THE BOOK--Playing The Percentages In Baseball

Are you suggesting that we use “Jim Rice” as a unit of measurement? If not, I am.

Think about it:

“Should Blyleven be in the Hall?”

“Well, once you get past the B.S., he has a lifetime 2.13 Jim Rice Rating.”

Or should it be “He’s over 2 Jim Rices?”

7 through 10 surprised me:
Blyleven 90.3
Gibson 85.5
Ryan 84.8
Carlton 84.4

33. Koufax 54.4
69% of that came in last 4 years.  If he had only pitched his last 4 years, he would be number 74 on the list.

92. Catfish Hunter 32.5
He had a 8.6 WAR in his famous $3.75 million/5 year contract w/ the Yanks.

It surprised me, but only a little.  I would have expected Carlton to rank higher than Bert, but they were pretty close.  Bert being overall more effective than Ryan is not a new idea for the sabermetrically inclined.  Gibson was great, but since WAR is a bulk measurement, he’s at a bit of a disadvantage not having as much career length as the other 3.

I was surprised by Catfish, but we are talking about a player with a career ERA+ of only 105, who was a pitch to contact guy in front of great defenses.

“2.13 Jim Rice Rating”

That’s better than an 8 Couric rating.

If you give Smoltz credit for full-leverage during his years as a closer, and account for post-season performance (which aids Smoltz and Schilling), Smoltz and Schilling end up noticeably ahead of Glavine and Brown. 

Rivera probably gets 1/3 of his value from the post-season.

The key point is that all 9 of today’s generation of starters are historically great.

How can Niekro and Perry be so close to Maddux?  Their IP lead is not that large, and Maddux’s ERA+ is dramatically better.  I know that WAR vs. ERA+ uses replacement vs. average and includes all runs vs. just ER, but it doesn’t seem that these two factors should be enough to explain Niekro and Perry’s strong showing compared to Maddux.

Does WAR include a pitcher’s offensive contributions as well?

I think I agree with #4, but Glavine was pretty good in the postseason also.  Fun fact: Glavine and Smoltz have identical 2.47 ERAs in 8 World Series games started each.  And Glavine had more IP and a significantly better WHIP (.91 to 1.28).

Pitching only (and pitcher defense indirectly since it’s part of run prevention).  The batter WAR pages are not integrated with the pitching pages, this could be a future enhancement.

Good question on Maddux vs Niekro and Perry.  I’ll compare to Perry.

Maddux, 5000 innings of 132 ERA+, about 563 runs above average.  Perry 5350 innings of 117 ERA+, 314 above average.  Edge of 250 for Maddux.  My system shows Maddux with 926 runs above replacement, and Perry with 834, a difference of 92. (They are essentially even in WAR because Perry pitched in a lower run environment, it took fewer extra runs to produce a win).  That means I still have 160 runs to account for.

Perry’s 350 extra innings are worth about 38 runs above replacement, so I’m down to 120.  Maddux’s teams saved him 64 runs over his career (mostly when he was a Brave) while Perry’s defensive support was -26 runs, at least by my figures.  At this point I’m down to 30 runs which I can’t explain without really digging into the data.

Two possibilities:  Perry might have faced slightly better opponents, relative to the league as Maddux did. (I’m evaluating pitchers with weighted opponent RPG instead of league average RPG) or more likely, it is because of the league adjustment - he pitched in the NL, which was inferior to the AL for most of Greg’s career.  During most of Perry’s career, the leagues were closer to even.

I agree 100% that postseason should count towards any player’s measure of greatness.  Some day, I will include it.

As for weight of postseason play, that’s a tough one.  Tango suggested twice the weight, but if you look at it as contributions toward a championship, it might be 50x more valuable.  At least for the World Series.

That is looking at only a World Series win having value, and a win everywhere else merely increasing your odds of getting to the WS and winning that championship.  If Pennants, division titles, and even regular season games have value in and of themselves, then your weights don’t have to be so extreme.  Tough call, and making it will be quite subjective.

In regards to Maddux vs. Neikro and Perry:

I am not familiar with how the team defensive support is calculated, but is it possible that Maddux is being hit too hard here? 

The glossary points out that a pitcher’s BIP rate is prorated, but does it assume the same GB/FB rate for each pitcher?  The strength of the Braves’ defenses were their outfields, and Maddux was an extreme groundballer.  If DEF assumes a league average GB/Fg rate, than Maddux gets unfairly docked. 

Also, Maddux had the unique situation of having a bunch of other true BABIP reducers on his team for a long time.  Glavine, Smoltz, Avery, Neagle, Wohlers, Stanton, and I am sure others all have BABIP lower than teamates to the tune of 1 standard deviation or more. I know it is imposible to seperate the responsibility perfectly on a team like this, but maybe all of these Braves pitchers are getting hurt unfairly by DEF - assuming too much responsibility on the defense and not enough on the pitchers.

I have seen other calculations that show Maddux losing 20-30 runs by his defense, but never close to 60.  Of course that was 20-30 runs prevented compared to average...but wouldn’t runs lost vs. average be the same as runs lost vs. replacement?

Likewise...did the 60’s - 70’s Braves really lose Niekro 100+ runs?

Any thoughts?  Trying to figure all of this out - a relatvie sabermetric newbie.

"As for weight of postseason play, that’s a tough one.  Tango suggested twice the weight, but if you look at it as contributions toward a championship, it might be 50x more valuable.  At least for the World Series.”

But it’s still a zero-sum game, right? It seems pretty harsh to dock some pretty good pitchers with a bad World Series appearance that much.

It doesn’t look at GB/FB mix, but the numbers don’t seem extreme to me, we’re only talking 3-5 runs per year over a 20 year career.

