THE BOOK--Playing The Percentages In Baseball

An overall fairly reasonable list by the readers.  I’d have put Strasburg after Johnson and Verlander.

I would find a way to stick Adam Wainwright in here.  3.24 FIP over the past three season in ~450 innings.  His ERA over that span is 2.73.  He’s past his prime, but has improved every year he’s been in the rotation and is durable as hell.  Besides he’s a guy who is successful without having a 95+ MPH fastball like the rest of the guys on this list.  I would think fastball velocity is something that is most likely to decline due to injury, and that could really hurt a guy like Strasburg or Johnson or Verlander.

Plus Wainwright has a career .258 wOBA as a hitter in 275 plate appearances including 5 homers, and if you watch him hit it’s clear that he actually has some idea of what he’s doing.  If you regress that down to .230 and consider the average pitcher is around .160, that’s another 5 runs in 80 plate appearances over 200 inning season, which is worth another .225 points of ERA.

I’m probably a little biased towards Wainwright being a Cardinals fan, but all of the evidence points towards him likely being better than Felix or Verlander or Jimenez.

I think I selected everyone born after 1983, and Wainwright was born in 1981.  I didn’t want age to be anything more than a tiny consideration.

No Jon Lestah!? You ah cleah-ly mistaken. NO ONE DENIES THIS!

This is how CHONE see’s all starters with a nERA of 4 or lower and 27 or younger:

nERA A Team IP Player
3.02 26 SFN 137 Tim Lincecum
3.18 26 KCA 134 Zack Greinke
3.28 26 BOS 129 Jon Lester
3.38 26 COL 127 Ubaldo Jimenez
3.44 27 DET 136 Justin Verlander
3.51 22 LAN 106 Clayton Kershaw
3.52 24 SEA 139 Felix Hernandez
3.60 25 CHA 119 John Danks
3.60 26 FLO 96 Josh Johnson
3.62 24 NYA 87 Phil Hughes
3.80 27 LAA 127 Jered Weaver
3.82 22 OAK 100 Brett Anderson
3.82 25 SFN 132 Matt Cain
3.86 25 LAN 126 Chad Billingsley
3.90 25 TOR 112 Ricky Romero
3.92 24 MIL 102 Yovani Gallardo
3.92 27 CHA 129 Gavin Floyd
3.95 26 OAK 99 Dallas Braden
3.96 24 ATL 118 Jair Jurrjens
3.98 23 ATL 105 Tommy Hanson

Also, not to be a slave to small samples but in Lincecum’s last four starts he has a 20/19 BB/K ratio in 23 IP and has been more 90-93 with his fastball.

Here is how MGL sees those same pitchers, where the average MLB pitcher is 4.50, 4.40 in the AL and 4.60 in the NL (Sean, are you adjusting for league?):

Lincecum: 3.18
Greinke: 3.51
Lester: 3.48
Jimenez: 3.30
Verlander: 3.66
Kershaw: 3.95
Hernandez: 3.81
Danks: 4.21
Johnson: 3.63
Hughes: 3.96
Weaver: 4.18
Anderson: 3.64
Cain: 4.30
Billingsley: 4.29
Romero: 4.88
Gallardo: 4.43
Floyd: 4.40
Braden: 4.53
Jurrjens: 4.47
Hanson: 3.72

This is Oliver’s list thru games of June 2, same criteria, current MLB pitchers (no Strasburg until next week), IP is projected, ERA is for neutral park

 era age  ip org name
2.77  26 218 SFN Lincecum, Tim 
3.15  26 203 KCA Greinke, Zack 
3.16  26 213 COL Jimenez, Ubaldo 
3.29  26 161 FLO Johnson, Josh 
3.43  26 201 BOS Lester, Jon 
3.51  25 216 SFN Cain, Matt 
3.53  22 168 LAN Kershaw, Clayton 
3.59  25 173 BOS Buchholz, Clay 
3.61  24 152 ATL Jurrjens, Jair 
3.61  27 215 DET Verlander, Justin 
3.64  22 145 COL Chacin, Jhoulys 
3.67  26 203 PHI Hamels, Cole 
3.71  23 174 ATL Hanson, Tommy 
3.71  24 219 SEA Hernandez, Felix 
3.73  24 164 TBA Price, David 
3.73  22 154 CIN Leake, Mike
3.76  24 168 MIL Gallardo, Yovani 
3.80  26 159 DET Scherzer, Max 
3.80  25 190 CHA Danks, John 
3.81  26 203 TBA Garza, Matt 
3.82  27 186 FLO Nolasco, Ricky 
3.84  27 161 CHN Wells, Randy 
3.87  27 191 ANA Weaver, Jered 
3.95  26 189 LAN Billingsley, Chad 
3.96  27 155 SFN Sanchez, Jonathan 
3.99  27 172 TBA Niemann, Jeff 

MGL I do believe they are league adjusted.  Would have to ask Rally to be 100% sure (yeah, I know confusing since we both have the same first name).

