Baseball Toaster Bronx Banter
Position Battles: Fifth Starter
2007-02-12 20:11
by Cliff Corcoran

If the right-handed first-baseman battle is unique among the Yankees' three spring-training position battles because it's the only one that features a head-to-head competition, the battle for fifth starter is unique because it's the only one that has a clear favorite entering camp. Well, maybe "favorite" isn't the right term to use. This is Crash Pavano we're talking about, after all.

The top four spots in the Yankee rotation are set with Chien-Ming Wang, Mike Mussina, Andy Pettitte, and Kei Igaway. If Mr. Glass is able to pitch with any level of effectiveness during spring training and manages to make it to Opening Day without slamming his finger in a door or cutting himself shaving, he'll be the fifth starter. Not that the odds are very good of that happening, or that there's much chance of him finishing the season with that job. If Pavano pitches well through April and May it's almost a lock that he'll be traded in June. The only variable there is the health of the rest of the rotation.

So while the fifth starter job is Pavano's to lose, the competition in spring training will be no less compelling as a result. Here's a quick look at the combatants:

mL career2006
Philip Hughes6/24/86237 1/32.1310.212.050.23AA1162.2510.712.480.39.266
Jeff Karstens9/24/82522 2/33.677.432.290.72MLB42 2/33.803.372.321.27.238
Darrell Rasner1/13/815203.446.872.390.52MLB20 1/34.434.872.210.89.254
Humberto Sanchez5/28/834544.168.844.560.56AAA51 1/33.867.543.510.35.327

The above chart does not include Tyler Clippard or the two starters obtained in the Randy Johnson trade, Russ Ohlendorf and Steven Jackson, because none of those three have thrown a single pitch above double-A. Of course, the same is true of Philip Hughes, who is far and away the best pitcher listed above. But then there are limits on Hughes' participation here as well.

Soreness in his pitching shoulder limited Hughes to 86 1/3 innings in his first full professional season in 2005. Last year, the Yankees cautiously held the 20-year-old to 146 innings. This year his innings cap will be 180. If Hughes were to make the Yankees out of camp, the Yankees would have to keep Hughes on short leash in order to keep him from exceeding that total over the course of a full season. Last year Chien-Ming Wang and Randy Johnson lead the Yankees with 33 regular season starts. Spread across 33 games, 180 innings averages out to just 5.45 innings per game, and that doesn't include potential postseason starts, which would consume the innings the Yankees might be able to save by skipping Hughes' turn on the few occasions when scheduled off-days allow them to do so.

The Yankees are to be commended for their careful handling of Hughes. They're doing all the right things regarding his workloads, and their impulse to have him start this season at triple-A is also a good one. True, Hughes could likely pitch effectively in the major leagues without a stint in triple-A, but there's no harm in easing him into things. Starting him out in Scranton is an excellent way for them to insure that he won't be extended beyond those 180 innings. Another is to using him sparingly out of the major league pen in the early months of the season as the Twins did with Francisco Liriano last year.

Let's crunch some numbers here using 2006 AL innings-pitched and ERA leader Johan Santana and complete games leader C.C. Sabathia as examples of the best and most durable American League pitchers. Both Santana and Sabathia averaged about 6.88 innings pitched per start last year. Assuming , then, that Hughes will average no more than 6 2/3 innings per major league start, he could make 27 starts for the Yankees in 2007 before exceeding his limit. Of course, when you factor in a full postseason run, the Yankees could require close to 40 starts from their best pitchers, which leaves Hughes inactive for significant chunks of the season. There are three possible alternatives to simply letting Hughes pitch until he hits 180 innings and thus facing the heat of the pennant race and postseason without him.

If the Yankees let Hughes pitch on a strict innings limit in triple-A for the first couple of months of the season, then promote him in June they might only get 19 major league starts out of him between the regular and postseasons, but they'll keep him active throughout the season. If the Yankees skip the fifth spot in the rotation every time the schedule allows them to in April and May, they'll first need a fifth starter in June on the fifth of the month. To put Hughes on schedule for that start, the Yankees could make him the fifth man in the Scranton rotation. That would give him 11 starts in triple-A before his major league debut on June 5. If they then skip Hughes' turn in the major league rotation whenever the scheduled off-days allow them to do so over the remainder of the season, Hughes would make 16 major league starts during the 2007 regular season. If they limit him to five innings in each of his 11 triple-A starts but otherwise let him loose in the majors (again assuming a maximum 6 2/3 innings pitched per start), he would finish the regular season with a maximum of 161 2/3 innings pitched between Scranton and New York. That would then allow him to make three postseason starts (say one in the ALDS and two in the ALCS) without meaningfully exceeding his 180-inning limit. If all of that goes according to plan and the Yankees find themselves in the World Series, I'm sure an extra two starts in pursuit of a World Championship would be perfectly acceptable.

