Let's Go Tribe!: An SB Nation Community

Navigation: Jump to content areas:


Pro Quality. Fan Perspective.
Around SBN: Baby Boilers no more: Michigan State-Purdue Preview

Over at Batter's Box, they look at the quality of opponents faced by Cliff Lee and Roy Halladay & find that, unsurprisingly, Halladay has faced stiffer competition. I don't know that I expect the Cy Young voters to give things this much analysis (in fact, I can pretty much guarantee they won't give it this much analysis), but it's interesting nonetheless.

about 1 year ago Abe_simpson_tiny zempf 17 comments 0 recs  | 

Story-email Email Printer Print

Comments

Display:

One problem I saw when I read the write up is that even though there is a difference between teams in regard to performance level, there are also differences within a team’s performance level throughout the season, so while it is true that the White Sox have been a good team all season, they have been a better team earlier in the season and have regressed somewhat as the season progressed (as an example), Same goes for Oakland, and even Boston. So to simply throw out the games played by teams that both Lee and Halladay pitched would be ignoring the variance in performance at different points in the season for these team.

by hans on Sep 8, 2008 6:06 PM EDT reply actions   0 recs

Agreed, and I think it’s worse than that. First of all, it doesn’t appear that they’ve considered the potency of the lineups on a given day (a generic R/G rate is used—what about splits?) I also don’t see why I should care about the opposing starter (except for the win totals). Cliff doesn’t seem to either. In all of the quotes I’ve read he’ll insist that he’s (paraphrasing) “just out there doing his thing.” He’ll call out AJ Pierzynski, and when he says that he really doesn’t care what the Sox might think, he means it. Cliff is totally focused.

I wonder if it would be a little more refined if you took into account every batter Lee and Halladay have faced over the season—considering splits, ballparks and so on—to produce a total scoring expectation; compare that value against the actual runs allowed and voila. You’d run into the same problem you bring up—batters also go through hot and cold stretches—but the data set would be large enough that I’d expect these variations to iron out. I’m not convinced that it’s perfectly fair to eliminate lineups and teams and compare instead an amalgamation of isolated batters, but I think it’d be more convincing than this report.

by jhon on Sep 8, 2008 6:49 PM EDT up reply actions   0 recs

Yeah, it’s definitely a bit flawed. Ideally you’d have the stats for all the batters minus their at-bats against Lee or Halladay, too — you can say that Halladay’s faced harder competition, but maybe the reason Lee’s competition looks worse is because Lee’s pitched better against them. And as the article says, it’s not like this is something Lee can control — he gets to face the Royals 5 times a year & Halladay doesn’t, not his fault.

Despite all of my best intentions, I have not, in fact, grown up to be a debaser.

by zempf on Sep 8, 2008 6:54 PM EDT up reply actions   0 recs

Triple the Difference

(Yes, I have given this comment a title.)

But there’s simply no comparison between the overall Quality of Opposition the two pitchers have had to deal with.

This is a dramatic and ridiculous overstatement.

I understand why they want to talk about quality of opposing starters and overall teams, because of its impact on W-L record — but while I understand it, I’m solidly in the camp that isn’t very interested in that subject.

Let’s talk about the opponents’ runs per game, though — 4.58 for Lee, 4.81 for Halladay. We’ll leave aside the fact that a tiny piece of that difference is because Lee’s opponents have faced Lee, 2.41 RA, and Halladay’s have faced Halladay, 3.10. The effect truly is tiny, I just felt like mentioning it.

Anyway, runs per game is very close to runs per 9 innings — sometimes a club doesn’t bat in the 9th, but then again sometimes it plays extras, so we’ll just go with 9 innings per game. The difference between the two “opponent pools” is 0.23 runs per game, so over 230 IP …

0.23 × 230 IP / 9 innings-per-game = about 6 runs

Six runs? I’m missing the part where there’s “simply no comparison.”

Lee’s opponents have scored 4.58 runs per game, which means that over Lee’s 201.2 IP, you would expect them to score 102.6 runs. They have in fact scored 54 runs against Lee, 51 earned and 3 unearned.