But from these three lists it certainly does like Jon Lester is missing from the voted on list.

Brian, what does Oliver think of B. Anderson and Hughes?  (Although I believe Anderson may be out with a serious injury now, so obviously that kind of hurts.) Also curious on Romero since MGL and CHONE have almost a run differential there.

 era age  ip org name
3.70  24 103 NYA Hughes, Phil
3.77  24 139 MIL Gallardo, Yovani
3.84  27 194 CHA Floyd, Gavin
4.21  22 137 OAK Anderson, Brett
4.34  26 132 OAK Braden, Dallas
5.62  25 156 TOR Romero, Ricky

I don’t know why Floyd wasn’t in my first list

Putting Anderson and Braden on OAK lowers their ERAs to 3.84 and 3.90 respectively.

CHONE is seeing a different Romero than mgl or I.

 era age  ip org name
3.42  24 130 NYA Hughes, Phil
4.06  27 197 CHA Floyd, Gavin
4.05  22 122 OAK Anderson, Brett
4.13  26 154 OAK Braden, Dallas
5.06  25 175 TOR Romero, Ricky

Disregard the values in post #11, as I pulled the values out of the wrong (unregressed) table. These are from the same table queried in #9

Brian, what is the average ERA in a neutral park?  Those numbers don’t mean anything unless we know that, right?  Is that average (park neutral) ERA different in the AL and the NL?

Sean, I thought YOU were Rally?

"CHONE is seeing a different Romero than mgl or I.”

Chone must know something about Romero that we (Brian and I) don’t. Not that FIP is the end all, but I don’t see anything in his 2009 or minor league numbers that suggest a very good pitcher.

I keep track of Chone, Oliver, Steamer, and Zips projections.  I park and “defense” adjust the projected ERA’s by these forecasters to get a “neutral” projection for each pitcher.  Then I update those projections myself by adding in this year’s neutral performance.

Here is what those 4 projection systems have Romero as, assuming that 4.40 is the league average AL pitcher:

Chone: 4.60
Oliver: 4.91
Steamer: 4.50
Zips: 4.59
MGL: 4.94

So I am not really seeing where Sean is getting his numbers from for Romero. Obviously I don’t know how he is updating his projections, but pre-season, he had Romero with a 4.76 ERA in a league (AL) with an ERA average of 4.26 (all pitcher projections pro-rated to their current number of IP).

Toronto is a fairly neutral park, so basically he has Romero pre-season as .5 runs worse than the league average pitcher.  Currently, after 77 innings, he has Romero as .6 runs (I am assuming that the average nERA of the league average pitcher is 4.50, but I don’t know that - he might be scaling it to the current AL ERA which is 4.21) better than league average.  I am afraid I don’t buy that.  77 current IP, no matter how good, cannot reduce a projection by .8 to 1.1 runs, when prior to 2010 the guy had 178 IP in the majors in 09, 164 IP in the minors in 08 and 92 IP in the minors in 07.

So, Brian, his projection for Romero is NOT really much different from ours or anyone else’s, at least pre-season.  It is just that he is apparently doing something very different than us in updating the projections in-season, and, as I said, what he is doing I am not buying, not that it is for sale I don’t think…

And please, please, please, Brian, Sean, and everyone else, tell us what the league average pitcher’s numbers are before you give us ERA, nERA, or whatever you call your pitcher ratings. Without the baseline, your numbers mean nothing.  Even if you give us your projected ERA, we still need to know what your average pitcher’s ERA is (prorated to every pitcher’s 2010 IP), because the run scoring environment can change a lot from year to year (due to umpires, baseballs, weather, new parks, etc.) and if your average ERA is not equal (more or less) to the current league ERA, you are going to shortchange yourself in your projections.  A projection can only be measured compared to the average player in your projection list.  If you want to make the league average ERA 5.50 or 3.50, that is fine by me. Just let us know.

For example, I do neutral projections for every pitcher and then I throw in the park and defense adjustments and then I scale every pitcher to the prior year’s league ERA.  So for this year, my pre-season AL projections were scaled to 4.45, last year’s AL ERA. If I had perfect projections for every pitcher I would be off by around .23 runs in ERA for every pitcher because the average ERA this year in the AL is only 4.21.  Obviously some of that could be different playing times for pitchers than I thought (better pitchers getting more playing time and worse ones less time), but it is more likely due to other things like luck, weather, addition of a pitcher’s park (Target Field), large strike zone, different baseball, weaker hitters, etc.