Alternately, the Yankees could limit Hughes to side sessions in April, promote him directly to the major league rotation when the fifth spot comes due on May 5, and again skip his turn whenever the schedule allows. That would result in 24 major league starts in the regular season and a maximum of 160 innings pitched prior to the postseason.

A third option, and the one that I think would ultimately be most beneficial to both the team's goals for 2007 and Hughes' long term development, would be to pitch Hughes in regular rotation in the major leagues beginning on June 5 (that is not skipping his turn when off-days allow). That would result in 21 regular season major league starts totalling a maximum of 140 innings. That leaves 20 extra innings which Hughes could pitch out of the pen in April and May (or, alternately, four minor league starts with a five-inning limit, or even five minor league starts with a four-inning limit). Again, that would result in 160 regular season innings and leave him free to make three postseason starts within his 180-innings limit with the possibility of two more World Series starts if needed beyond that limit.

Each of those three scenarios assumes a level of success for both Hughes and the team that is not a foregone conclusion, but even in these best-case scenarios it's clear that Hughes can't be used as the fifth starter in April, and likely not in May either lest he be shut down in advance of the playoffs and possibly even before the divisional race is decided. The Yankees will need a fifth starter three times in April and four more times in May. If Pavano gets a splinter, the Yankees will need someone other than Hughes to take those turns. Even if Pavano is able to take the ball each time, the Yankees will need to use spring training to decide who they'll turn to should the rotation spring a leak elsewhere.

Having reduced the candidates to Karstens, Rasner, and Sanchez, I'm going to further cull the field by eliminating Sanchez. Sanchez didn't hit triple-A until the middle of last year and made just nine starts at that level before being shut down with tenderness in his pitching elbow. A highly touted prospect in his own right, though certainly not in Hughes' class, Sanchez should also be handled with kid gloves this year in the hope that he can establish himself in triple-A, stay healthy, and emerge as a candidate for the 2008 rotation or, in a best-case scenario, down the stretch this year.

That leaves Karstens and Rasner, the two rotation patches that the Yankees used late last season. Although Karstens is the more familiar face due to his having made six starts last year to Rasner's three, I've always been partial to Rasner. The stats above show why. While both men where rather fortunate on balls in play last year, Karstens was both more fortunate and significantly less likely to record a strikeout, a scary combination that could lead to a huge increase in hits and thus runs allowed this year. What's more, Karstens has come by his nickname "Scary Fly Ball Guy" (props to Steven Goldman) honestly. Consider the home run rates above. Then consider that Karstens groundball-to-flyball rate last year was 0.67. While that came in a small sample, only one of the 80 major league pitchers who qualified for the ERA title last year had a more extreme fly-ball rate. By comparison, Rasner's groundball-to-flyball rate, in an admittedly even smaller sample, was 1.13, which would have ranked him a more respectable 51st out of those 80 qualifiers. Karsten edges Rasner in the important category of minor league K/BB ratio, but Rasner edged Karstens in nearly every peripheral stat you can dig up from their major league stints last year and, frankly, I was more impressed with Rasner's stuff and composure after seeing them both pitch last September. Still, just as Pavano would be keeping the fifth starter spot warm for Hughes, Rasner or Karstens will hopefully be doing no more than that. While Rasner may be a perfectly viable fifth starter, he'd be a significant downgrade from what's expected of Wang, Moose, Pettitte, or Igawa, and the rest of the cookies Cashman collected this winter still need to bake a bit longer.

2007-02-13 01:17:09
1.   Cliff Corcoran
Rasner still has options left, by the way, he was first added to the Natspos' 40-man after the 2004 season, so the Yankees are in full control of all of their young pitchers.
2007-02-13 02:49:05
2.   yankee23
Does this still count as leadoff? And what's all this "Crash" Pavano stuff? Porsche into the garbage truck aside, I say we stick with "Meat."
2007-02-13 04:30:11
3.   Levy2020
My understanding is that Igawa was more likely to lose his spot in the rotation since his contract was cheaper, he would solve the no lefties problem, he is even more unproved in the major leagues than Pavano, and the Yankees are hoping to increase Pavano's trade value in case they could get rid of him.