Halladay’s opponents have scored 4.81 runs per game, which means that over Halladay’s 218 innings, you would expect them to score 116.5 runs. They have in fact scored 75 runs, 64 earned and 11 unearned.

So Lee has allowed 48.6 runs fewer than expected against his specific opposition, and Halladay has allowed 41.5 runs fewer than expected against his specific opposition. So I’m missing the part where this shows that Lee has accomplished less.

Halladay has been a great, great pitcher for a while, and he does some very impressive things that real aficionados love — like, pitching gobs of complete games. There is a natural inclination to downgrade what appears to be a fluke season for Lee and to elevate Halladay in comparison. I understand that, and I have those impulses myself. Halladay has been damned impressive for several years and deserves a Cy Young of his own, maybe more than any other pitcher.

Here’s the dirty little secret getting ignored here — maybe you noticed it above — the unearned runs. One of the first things advanced statistical students of the game figure out is that it makes more sense to look at total Runs Allowed rather than just Earned Runs Allowed — it correlates better year-to-year and with other defense-independent stats. The simple way to put it is that it’s more accurate to make the pitcher totally responsible for unearned runs than totally not responsible.

The ERA race is relatively close — 2.64 to 2.28 — and you can’t have Lee’s W-L record without generous run support, so again, there’s a natural suspicion that Halladay has really been better. But when you look at the RA race, it ain’t all that close — Lee leads 2.41 to Halladay’s 3.10 — a margin of 0.69, exactly triple the difference in the quality of opposition touted in this article.

If we’re going to look at a difference of 0.23 runs-per-game and say that there’s “simply no comparison,” then what do we say when there’s a difference of 0.69 runs per nine innings?

I’ll tell you what we say: There’s simply no question who deserves the Cy Young.

by Jay on Sep 8, 2008 9:46 PM EDT reply actions   2 recs

This is a very nice analysis. On the R/G issue, it would be nice, as zempf noted above, to remove the games in which Lee/Halladay have pitched in order to estimate a runs/game value that’s independent of the two pitchers’ performances. That way, if you want to adjust their ERAs based on the differential, you’re not adding two things that will be decently correlated.

My second minor comment is on total Run Average correlating better with FIP, etc. than ERA. Do you think most of that is due to situations such as those where an error occurs with 2 out, and then a bunch of runs score, all unearned? I suspect that if we redefined the concept “Earned” for this purpose to include all runs except those that directly resulted from an error (i.e. where the scoring baserunner reached on an error or where the actual scoring play wouldn’t have been a scoring play without an error). If RA does better than ERA, a more nuanced system is likely to do better still, since RA undoubtedly systematically overestimates the runs a pitcher is responsible for.

A more “theory neutral” way to balance RA and ERA would simply be to run a multiple regression, predicting your favorite defense-independent pitching statistic using both ERA and uERA (unearned run average — used instead of RA to reduce the correlation between the predictors), and then use the R^2 (variance accounted for) of the unearned run component to figure out a weighting coefficient for “what proportion” of unearned runs a pitcher can be expected to be responsible for in general.

Anyway, the result may not be hugely different from your method, but it’s less susceptible to arguments that Lee is getting too much credit for a superior defense playing behind him.

by Logodaedalus on Sep 9, 2008 6:36 PM EDT up reply actions   0 recs

It would also be nice to figure out a similar weighting coefficient to figure out what proportion of a run should be the responsibility of the starter when the starter leaves with men on base, and a reliever allows one or more to score. As it is, those are charged 100% to the pitcher who put the runners on in the first place, which doesn’t seem fair, though of course you can’t charge them 100% to the reliever either.

This might be a more difficult thing to calculate because it would differ more transparently by individual — I think you’d want to come up with a probability distribution for what would have happened in successive PAs had the starter stayed on, and use that to determine an overall distribution for how many of the runs would have scored. The expectation of the latter would be how many runs get charged to the starter. The prohibitively difficult part would be to determine what information you use in calculating probabilities of how PAs turn out, which, if we could do, would result in a lot more gambling in baseball.

by Logodaedalus on Sep 9, 2008 6:54 PM EDT up reply actions   0 recs

Look, all of that is and/or would be nice … but my only point is that RA is more relevant than ERA, if you’re going to pick only one to use and those are your only choices, and you won’t find any statistically inclined analyst who will tell you otherwise.

