Sunday, May 24, 2020
Catcher WOWY
In 1979, in games started by Gary Carter, the Expos allowed 463 runs and made 3705 outs, for a runs per 27 outs of 3.37. That season, in games started by his backups, the Expos allowed 5.08 runs per 9 IP. That difference is a whopping 1.71 runs per 9 IP, or 66% as much as his mates.
However, Carter had 136 starts compared to the 24 from his mates. In terms of the WEIGHT of that 1.71 difference, we would use the harmonic mean, which is 41 games (or 1073 outs).
We can go through every season of his career and repeat this process. And in his career, his weighted runs allowed is 86% of that of his mates (3.66 RA / 9IP with Carter starting compared to 4.28 when his mates start). In other words, his team has allowed 0.62 fewer runs to score per 9 IP, when Gary Carter started the game behind the plate, compared to his mates that season.
Gary Carter started 1954 games (or if you use outs, the equivalent of 1964 9-inning games). And 0.62 runs per game times 1964 games (unrounded) is 1204 fewer runs allowed. That is the best figure in the last 100 years.
Using this method, here are the top 11:
- 1204 runs reduced: Gary Carter
- 1004: Mickey Cochrane
- 961: Tony Pena
- 871: Yadi Molina
- 807: Brad Ausmus
- 757: Andy Seminick (who??)
- 750: Rick Ferrell (who??)
- 701: Johnny Bench
- 682: Jason Varitek
- 652: Al López (who?? he did have MVP votes in 7 seasons)
- 647: Russell Martin
Fans of WOWY will recognize this as...well, WOWY.
The main problems, at least for THIS iteration, are as follows:
- Are the backups of Gary Carter disproportionately worse than the backups of Cochrane and Pena and so on?
- Was Gary Carter paired disproportionately with the best pitchers on his staff, compared to his mates?
- How much can Random Variation affect these results?
The Random Variation one is a big one. In the Gary Carter starts, his teams allowed 7246 runs. And since this is 1204 lower than his mates, his mates would come in at a pro-rated 8450 runs. Just because we've OBSERVED 8450 (pro-rated) runs doesn't mean that's the true rate. How much Random Variation is there in runs allowed? I should figure it out at this point, but let's say it's one standard deviation is 3 runs per game. With 1964 starts, you take the square root and multiply by 3 runs and that gives you 133 runs. Since we have observations on both Carter and his mates, the Random Variation of the DIFFERENCE is 133 times root 2, or 188 runs. So, based on that number 3 that I totally made up, we can reduce the 1204 by twice 188 runs due to Random Variation (for 2 standard deviations). That still leaves us with a whopping 828 runs lowered. So, whatever you can say RAndom Variation contributes to the noise in that 1204 runs, it won't be able to wipe away even half of it. Most of it is real.
Anyway, that's all I've got for Iteration 1. If an Aspiring Saberist wants to take it from here, go for it.
***
Note: reason I started this was because of Ryan Doumit, since he was getting hurt by the framing numbers. Using this method, his teams allowed 188 more runs. And if you go all the way to the bottom of his Fangraphs page:
And look right under the Defense column, all the way to the last line, you will see this number: -178 runs. In other words, this WOWY method (188 runs) supports the evaluation of Doumit and his framing numbers (178 runs)!
***
Note 2: since someone will ask, Mike Piazza was better than his mates by 322 runs. Make of it what you will. I just know whatever number this method spit out, someone is going to complain.