Also, thanks for not using Clemens in your calculation. There's certainly been enough fo that.

2007-02-13 05:13:23
4.   Sliced Bread
Thanks for another analytical feast, Cliff.

I'm with you re: Razzle Dazzle Rasner.

With Randy's back barking, I would have chosen him and Karstens over Lidle (RIP) and Wright last postseason, and probably Villone, as well.

You've given a lot of careful thought as to how best to proceed with Hughes. I agree his ML innings should not be wasted in April and May, while there are other servicable options. Still, I CAN'T WAIT TO SEE HIM!

I also agree it's best not to expect too much from Pavano at this point. For now, his progress and trade value should be measured one pitch at a time.

2007-02-13 05:40:12
5.   Jim Dean
Thanks again Cliff.

I with 3 in that I'm not sure Igawa is a legit 4 or isn't also a significant downgrade from Andy Wang Moose.

Back to choking on gasoline.

2007-02-13 06:05:00
6.   jayd
I think you're going to have to hold off writing about the starting rotation until the end of spring training. Simply stated, this starting rotation is to this year's real rotation what centerfield was before the arrival of Johnny Damon.

Kei Igawa is possible long relief maybe starter but no one to plan a season around. Pencil in as pleasant surprise we hope, could be gone by May. Ditto Pavano but wait to see how the big guy deals with the pressure. Does he really want to play? I say yes big time, but I seem to remember being wrong before. All right. I seem to remember being wrong three or four times before...Pavano comes up big and we have a season to remember.

So Wang, Moose and a fragile Pettite; make that a fragile Moose, too. Or do I mean a fragile Wang -- no guarantee he stays healthy through the whole season. Plenty of guys to throw in there for 2 or 3 missed starts so fragility is part of the game these days. But I really only count three starters so far....and I've written "fragile" too many times. At least I don't see Julian Tavarez' name anywhere.

Looking forward to seeing Hughes as well but not expecting as much as all the hype. Talk to me in 2008 and I will give you expectations.

You know what that means...A Shoe Drops In St. Paul... a new children's baseball book by....

Oh the suspense is killing me.

Johan my Johan...

2007-02-13 07:18:48
7.   rbj
Ah, in darkest, coldest February
comes three little words to make me merry.
Three little words I long to hear
three little words that fill me with cheer.
Three little words to bring hope to you
Pitchers and Catchers!

Woo hoo.
Good analysis, Cliff.
Re: Pavano, if he's pitching well (ha ha ha ha ha, sorry, it's just too funny), then keep him. You can never have enough pitching, and Moose will spend a stint on the DL, so will Andy, and possibly the WormKiller too.

2007-02-13 07:32:01
8.   Cliff Corcoran
7 I have to disagree about Pavano. If he's able to reclaim any value, the Yankees need to trade him and trade him fast. That said, you're right about never having enough pitching. I just don't think a healthy Pavano can accurately be described as "pitching" in much the same way that "magic" is more appropriately called "sleight of hand."
2007-02-13 07:48:24
9.   bobtaco
Anyone notice that Hughes and the Captain are born two days apart? I'm not sure if there is any significance there, but there might be...
2007-02-13 07:49:54
10.   ny2ca2dc
8 I think Cliff is right here, and Cashman has shown the inclination to trade away guys like Wright and RJ who, while decent/OK enough pitchers, are not getting it done, and are injury risks. It's a real risk getting rid of any servicable starter (assuming Pavano is servicable, and if he's not we're stuck with him anyway) when you don't have someone proven to replace him... But Cashman might do it.

Ideal for me would be Igawa to the pen to be the lefty set-up (or the villone role) after Hughes come up full time in slot 5, and an unnamed texan to take the #4 slot as pavano's bum is re-bruised by the doorknob on his way out (to SEA for sexon? ;) haha)

2007-02-13 07:53:08
11.   Yankee Fan in Chicago
I've come the conclusion that any innings of the alloted 180 that Hughes pitches in Scranton are a complete waste.

Start him out in the pen for at least April and May, maybe June too, then into the rotation.