The reason is simple: Most unearned runs are mostly the fault of the pitcher. The mere fact that the pitcher can give up an infinite number of runs, none of them “earned,” after he gets three outs+errors in an inning by itself necessitates that this is true. As you seem to suspect, the great majority of unearned runs aren’t scored directly on the error, but subsequent to it.

Once the error has been committed, as an oversimplification, the best pitcher will get the “fourth out” immediately, the second-best will give up one run before getting the last out, the worst will find a way to give up five unearned runs before getting that one out. ERA treats these three cases as the same from the standpoint of what the pitcher has “earned,” but they are plainly not the same.

To put it another way, unearned runs are predominantly a function of how well a pitcher performs after an error has been committed, and only slightly a function of how many errors are committed. Conversely, as a measure of overall defensive quality, unearned runs provide very information — they don’t tell you anything about the range of the infielders and outfielders, or about their ability to create outs on the bases.

The pitcher’s job is to get three outs before any (or many) runs score, and it is a fact of the game that the pitcher sometimes has to induce four “out plays” in order to get three outs — just as it’s a fact of the game that sometimes the pitcher gets bailed out by his defense.

By the way, it’s pretty much a given that Halladay has had a better defense behind him than Lee has, and I don’t think you’ll see anyone making that argument.

by Jay on Sep 9, 2008 11:40 PM EDT up reply actions   0 recs

I absolutely agree that no one with influence is going to use anything so complicated as the things I suggested, and that simplicity often trumps accuracy. I’m just engaging in a bit of statistical masturbation, as it werel. Although I think it would be nice, and not complicated, to see a three-tiered run assignment system: “earned runs” (as they are now), “unearned runs” (those that result directly from errors), and “post error runs” (or something).

by Logodaedalus on Sep 10, 2008 12:35 PM EDT up reply actions   0 recs

What you really want is Opponents’ Runs Created.

by Jay on Sep 10, 2008 4:26 PM EDT up reply actions   0 recs

Joe Sheehan at BP offers basically the same quality of opposition piece between Lee and Halladay in his own “unfiltered” post today. Ranks the opposing teams by EQA. Pay content I think.

by cheech99 on Sep 8, 2008 11:05 PM EDT reply actions   0 recs

Pretty sure the Unfiltered blog is all free content … Sheehan’s entry is here.

by Jay on Sep 8, 2008 11:20 PM EDT up reply actions   0 recs

Blue Jays fans have as much invested in Halladay getting some hardware as Tribe fans have in Lee winning the CYA, and for essentially the same reason in this lost season. The bottom line is that Lee can’t control when he pitches or who he pitches against – if Wedge knew 2008 Cliff would be so unlike Spring Training Cliff or 2007 Cliff, he’d have a schedule a lot like Halladay’s. This particular “analysis” shows a bit of desperation in the stat-spinning. Against quality opposition, both of these guys have been pretty good, although Lee has given up 12 fewer ER in six more innings (each 8 total starts):

ANA Lee 1-0, 2.00 ERA; Halladay 1-0, 2.57
CHW Lee 1-0, 0.53; Halladay 1-0, 1.23
MIN Lee 2-1, 2.35; Halladay 1-0, 5.40
TBR Lee 2-0, 1.38; Halladay 2-3, 4.11
Total Lee 6-1, 1.60; Halladay 5-3, 3.70

Cliff’s still waiting for his first start against Boston this year, while Halladay is 2-1, 2.10 in three outings.

by FredOx on Sep 9, 2008 9:25 AM EDT reply actions   0 recs

The Angels, White Sox, and Rays are all quality opponents, yes, but not necessarily because of their offenses. The Angels and Rays rank 9th and 10th in the AL in runs scored, and the White Sox are 7th (Indians are 8th). It’s misleading to highlight these games as proof that Lee has pitched well against teams with good offenses.