What about having a separate page or section for post season contributions? B-Ref has separate sections for playoffs, all-star, etc. That way you don’t have to worry about subjective weighting, people can weight it as they wish.

Or you can just use the approach linked to in my name, from Gassko.

Rally

I’ve been meaning to ask - are there known gaps of players?  I’ve been thinking about swithing over my draft database from WARP to your WAR, but have found some missing players.  The two examples I remember are Andy Dominique and Sandy Martinez. 

Dominique only has 14 career AB, but Martinez has over 600 over 8 seasons.

I doubt you’re getting many people searching for either, but for my purposes I want complete listings for a lot of pretty low level players.

Just curious if other people have mentioned some missing players and if there’s an easy explanation.

Thanks.

Rally has the players sorted by Retrosheet ID, not player name.  So, he is there:

http://www.baseballprojection.com/war/m/marta002.htm

If you have Firefox, it makes life easier.  You just type “Sandy” when you are on the M page, and boom…

Rally,

I was wondering if you might give us some names of players that had a decent amount of playing time that had career WAR of around 0 so we could actually put some names to those that meet the concept replacement player, preferably with all positions accounted for?  It is a concept many being introduced to idea struggle with, and actual names might help.  I noticed John Mabry was career -0.2, so he would be an example of a career replacement level corner if/of utility player.

SirK, great question!

Yes, how about the list of players, in descending order of PA and IP, with a career WAR of 0 +/- 0.5.

Sounds like a good idea.

I think my cutoff point was something like 50 career plate appearances to earn a card.

The defensive support question is very good.

The strongest version of DIPS, which I think no one (not even Voros) ever supported, is that all pitchers have the same true talent BABIP. Tom Tippett’s article in which he showed that very good pitchers tend to have below-average BABIP disproved this conclusively.

Similarly, if it turns out that the top pitchers all got above-average defensive support in Rally’s system, he is shortchanging the pitchers and over-crediting the fielders.

I have no idea if this is true. (I hope it isn’t!) I am going to look at the top 25 pitchers on the list and see.

Pitcher, Def runs per 1000 IP
R. Clemens, -1.4
T. Seaver, -1.3
P. Niekro, -20.0
G. Maddux, +12.8
G. Perry, -4.9
R. Johnson, +1.7
B. Blyleven, +0.4
B. Gibson, +4.1
N. Ryan, -6.7
S. Carlton, +0.8
F. Jenkins, -6.4
P. Martinez, +10.4
M. Mussina, -2.2
D. Sutton, +9.8
C. Schilling, +6.1
T. Glavine, +18.6
R. Reuschel, -19.2
D. Drysdale, -0.6
K. Brown, -4.6
J. Smoltz, +15.9
J. Marichal, -5.4
J. Palmer, +36.5
J. Bunning, +1.1
L. Tiant, +8.0
T. John, +4.2

11 negative, 14 positive
median: Carlton, +0.8
I’m reluctant to calculate the mean, since it may be skewed by Jim Palmer and the three ‘90s Braves pitchers.

Conclusion: looking good so far!

Even with that, the mean of that dataset is only +2.3.

The defense is calculated at the team level (just the sum of TZ, ofarm, dp, and catcher ratings from the hitter pages) and pro-rated by BIP allowed by the pitcher.

The intent is to provide an adjustment for defensive support, but still give credit to pitchers for exceptional hit prevention.  For this, I really don’t care how much of the low BA allowed is luck or not, just giving credit for what happened.

Thinking more about this in regards to the Braves pitchers.

Also, Maddux had the unique situation of having a bunch of other true BABIP reducers on his team for a long time.

In the THT article, Rally wrote:

I looked at how many balls in play a pitcher allowed, relative to his teammates, over his career, and regressed this amount by 50 percent when a pitcher has 3,550 balls in play allowed. Regress more if they have fewer balls in play, less if they have pitched more than that.

This is the best method I can think of, assuming the 3550 figure is right, as long as the teammates as a group have average BABIP talent. If they don’t, it’s a problem just like Matthew said.

Really this is the same problem as adjusting a team’s W-L record for strength of schedule. The pitchers Maddux was “competing against” (i.e. his teammates) were themselves above average. There are two ways to solve the strength of schedule problem: run an iteration procedure until it converges, or do it in one fell swoop with linear algebra.

Both of those methods should work in this case. By the way, there’s a brief explanation of the linear algebra method on page 17 of this PDF.

That would apply to Jim Palmer, but for the Braves during Maddux’s career I don’t worry about or adjust for pitcher BABIP (except for 2000-2002).  For those years I have GB/FB/LD/POP data, and for the 90’s, project scoresheet hit location.

For those years, TZ defense is based on the bb type and hit location.  If a pitcher allows a lower BABIP because he gives up fewer line drives, more popups, or both, then everything works out fine.

If a pitcher has real skill beyond that, say his ground balls or fly balls are easier or harder to field, TZ does not pick that up.

It is interesting that Koufax has the highest WAR peak (best 2 or 3 seasons), and in particular, higher than Pedro, Clemens or Maddux.  When you include the post-season for those seasons (’63, ‘65 and ‘66), it is, according to this measure, not particularly close.

Well, if TZ does not pick up on harder to field groundballs or flyballs - you may have your answer as to why Maddux is being shortchanged in comparison to Perry and Niekro.  Maddux is an extreme groundballer, but still has a great BABIP average.  This is a rarity.  Glavine was less of a groundballer than Maddux, but still a groundballer with a lower than expected BABIP (even vs. teamates).  Glavine’s LD% and IFF% are leage average (if not worse), but still produced a BABIP .017 better than league and .008 better than non-Maddux/Smoltz teamates.