2007-02-13 07:55:35
12.   Yankee Fan In Boston
another gem.

but... can we let up on comeback carl? a little? after all, he tossed the ball around yesterday and seemed to be alright.

you can see photographic evidence here (along with what i'm sure will be a slew of witty cutting remarks) here:

didn't we learn from last season that bombarding a player on your own team with negativity doesn't help? (and that was a player with exponentially more talent.)

i'm rooting for the guy. if he does well, it only helps the team.

can't we all just get along?

2007-02-13 07:59:51
13.   Cliff Corcoran
With regards to concerns about Igawa, here's his PECOTA projection from Baseball Prospectus:

29 GS, 180 IP, 4.42 ERA, 6.1 K/9, 2.8 BB/9, 12-8

Igawa projects to be almost exactly league average in every way, which is just fine for a fourth starter.

Last year the Yankees' fourth starter was Jaret Wright. Here's his 2006 line:

27 GS, 140 1/3 IP, 4.49 ERA, 5.4 K/9, 3.7 BB/9, 11-7

Igawa should be an improvement.

2007-02-13 08:26:37
14.   baileywalk
Those scenarios for Hughes are pretty convoluted. I think we can scratch one idea: it's doubtful they pitch him out of the 'pen. They are being as careful as they can with Hughes' arm, so I seriously doubt they do anything but have him start every five days -- they don't want him to warm up and sit down and put that wear and tear on his arm.

I agree we shouldn't be penciling in Igawa over anyone else. The reports on him is that he throws a nice change, but relies heavily on a 90-mile-an-hour fastball. When healthy, we know Pavano can get major-league hitters out consistently. We don't know that about Igawa.

I think if it's June and Pavano is pitching effectively, and Igawa is merely average, you'll see Igawa moved to the 'pen for the rest of the year and Hughes made the fifth starter. Hopefully Clemens won't enter the picture (I hope he goes back to Houston).

2007-02-13 08:26:54
15.   Rich
No matter which team he ends up pitching for, I predict that Pavano will put up better stats than Igawa.
2007-02-13 08:28:22
16.   Jim Dean
13 Sorry Cliff, but I'm very skeptical of projections for pitching even if that's the best we have to go on, and especially for Japanese imports. When a guy like Clemens (the greatest of his generation) can go from a 226 ERA+ and 176 ERA+ in TOR (1997 and 1998) to a 97 ERA+ in 1999 in NY and still never regain that TOR form until he got to Houston, I'm going to remain skeptical of pitching projections. Further Mussina killed his projections because he found a new changeup last Spring.

I'm not saying there's anything different about NY (though there might be), just that pitching is very unpredictable. Plus, that's before trying to project Japanese numbers into MLB.

As far as I'm concerned, Igawa and Dice-K are huge questions marks. And while the latter will have the whole season to prove himself, Igawa may be out of the rotation in May/June if Mr. Green Tea doesn't like the karma. Pavano by contrast will get chance after chance to prove himself for Torre (assuming health).

2007-02-13 08:35:30
17.   Rich
M'zaka is far less of a question mark than Igawa given how hard he throws.
2007-02-13 08:43:57
18.   Cliff Corcoran
16 Projections in general are sketchy because no one can predict the future no matter how seemingly foolproof their method, and pitching is indeed harder to predict than hitting, but your Clemens example doesn't speak to that because you don't mention what his projections were for those years. What if he was projected to decline considerably in 1999? (incidentally, PECOTA didn't exist then and BP didn't project pitchers until at least 2003).

I agree that Igawa and Matsuzaka are big question marks, but I expect both to do well. Japanese pitchers have a pretty solid track record, particularly in their first season in the majors.

What I don't agree with is that Pavano will get repeated chances from Torre. I think he's on the shortest of short leashes with the entire organization. The way I see it, the only reason he's being given any chance at all is so that he can develop enough value to allow Cashman to trade him.

2007-02-13 08:47:49
19.   Cliff Corcoran
Oh, and for yucks, here's Mussina's 2006 PECOTA:

28 GS, 175 1/3 IP, 4.08 ERA, 6.6 K/9, 2.3 BB/9, 11-9

And his actual 2006 line:

32 GS, 197 1/3 IP, 3.51 ERA, 7.8 K/9, 1.6 BB/9, 15-7

2007-02-13 08:59:15
20.   Yankee Fan In Boston
19 his changeup adjustment threw them a curveball. (or a knucklecurve, i guess.)
2007-02-13 09:02:04
21.   Jim Dean
18 On what basis would a projection system predict that significant of a decline for Clemens? Or the rebound in Houston? There just seems to be too many variables (and randomness) for projecting pitching with any precision - and it might always be that way.