I’ve got AL at 4.50, NL at 4.38 in avg ERA as calculated from wOBA allowed.

Park Neutral is also league neutral, generic MLB, which are the numbers I quoted here. The THT projections are adjusted for which MLB team the player is currently signed to.

For in-season projections, I simply vary the weighting.

At the start of 2010, 2007=0.6, 2008=0.8, 2009=1.0
At the end of 2010, 2007=0.0, 2008=0.6, 2009=0.8

where p=pct of 2010 schedule played
inseason = start*(1-p)+end*p

If 2010 is p=0.3 for 2010
2007 = 0.7*0.6 + 0.3*0.0 = 0.42
2008 = 0.7*0.8 + 0.3*0.6 = 0.74
2009 = 0.7*1.0 + 0.3*0.8 = 0.94

inseason = 2010+0.94*2009+0.74*2008+0.42*2007

Effective weight of 2010 is 0.30, that plus weights of previous three seasons will always sum to 2.4 as long as all seasons have stats

If 600 PA each season 2007-2009, 180 PA in 30% of 2010, sample size will be 1440, then divided by sum of weights (2.4) = 600 for weighted mean yearly value. If some years don’t have values, sum of weights will be less than 2.4.

MGL again, different Sean, CHONE creator generally only posts as Rally here or AROM at BTF.  I probably should use a different name as it’s pretty generic though.  Rally released updated CHONE projections on his site a couple days ago and I simply just pulled the data as I thought it would be interesting to add to the conversation.

Sean + last name
Sean + last initial
‘The other Sean’...but Foreman’s out there
or just create a pseudonym

it does cause confusion when you start discussing CHONE

I suggest changing your name to:
Not the CHONE Sean

I think Sean is OK.  I always post here as Rally.  The ERAn stat is adjusted for everything, what the system projects in a neutral environment.  An average pitcher should be whatever last year’s ERA was, or close to it.

Romero had an ERAn of 4.41 coming into the year.  Now he’s at 3.90.  So he’s improved by half a run, not .8 to 1.1, unless you’re comparing the preseason Toronto ERA to updated ERAn, which is not right.  Apples to apples, please.

Whether it’s right or not, we’ll see.  But the quickest way to improve your CHONE projection is to start striking out a lot more people.  Romero has done that with 79 in his first 77 innings.  That is the stat which is regressed to the mean the least, and where changes can signify real changes in ability the quickest.

Today, Jason Stark over on espn.com has an article entitled “Starting Pitching Stock Market”.  He surveys 22 GM’s and asks which starting pitchers they would buy stock in. 

Stipulations were…

All these pitchers already had to be in the big leagues to qualify (i.e. no Strasburg). And all had to be 26 or younger as of July 1.

http://sports.espn.go.com/mlb/columns/story?columnist=stark_jayson&page=rumblings100604&campaign=rss&source=MLBHeadlines

"Romero had an ERAn of 4.41 coming into the year.”

Is that scaled to the AL ERA from last year, which is around 4.45?  If yes, you are saying that pre-season you had him right around a (AL) league average pitcher, a tad less.

Yet....

According to your projection file that I downloaded from your site, his pre-season projected ERA was 4.76.  If I prorate all of your projected ERA’s in this same file to each pitcher’s current (2010) IP, I get an average ERA of 4.26. So you were essentially projecting Romero (before the season started) to have an ERA of .5 runs worse than the average 2010 pitcher.

“But the quickest way to improve your CHONE projection is to start striking out a lot more people.  Romero has done that with 79 in his first 77 innings.”

Yes, I know.  But it is still 77 innings as compared to over 430 IP in 07-09, albeit with less weight.  I’m still not buying it.

Let’s stick to the facts:

As I said, unless the file I downloaded from your site was wrong, you had Romero, before the season started, as exactly .5 runs in ERA (which is actually more than that in RA, but we’ll forget about that) worse than the average 2010 pitcher (starters and relievers combined).  Again, how did I get that?  Simple.  You had him projected with a 4.76 ERA.  TOR is a more or less neutral stadium.  If we take all of your pitcher projections and average them weighted by their 2010 IP, we get 4.26 in the AL.  You can try it if you don’t believe me.  Hopefully I did that correctly.  So you uncontrovertibly think (at least based on pre-2010 numbers) that Romero is exactly .5 runs worse in ERA than the 2010 league-average AL pitcher.

