THE BOOK--Playing The Percentages In Baseball

The latest PECOTA has him at 4.52 ERA, 1.9 WARP for 2010.

PECOTA also lists Lewis with a very high UPSIDE score of 226.8, slightly above Jake Peavy and Roy Halladay, and comparables of John Denny, Mike Scott, Curt Schilling, and Jason Schmidt.  (Which doesn’t seem right.)

OOPS, a new PECOTA came out 2 days later, and Lewis changed into a 4.14 ERA, 2.7 WARP, 290.4 UPSIDE pitcher.  (The UPSIDE is right between Roy Halladay and Matt Cain.)

+-------------------+------+
| name              | war  |
+-------------------+------+
| Darvish, Yu       | 12.9 |
| Lincecum, Tim     | 12.4 |
| Haren, Dan        | 10.8 |
| Sabathia, CC      | 10.7 |
| Halladay, Roy     | 10.5 |
| Vazquez, Javier   |  9.5 |
| Greinke, Zack     |  9.1 |
| Hernandez, Felix  |  8.8 |
| Lilly, Ted        |  8.7 |
| Wainwright, Adam  |  8.5 |
| Lee, Cliff        |  8.2 |
| Lewis, Colby      |  8.2 |
| Beckett, Josh     |  8.1 |
| Santana, Johan    |  8.0 |
| Greisinger, Seth  |  7.9 |
| Jurrjens, Jair    |  7.8 |
| Cain, Matt        |  7.7 |
| Hamels, Cole      |  7.7 |
| Jimenez, Ubaldo   |  7.7 |
+-------------------+------+

Currently my NPB pitching translations are based on 224 pitchers who played on both sides of the Pacific.

Coming to the US, the pitchers are working in generally larger parks but against better hitters, so on average base hits and homeruns are a wash. I have a BH factor of 1.03, HR 1.01, BB 0.87, SO 1.07,

Lewis pitched for Hiroshima which was the most homer friendly park - factors of 2.10 for batters and 1.27 for pitchers topped both categories. I project Lewis at only 13 HR, with 152 H, 29 BB and 152 SO in 165 IP, for a 3.07 ERA. Pecota’s latest is 21 HR, 162 H, 40 BB and 148 SO in 164 IP, for a 4.14 ERA. So Pecota has +8 HR, +10 H, +11 BB, -4 SO

Here’s my unregressed MLEs for the past three years

Year Org   ERA   IP    H  K  BB HR K/9 BB/9 HR/9  WAR
2007 OAK  3.57 131.2 121 102 39 18 7.0  2.7  1.2  2.7
2008 HIR  2.81 174.1 155 172 32 10 8.9  1.7  0.5  5.1
2009 HIR  2.80 177.6 158 173 23 11 8.8  1.2  0.6  5.2

Lewis had a couple good years in the minors in the early 00’s, but hurt his arm and was out almost two seasons. 2006 was a comeback year with middling results. in 2007 in the PCL he was lights out, but rocked in a handful of starts for the A’s. Off to Japan for two years where he duplicated his 2007 success at Sacramento.

"Lewis pitched for Hiroshima which was the most homer friendly park - factors of 2.10 for batters and 1.27 for pitchers topped both categories.”

Brian, what is the difference between a pitcher and batter park factor? I never heard of that distinction before.  When you see PPF and BPF in Palmer’s stuff and on B-R they are simply referring to the park factor adjusted for the fact that a team doesn’t face its own batters or pitchers.  So it is merely an opponent adjustment and the BPF and PPF are only applied to players who play for that team.

MGL, that’s a combination of the park and league factor. A hitter coming from Japan to MLB will on average have his homers reduced by a factor of 1.71, while pitchers will have their HRs reduced by only 1.01.

Coming to MLB, the batters face a double whammy of better pitching and larger parks. The pitchers face better hitters, but that is offset by the larger parks. Batters—Pitchers +-

I don’t understand the Seth Greisinger forecast.

| Lee, Cliff | 8.2 |
| Lewis, Colby | 8.2 |

You can’t stand behind that forecast, can you?  That forecast says two things:
1. the mean of Cliff Lee = mean of Colby Lewis
2. the UPSIDE of Colby Lewis is much higher than Cliff Lee (as well as the downside)

How do I know the upside is higher?  Because the uncertainty around the mean is far higher for Lewis than it is for Lee.  Given that you have the same mean, that means Lewis has a better chance of winning the Cy Young than Cliff Lee.

***

There’s got to be a better regression built-in for players playing against totally new competition.  You can’t just make the translation as you are doing.  Not to mention that the translations you have of your 224 pitchers certainly doesn’t have anyone that forecasts like Cliff Lee does it?  So, in addition, you are extrapolating beyond your data points.

I just can’t believe that a pitcher, even if he was a first rounder, that pitched for 217 IP in MLB with a 6.71 ERA will come back to MLB at the age of 30/31 and put up Cliff Lee numbers, and with a better upside.

Ok, I guess it’s up to me to come up with my own forecast, but how can this pass the sniff test?

If a GM came to you and said “Should I spend 25MM$ over the next TWO years on Colby Lewis”, what are you going to say?  That it’s a bargain?

That’s one set of b-lls you’d have to do that.

Tango, you are right that they have the same means but would have different uncertainties.

I’ve generated some error bar numbers based on sample size, but I agree that where the player put up the numbers also contributes to the uncertainty.

I can run some variance tests to compare the spread of errors based on league before MLB.

but one of the tings I am doing towards that end, and which ties into a discussion a couple days ago, is regressing each player to the MLE of an average player in his league.

Say a guy plays 90% in AA and then 10% in MLB. I will regress him to a mean that is 90% what an average AA player projects to and 10% an average MLB player.

Guys from Japan get regressed to a lower mean than those in MLB.

Tango, I don’t have my projections done yet, and I don’t do any Japanese league translations.

I have my doubts about the Colby Lewis projection.  It sure doesn’t seem right, but he was one of the two best pitchers in Japan the last two years (along with Darvish), and that’s what the numbers say.  The same system has done well for other Japanese pitchers coming over.  Almost identical in Kuroda’s case.  It doesn’t feel right, but I’m not going to monkey with the system just because I’m not sure about one guy.

“Not to mention that the translations you have of your 224 pitchers certainly doesn’t have anyone that forecasts like Cliff Lee does it?”

Actually, it does.  Dice-K.  His last two FIPs in Japan were 2.71 and 2.55.  Lewis has been 2.58 & 2.62.  Dice-K didn’t quite live up to the forecasts (3.28 CHONE, 3.82 ERA first two years before getting hurt).  If the Rangers get two years of 3.8 ERA in 180+ innings this will be the deal of the offseason.

Speaking of Cliff Lee, he was horrible in 2007.  Had his contract been sold to Japan and he made the same adjustments that he has in the majors, his record probably would look a lot like what Colby Lewis did in Japan.  Lee has an outstanding 4.5 K/W the last two years against MLB hitters, Lewis is 8.0 vs. the inferior competition he’s facing.  If that had happened and Lee were returning to the US for 2010 there would be a lot of doubts about his projection.

BTW, my projections in 2007 preseason when Lee was coming off a terrible year and Lewis off another dominating AAA season combined with terrible MLB results, I had these ERAs projected:

Lee 4.45
Lewis 4.99

I meant 2008 preseason.  Or 2007 offseason. d’oh

looks like these colby lewis forecasts were dead on thus far…