's candidacy is the one statistic that the pitchers have the least control over, and Santana dominates Schilling across the board in all other statistics.