Now, if you think that he is only league average now, 50 games or so into the season, then his projection has gone down only .5 runs.  But if you think that he is .5 runs better than league average now, then his projection has gone down 1 run.  Again, these are all incontrovertible facts, unless I have the wrong file or I looked at the wrong column for his pre-season projected ERA (4.76) or all your AL pitchers do not average to 4.26 (weighted by their 2010 IP).  All of that can be easily verified.

Now, in order for you to have an updated projection of .5 runs better than league average, let’s see how Romero would have had to pitch for those 77 IP, using a reasonable weighting system for the last 4 years, 07-10:

Again, in 09, he had 178 innings (major league), in 08, 164 (minor league), and 92 in 07 (minor league).

We’ll use Brian’s 3/4/5, which is essentially what I use as well.  If we include 07 (and 10 of course), we have a weighting system of around 2.4/3/4/5 for 07-10.

So, in order for his current projection to be .5 runs better than league average, we have the following equation (everything is simplified because I am not including any regression, although the regression is obviously built in to all the projections by essentially adding some number of league average ERA for a left-handed starter, I assume):

(I am calling a league average ERA 4.50, so, for example, the “5.00” below is .5 runs worse than league average, Chone’s pre-season Romero projection.)

(2.4 * 92 * 5.00 + 3 * 164 * 5.00 + 4 * 178 * 5.00 + 5 * 77 * X) / (2.4*92+3*164+4*178+5*77) = 4.

We have to solve for X, which is his 2010 “ERA” in order to make his projected ERA .5 runs better than league average when it started at .5 runs worse than league average.

X comes out to almost zero.

Even if you have him currently projected at only .5 runs better than you did pre-season (for that to be the case, you would have to have him as a league average pitcher now), X would have to be around 2.5 or 2 runs better than league average.

So, I ain’t buying your updated projection!

You are confusing the issue with your estimation of a league average.  I don’t report a league average, I don’t set one, and I don’t update the rates to regress to when I do the midseason projections.  Whatever league average you compare him to for pre-season, that exact same league average MUST stand when you’re looking at updated projections.  If you think I’m saying he’s improving by a full run, then you are needlessly confusing the issue.

So using the Toronto ERA, he improves from 4.76 to 4.18.  0.58 ERA.  Using the ERAn, he improves from 4.41 to 3.90, or 0.51 runs.

4.18/4.76 = 88%
3.90/4.41 = 88%

If his projection improved by .55 runs, and his pre-season ERA projection was .5 runs greater than league average, which it was, then he is now right around league average.  If you say that he is NOT right around league average now - a little less, then something is wrong, period.  So, if you say that his current projected ERAn is 3.90, then the average pitcher in the AL must be 4.00 or so in ERAn.  Are they?

And, as I point out in my last post, he cannot be .58 or .51 runs better now in his projection than he was pre-season. K rate or no K rate, in order for that to be the case, you would have to be giving way more weight to current season performance than the typical forecaster does.

Preseason rates:
BB/PA .108 SO/AB .188 HR/Con .039 babip .313

Updated: .101, .217, .032, .311

I probably am giving more weight to current season than most forecasters in some categories, particularly strikeouts.  If I’m wrong I’m wrong.

If you’re not buying, you can have a full refund for every dollar you paid for these projections.

As for the comparison to league average, if you believe he’s half a run worse preseason, ERA pf 4.76 compared to 4.26, your calculation:

Then his current projection of 4.18 is just a bit less than that 4.26 league average.

Why do you find it so hard to compare his ERA of 4.76 preseason to 4.26 current or his ERAn of 4.41 preseason to 3.90 current?  Your post #26 sounds like your looking at his Toronto ERA for preseason, and the much lower, across the board, ERAn for current.  Why can’t you compare apples to apples and oranges to oranges?  Tango gets it in #25. Either way his CHONE ERA projection is 88% of what it was preseason, but if you continue to compare the wrong numbers then there is no point continuing this discussion.

Rally, of course I know which are apples and which are oranges.  If his current projected ERA is 4.18 in a 4.26 environment, then your projection indeed has gone down by .58 runs.  No problem there.

I was assuming that the 3.90 ERAn was based on a league average ERAn of 4.50, which is apparently not the case.  One of the problems is that you (and Brian) are throwing out numbers for these pitchers without telling us in advance or at the same time what the league averages are.  As I said, how can you do that?  YOU are creating the problem.

