Little-known Roy Halladay facts (with apologies to a list of Chuck Norris facts):
– You don't decide your destiny, Roy Halladay chooses it for you.
– Roy Halladay makes mission impossible….possible.
– Nothing beats a failure but Roy Halladay.
– Roy Halladay can divide by zero.
– Roy Halladay was a top pitcher with two starts tied behind his back.
That's right, we were missing two starts for Roy Halladay, and some starts by some lesser hurlers as well. We'd been investigating the reasons for Halladay being anywhere besides the top of the pitching lists, and finally figured it out. No changes have been made to the formulae, but the various leader boards for pitchers now give him his full due. We apologize for the error and thank you for your patience.
Thank you for reading
This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.
Subscribe now
See the batting report, filtered to just pitchers:
http://www.baseballprospectus.com/sortable/index.php?cid=1018548
Halladay: -0.5 BWARP, Lowe: +0.4 BWARP. BWARP here is non-pitching WARP, comprised of batting, fielding, and baserunning.
Halladay: -0.6 FRAA, -3.8 BVORP, -2.5 BRR ==> -0.5 BWARP
Lowe: +1.4 FRAA, +2.0 BVORP, -0.2 BRR ==> +0.4 BWARP
By contrast, baseball-reference.com appears to have 0's for fielding stats, batting runs above replacement computed by taking runs above average, positional adjustment, and then a replacement-level mod. I'll list the numbers in the same order as for the BP stats:
Halladay: n/a, -3 (-9+5+1), -0 ==> -0.3 WAR
Lowe: n/a, +3 (-2+4+1), -0 ==> +0.2 WAR
So, B-R has a difference of 0.5 WAR (+fielding), compared to
0.9 for BP. If you figure the 2 FRAA runs add 0.2 WARP, and take that out of the BP values, that becomes 0.7 for BP vs. 0.5 for B-R. Then, when you consider that each component at B-R is rounded to a whole number, who really knows how much difference there is? But it's not extreme.
My take has been that we may need to revisit the baserunning runs - those values seem to have too much magnitude, and my suspicion is they might not be getting regressed enough. This may or may not be the case, no promises. But I think this does indicate that *if* there is a bug in the coding, it's relatively small in scale, and the system - as designed - is the best available (though I think it took everyone multiple read-throughs of Colin's articles to adjust intuition on some things).
Pitcher A: 107 IP, 105 hits allowed, 146 total bases allowed (7 HR), 40 UIBB, 80 K
Pitcher B: 110.7 IP, 93 H, 127 TB (5 HR), 25 UIBB, 65 K
Pitcher C: 107 IP, 94 H, 140 TB (9 HR), 22 UIBB, 106 K
According to BP's pitcher's VORP, Lowe (pitcher A) is the best (34.1 PVORP). Jurrjens (pitcher B) is more than a win less valuable at 23.3 PVORP, despite what many would consdier to be better stats put up for the same team. Pitcher C is cliff Lee scaled to Derek Lowe's IP, and Lee's scaled line is good for 21.4 PVORP, despite allowing fewer hits, fewer total bases, fewer walks, and more strikeouts.
Hell, let's throw in Tim Hudson:
Hudson: 106.0 IP, 95 H, 133 TB (8 HR), 22 UIBB, 76 K.
Lowe: 34.1 PVORP. Hudson: 16.1 PVORP. In basically the same IP for the same team, Hudson has 10 fewer hits allowed, 13 fewer TB (though 1 more HR), 18 fewer UIBB, and 4 fewer Ks. 1 HR + 4Ks = 18 PVORP?
Which group of pitchers is better:
Group A: Roy Halladay, Felix Gernandez, Jered Weaver
Group B: Jon Niese, Derek Lowe, Tim Stauffer
According to PVORP, they're basically equivalent (100.2 PVORP fro Group A, 99.6 for Group B). I doubt there are very many people who would need to flip a coin to pick between which group they'd rather have. (baseball references pitching runs above replacement has Group A at 109, group B at 48. Fangraphs has Group A at 12.5 WAR, Group B at 5.5. Those numbers are much more in line with reality, IMHO)
Why not keep it going?
Roy Halladay is his own control group.
Roy Halladay makes everyone a standard deviation.
Roy Halladay was christened ROY.