Halladay has three starts against Boston (Lee has zero), but Lee has three starts against Detroit (Halladay has zero). They both have two starts against Texas – Halladay has pitched extremely well, Lee has pitched extremely poorly.

A big difference is interleague. Halladay had to face the Cubs (1st in NL scoring), Phillies (3rd), Pirates (9th, but a LOT better when they faced Halladay), and Reds (12th) in interleague. Lee got to face the Padres (16th), Dodgers (13th), Giants (15th), and Reds (12th). That’s a huge difference.

Look, there’s no shame in Lee having faced an easier schedule than Halladay. But comparing Lee and Halladay is not an apples-to-apples comparison. I’d still vote for Lee for Cy Young, though.

by Peter Bendix on Sep 9, 2008 3:59 PM EDT up reply actions   0 recs

The problem with all of this is once you start, when do you stop? Do we account for the fact that Halladay faced the Rangers in April, when they featured Ben Broussard at 1B and Frank Catalanotto at DH (.229 OPS entering the Toronto series)? How about Texas’ overall offense being significantly better when Lee faced them (average OPS .835 as of June 3 vs. .711 on April 11)? How about factoring in splits, since we have a lefty and a righty? Or accounting for any of the myriad other factors affecting performance?

by FredOx on Sep 9, 2008 4:58 PM EDT up reply actions   0 recs

Peter … see my post above.

The difference in scoring amounts to less than a six-run difference over 230 IP. It’s notable but hardly huge.

A lot of this stems from people overestimating the difference between facing good teams and bad teams. You know, 4.58 is more than 95% of 4.81. They’re two different numbers, but they’re not dramatically different numbers. If you had two pitchers with those two ERA’s, you wouldn’t think of them as being that much different.

by Jay on Sep 9, 2008 11:49 PM EDT up reply actions   0 recs

Halladay has better IF defense.

This thing is QED

by jhon on Sep 9, 2008 4:59 PM EDT reply actions   0 recs

Comments For This Post Are Closed


User Tools

Constantly updated Indians news, lots of in-depth analysis, live in-game discussions — and more fanatical and thoughtful Indians fans than every other web site combined.
Start posting about the Indians »

Join SB Nation and dive into communities focused on all your favorite teams.

FanPosts

Community blog posts and discussion.

Recent FanPosts

39135485-59af19dbb26654095f910f34176af094_4ae8a81e-scaled_small
Predictions Group
Small
I think we found our utility infielder
Small
Baseball in Japan
3444ant_black_small
Spring Training Trips
Hans_small
Trade Peralta?
3444ant_black_small
Beware the year of the Ox
Bos-fod_small
award-winning independent baseball documentary released
Calavera_small
Create your own 2010 BA Top Ten list (then sign Grady to an extension)
Img_0108_small
Jason Grilli Signs Minor League Deal?
Mariahcareyglitter_medium_medium_small
2009 AL Central Off-Season Transactions and Rumors

+ New FanPost All FanPosts >

Featured Poll

Poll
Who would you like to see hired to manage the Cleveland Indians?
Bobby Valentine
106 votes
Travis Fryman
41 votes
Manny Acta
113 votes
Don Mattingly
78 votes
Torey Lovullo
30 votes
Other
51 votes

419 votes | Poll has closed

FanShots

Quick hits of video, photos, quotes, chats, links and lists that you find around the web.

Recent FanShots

Lee has Foot Surgery
Jamey Wright Signed to Minor League Deal
2010 Cleveland Indians draft preview
Jim Ingraham Angers You.
BP Calls John Hart One of the, STRIKE THAT, BEST GMs of the 90s
Sizemore Addresses Offseason
Can we reach 1,000 on a post about Adam Kennedy?
"At this time, we're looking to make the biggest impact possible on the...
Casey Blake shaved his beard
ICBWDSTGFO550K

+ New FanShot All FanShots >

Sweet Baboo

Sweet Baboo, now with glitter


Managers

427px-nap_lajoie_1913_small Ryan

Dosequisman_small Jay

Authors

3444ant_black_small APV

47b8dd28b3127cceb64839d9746800000026102bauwjrq3za_small afh4