Pitching might be like weather. You know good and bad when you see it. But no one knows what tomorrow or next month or next year will bring apart from the most general trends.

I'd probably feel much better about a projection system if they started showing their error rates and for particular players and years. Unless they do? Like Mussina last year - what was the projection? Or Josh Becket? To take two extreme examples.

True on the first year for Japanese pitchers. Still, the only thing that will convince me is their performance.

We'll have to see on Pavano. But I'm still not sure, given the "veteran" proclivities of the manager, how Meat doesn't get more time than Igawa to prove himself.

2007-02-13 09:10:12
22.   Jim Dean
19 But see, is that a good projection? It's not bad, but it's not great either.

Let's use that variance (as I have no idea what the actual variance of PECOTA is) and look again at Igawa:

Good Iggy:
33 GS, 200 IP, 3.92 ERA, 7.3 K/9, 2.1 BB/9, 16-6

Fat Toad Iggy:
25 GS, 160 IP, 4.92 ERA, 4.9 K/9, 3.5 BB/9, 8-10

Of course, if we consider the difference between predicted and actual 2006 Moose as somehow representative of the projection system, then Igawa is just as likely to be very good as very bad. So really what have we learned? What's the point of paying their subscription fee if that's the best they can tell us?

2007-02-13 09:10:34
23.   rbj
19 What happens to older pitchers the year after they exceed their PECOTA? Do they come back down to earth or do they maintain a higher level of performance?
2007-02-13 09:10:39
24.   Cliff Corcoran
21 The switch to the NL alone might result in a projected improvement in Houston. As for the decline in New York, age and the unsustainability of the level at which he was pitching in Toronto might lead to a projected decline. I'm not saying the projections would have done that, but if you want a reason why they might have, there you go.

Mussina I posted above. Here's Beckett:

PECOTA 2006:

29 GS, 184 1/3 IP, 3.87 ERA, 7.3 K/9, 2.7 BB/9, 12-8

Actual 2006:

33 GS, 204 2/3 IP, 5.01 ERA, 6.95 K/9, 3.25 BB/9, 16-11

Finally, Torre also has major trust issues, especially when it comes to pitchers, and he's been give every reason not to trust Pavano.

2007-02-13 09:20:27
25.   Jim Dean
24 Fair enough, thanks.
2007-02-13 09:27:12
26.   Yankee Fan in Chicago
14 We're not talking about making Hughes a closer or some other situation where he warms/sits/warms/pitches. He'd warm up, same way he'd do as a starter, and come in for an inning or two of work, starting innings rather than coming into jams.

I believe this was how the O's broke in their great rotation of the 70s, Palmer et al.

Again, why waste some of his admittedly limited innings in the minors. He's already dominated AA. Great pitchers traditionally, I've read, don't spend much time in AAA. For Hughes it ought to be nada.

2007-02-13 09:32:02
27.   Jim Dean
24 One more note on Beckett: The difference for ERA and W-L is in the HR/9. It went from .71 to 1.58. And there was no way of knowing or projecting that.

He's still a big question mark for them. Funny when I was terrified of what he'd do over there.

Okay, I'm done.

2007-02-13 09:32:29
28.   C2Coke
Baseball season is starting! And with Cliff back in business, I am excited all around. It feels as if all my nerves are itching.

"cutting himself shaving"...why can I already see this image in my head? Mr. Glass's got a long long way to go.

2007-02-13 09:38:40
29.   C2Coke
13 In addition to that, Igawa is less injury prone according to his health history.
2007-02-13 09:52:21
30.   JL25and3
26 That was part of Earl Weaver's dogma. Weaver's Eighth Law: The best place for a rookie pitcher is in long relief. But because of his Seventh Law - It's easier to find four good starters than five - there were also more opportunities for spot starts along the way.

That's how Weaver broke in guys like Flanagan, McGregor, and Dennis Martinez - but not Palmer. He was one of the "Baby Birds" of an earlier, pre-Weaver era, a crop of very young pitchers who were thrown into the rotation and promptly blew out their arms. Palmer and Dave McNally recovered; Wally bunker and Steve Barber never really did.