Here are my current projections for my top 3 pitchers in baseball:

Pitcher A: 2.87 (in ERAxfl)
Pitcher B: 2.99
Pitcher C: 3.06

You get my point:?  What the hell do those numbers mean?  They are worthless.  You do know who my best pitchers are, but you have no idea how much better they are than the average pitcher, which is pretty important.

Now, that being said, here is what you originally put for Romero:

3.90 25 TOR 112 Ricky Romero

So, 3.98 is the league average ERAn (if his projected ERA is .08 runs better than league average than any normalized ERA would have to be around the same, since TOR is a neutral park)? 

Romero is the 15th best young starter in baseball and his projected ERA is only .08 runs better than league average?

That’s what it looks like.  I double checked other Sean’s list and get the same thing.  Eliminate the non-starters and anyone over 28 and Romero is the 15th best.  Behind Cain, Billingsley, Anderson, and Weaver, ahead of Gallardo, Floyd, and Braden.

American league pitchers have a tougher time than most having to face the DH plus the quality of leagues.  So IF the AL average is 4.26, then 3.98 seems reasonable for a neutral ERA of an AL pitcher.  Neutral being that mythical league in an average park that is halfway in quality between the American and National league.

So, there are only 14 starters younger than 28 in baseball who are more than .08 runs better than league average?  I find that hard to believe.

Here are my list of starting pitchers who are age 27 or younger who I have projected better than .08 runs above league-average:

1) Brett Anderson
2) Billingsley
3) Clay Bucholtz
4) Cain
5) Danks
6) Floyd
7) Garza
8) Greinke
9) Hamels
10) Hanson
11) King Felix
12) U. Jimenez
13) Josh Johnson
14) Ian Kennedy
15) Kershaw
16) Lester
17) Lincecum
18) Matusz
19) McGowan
20) Price
21) Verlander
22) Volquez

Whatever.  As I’ve said, I’m not committed to any league average.  It’s a topic that makes me yawn.  On my site, that’s you can see how I rank them compared to each other.  Set league average wherever you feel like.

No big deal.  In general, I am a big fan of your projections! I just think there is something screwy going on here.  I’m sure you can find some screwy things going on with some of my projections as well, for whatever reasons.  And I am not saying that your projection for Romero is wrong.  I suppose we will have a slightly better idea when the season is over and we see how Romero performs from now until…

BTW, Rally, do you have current projections in downloadable format (a spreadsheet file) on your web site?

I don’t expect everyone to agree with my projections.  They’ve been wrong before.  Sometimes very wrong. But sometimes when mine disagree with everyone else they turn out right (Wieters last year, Colby Lewis so far this season).

As for a download, just load the web page, select all the rows, copy, and paste into a spreadsheet.  That’s what I do every time I’m at work, away from my main computer, and need to do sorts or calculations.

Got it thanks. For some reason I thought you only had the updated projections player by player (one web page at a time).

I’ve always said that an interesting way to compare various projection systems is to look at players for whom the different systems differ by a lot and see who is closer.  I am not sure if anyone has done that.  To take all the players for whom everyone has just about the projection and then see who has the lowest combined RMSE or something like that is pretty uninteresting.  Do we really care if forecaster A is off by an average of .15 runs in ERA and forecaster B is off by .22? I don’t think so.  We want to know who is better when forecaster A projects an ERA of 4.75 and forecaster B projects 3.75.  Of course, we’ll probably find that the players end up somewhere in the middle, I would suspect, which puts us right back where we started, which is that just about all the forecasters are the same, no matter how you slice it.

http://www.baseballprojection.com/2010/june2010p.htm

With about 2/3 of the season to play, that would be 3240 starts to go.  There are 149 starters totalling 2973 starts.  All starts is 9425.  So, be careful in using the “totals” to mean anything, as clearly Rally is using minor league data to extrapolate totals to the MLB level.  I’m pretty sure he noted something that the counting stats aren’t really relevant for non-MLBers, just the rate stats.

To that end, a column that shows something like “82%” meaning 82% of the forecast is from MLB performance and 18% from Minors would be helpful.

***

If we take everyone with at least 60 IP forecasted, and sort by nERA, and take everyone with an nERA of better than 4.80, we have 183 starters, totalling 3233 starts.  This becomes our pool of MLB talent.  I’ll point out that Carlos Silva just makes the cut with an nERA of 4.78.  The average nERA for these starters is 4.27. 

But, the average is not really relevant anyway.  All we really care about is how they compare to each other.

#10 in nERA is 3.51
#20 is 3.82
#30 is 3.92

So, just use those numbers as your scale.

Using the 4.27, Lincecum is #1 with an ERA of 71% of “pool average”.  So, it would seem to me that the pool average is pretty close to the league average.