O/U pitcher/team ratings

[nfleqbc]

This is understandably vague, but the explanation of the basic principles I use may be useful as a starting point.

Basic methodology: for every team/pitcher combination (with a minimum number of games), thresholds for exceptional games (on both the under and over) are set. The assumption is that the opposing pitcher/team combination accounts for these exceptional games, and therefore the tendency of a given pitcher/team combination to produce exceptional games can be quantified.

While not exactly the method I use (my internal ratings take into account the quantified tendency, not the qualitative tendencies listed here), these ratings may prove useful to at least one of you, under a possibly bizarre set of circumstances. One way to use this is to set a threshold for totals. Two pitchers listed as OVER tending going against each other would tend to go over the simple total (that obtained by averaging their scoring totals, for instance, obviously taking park factors into account).

So here are the qualitative ratings, with a simple total for each pitcher/team (my simple totals are not as simple as may appear... only games versus pitcher/team combos with enough recent appearances to make reasonable conclusions are considered) ... park factors are left as an exercise for the reader.

Top AL OVER Pitcher/Team combos [sorted by strength of tendency]
=====
Carlos Silva/MIN (14.1)
Cliff Lee/CLE (11.2)
Josh Beckett/BOS (11.9)
Paul Byrd/CLE (11.9)
Mark Redman/KC (11.6)
Ervin Santana/LAA (9.3)
Roy Halladay/TOR (10.1)
Joel Pineiro/SEA (10.9)

As an example... without compensating for park factors, Silva vs. Beckett should finish OVER 13.

Top AL UNDER Pitcher/Team combos
=====
Dan Haren/OAK (8.2)
Francisco Liriano/MIN (6.5)
Nate Robertson/DET (7.3)
Jeremy Sowers/CLE (7.7)
Scott Kazmir/TB (8.4)
Johan Santana/MIN (8.7)
Jose Contreras/CHW (9.6)
Casey Janssen/TOR (8.3)

Thus, for example, before taking a park factor into account, Liriano vs. Robertson should finish UNDER 6.9.

Top NL OVER P/T Combos
=====
Jon Lieber/PHI (10.9)
Ryan Madson/PHI (10.8)
Cole Hamels/PHI (11.9)
Derek Lowe/LAD (10.0)
Doug Davis/MIL (10.5)
Miguel Batista/ARZ (11.0)
Tim Hudson/ATL (10.8)
Steve Trachsel/NYM (11.3)
Chuck James/ATL (12.2)
Juan Cruz/ARZ (9.8)
Tony Armas/WSH (11.1)
Jason Marquis/STL (9.9)

Top NL UNDER P/T Combos
=====
Chris Carpenter/STL (8.3)
Chris Capuano/MIL (8.0)
Jason Schmidt/SF (7.9)
Aaron Cook/COL (7.4)
Ramon Ortiz/WSH (7.9)
Ian Snell/PIT (7.3)
Byung-Hyun Kim/COL (9.3)
John Maine/NYM (7.0)
Pedro Martinez/NYM (8.3)
Anibal Sanchez/FLA (7.8)

[dinque]

the fact that capuano is a leading under pitcher,while having an awful year,makes me believe that his team is just not scoring when he pitches.isn't that part kind of random?

[blogguy]

You might have Capuano confused with another guy.

He's 10-8 for a team that's 52-59
He's got an ERA of 3.87...with a K/BB ratio of 132/31

He's fallen back to earth a bit since the ASB...but I don't think you could call THAT an awful year.

But...I do agree with the general point that run support plays a big impact here. I'd prefer a methodology that didn't include run support. That's been proven to be random about 1,000 times over I think by stathead studies the past 15 years.

I think Halladay on the Over might be a better example of the dangers. You wouldn't normally think of him as a guy in Overs. The fact that he has been this year isn't necessarily an indicator that it will continue. Most of the guys listed do have team O/U records going in the direction NFLEQ would have suggested. Might be hard to separate out the randomness from the indicator....so you just have to live with some amount of finger-crossing in the percentages...

I'd also prefer seeing an algebraic formula lead to the "number" then just averaging the two together. If 12.0 guy is facing a 12.0 guy...those influences would typically magnify each other in a way that lifts the number above 12.

If an NFL team scores 24 on offense...and is facing a defense that allows 24...then the offense is more likely to score 28-34 than they are 24...