2007-02-13 11:00:02
31.   YankeeInMichigan
30 Absolutely correct. The long relief role is far more significant in a 4-man rotation. In a 5-man rotation, the long relief guy can end up sitting for weeks.

1 Thanks. I had seen elsewhere that Rasner was out of options, but I'll go with your word. I've also seen that Sean Henn is out of options. If so, he has an inside track on a bullpen spot.

2007-02-13 11:14:56
32.   Cliff Corcoran
31 I'll have to check on Henn, but if he is out of options he looks more like Alex Graman redux than a potential member of the Opening Day roster to me.

27 Actually the move from the pitching-friendly Dolphins Stadium in the NL to the homer-happy Fenway in the DH league suggested an increase in home run rate, so I wouldn't say there was no way that could have been predicted. PECOTA clearly thought highly of Beckett and thus limited that homer rate increase to an '06 projection of 0.9 HR/9. A more pessimistic projection could easiliy have pushed that past 1.0, though I will admit that his actual 1.58 HR/9 was pretty surprising.

2007-02-13 11:54:15
33.   pistolpete
This made me laugh out loud:

"If Pavano gets a splinter, the Yankees will need someone other than Hughes to take those turns."

2007-02-13 12:11:50
34.   Cliff Corcoran
22 Jim, somehow I missed that post. BP's PECOTA cards on their site actually do give you a range of projections. What I've been quoting has been the weighted mean projection, but they give good (90th percentile projection) to bad (10th percentile).
2007-02-13 13:14:03
35.   OldYanksFan
10 I think that BECAUSE we unloaded both RJ and Wright that Cashman will nned to keep Pavano if he is league average or better.

I also think we are already having expectations of Hughes. Hopefully we won't NEED to bring him up.

We are short on SP right now. Lots of questions. I don't think we can afford to get rid of ANY league average pitchers, ie: possibly Meat unless:

...One of the kids is consistantly better then Meat and can hold up for the PS
...We get a mid-season guy, Clemens or whoever.

Since 'you can never have too much pitching' and pitching is the reason we keep losing in the PS, I say Meat stays with us.

2007-02-13 13:15:59
36.   OldYanksFan
32 I thought Fenway was Homer neutral, with a little boost to BA and doubles. You got some stats?
2007-02-13 13:29:14
37.   Jim Dean
34 Even then of the few cards I've seen it seems they act like they have a much more precise estimate (with the distribution and all) and then the weighted mean projection skews that perception even more IMHO.

Rather I'd like to see the weighted mean with some standard error term (assuming chance variation) around each factor. For instance, it's alot different to give the Igawa weighted mean versus the very real range of 22 (if in fact that's his variance by chance alone). Those are two entirely different pitchers and he could wind up as either one based on nothing more than random variation from what we know right now (or correct me please if PECOTA gives those terms).

It's in that sense that I just can't buy too heavily into projections for pitchers.

32 And that's the thing about about Beckett's HR/9. A .9 or 1.0 would have been probably within two standard deviations from his mean - not likely but not that surprising either. The actual number was so far beyond any possible projection, it just makes me wonder how valuable they can be overall for pitchers. Even Moose too. Where did last year fall for him? his 80th or 90th percentile?

2007-02-13 13:49:44
38.   Jim Dean
P.s It just reminds me of something Neyer said in a chat - that the standings can be pretty well predicted within 10 games.

My response: That's a prediction!?

2007-02-13 13:59:19
39.   Cliff Corcoran
This just in: Humans Remain Unable to Predict the Future, A Nation Weeps.
2007-02-13 14:39:55
40.   Jim Dean
39 Come on, be fair if you're going to be critical.

What's the point of pretending 13 that the predictions are accurate if they're not. All along I've drawn the line at pitchers.

You present Igawa's weighted mean in 13 as evidence that he "should be better" than Wright. You're saying he's a league average pitcher. I'm saying he might be - but there's a huge variance on his 2007 outcome.

If you're going to go that far, then you've already drunk the Kool-Aid (with your subscription fee) far more than I'm comfortable with.

All I want is an error estimate and depending on how good it is I might think I can learn something from BP about pitchers. The evidence for Moose is less than compelling (where did it fall in their 2006 distribution for him?) because that same variance results in two vastly different Igawas and would for most pitchers.

The Beckett projection was much better, but there's no dancing around the big miss in the HR rate. And so any projection for him this year is just a wild guess as to whether he rebounds from that even if everything else stays the same. Do I really need PECOTA for that bit of logic?

2007-02-13 15:31:25
41.   MattinglyHOF
Id say handling Hughes with kid gloves is a good idea up to a point.We still have to play a season and win games and if Phil wins the 5 out of ST maybe 200 innings isnt out of the question.Just get him out early if hes not effective and after 5 if hes winning by a large margin and pace him out.
2007-02-13 17:11:56
42.   Cliff Corcoran
40 I said "sould be" not "will be."

The range of predictions for 2006 are no longer available, the weighted means for both Mussina and Becket I got from Baseball Prospectus 2006. Odd that you think the Beckett prediction is better, though.

Look, all I was trying to say when I brought up the Igawa PECOTA was that I think he'll be a perfectly fine fourth starter and look, here's an advanced analysis system--the most accurate there is in fact--that supports my point of view. I don't believe that the numbers in 13 are what Igawa will actually do in 2007, but I think they are a fair baseline. Certainly he could be better or worse. The point is there's room for him to be worse and still match Wright's production in the fourth spot last year and if he's better, well hot damn! Either way I think that going into the season expecting him to be anything less than the fourth starter is self-defeating. The guy was a strikeout champ and winner of the Japanese equivalent of the Cy Young in Japan and he's here to be a productive part of the rotation. Until he proves he can't do that, I see no reason not to assume he will.

2007-02-13 18:33:35
43.   Jim Dean
42 Okay. My only point is that there's variation around those numbers, and in order to feel good about PECOTA as "an advanced analysis system--the most accurate there is in fact" I have to know how much variation they take into account. I'm not demanding that of you (except when you post the weighted mean as what to expect) - more of the BPers.

In scientific reports, when someone shows a graph it's almost always necessary to show the variability, confidence interval, range, standard deviation, etc to fill out the numbers and to give a sense of how they can move around.

The weighted mean doesn't do that.

Me, I'd like to see:

Igawa 2007

29 (+/- 5) GS, 180 (+/- 20) IP, 4.42 (+/- .50) ERA, 6.1 (+/- 1.2) K/9, 2.8 (+/- .7)BB/9, 12-8 (+/- 4 W 2 L).

It doesn't look pretty but it conveys so much more information. And for Igawa it should be no surprise that his variability is much higher than for a Moose or Beckett because of the Japan-US translation of rates.

Indeed, the Becket and Mussina examples are illustrative.

The error in the Moose projection varied by

4 GS, 22 IP, 0.57 ERA, 1.2 K/9, 0.7 BB/9, 4 W and 2 L

By contrast, the error in the Beckett projection was:

4 GS, 20 IP, 1.24 ERA, 0.35 K/9, 0.55 BB/9, 4W 3L

So it's starting to look like the error for pitchers with a few MLB seasons:

GS is +/- 4, +/- 20 IP, ERA and W-L we know are a function of other things (BABIP, HR/9, BB/9, K/9).

And the rest looks like the particulars of the individual pitchers. Moose K'ed more guys but was also hunting for them (increased BB rate). I'll bet he also did better than his projected HR/9 (actual = 1.00)

I think they did better on Beckett because the error on BB/9 and especially K/9 was so much less. They just really missed on the HR/9 and that's ultimately what killed him (and their projection). If they had gotten that right then the rest (ERA and W-L) would have followed very naturally, I think. With Moose they missed wide on everything.

See, it's not that I think PECOTA is bad. I just they they could do so much better with typical error rates and leave it there. Percentiles become pretty meaningless precisely because of the variability. Instead they should stick to the mean and how it's likely to vary. But the two should be inseparable, esp. for pitchers.

Sorry for getting all geeked up.

2007-02-13 19:21:51
44.   yankz
You can't estimate what the error was based on just two pitchers, no matter how similar it appears.
2007-02-13 22:07:08
45.   rilkefan
43 - it's not really normal practice to quote (nearly) two sigma errors on an estimation. That's more of a worst-or-best-case scenario range and not that useful for comparing players. (I would also guess that the width on Igawa's estimate is rather larger than usual, since it depends to some extent on a smaller-sample measurement and much worse systematic uncertainty.)
2007-02-13 22:08:15
46.   rilkefan
45 - d'oh, shoulda read the whole comment before going off.

Comment status: comments have been closed. Baseball Toaster is now out of business.