I initially created Baseball Graphs to develop new and better ways of presenting baseball statistics. The 2003 season was my first full-season attempt, and it turned out pretty well. I now spend a lot of my time creating and formatting the baseball stats at The Hardball Times.
Along the way, I've also picked up the writing bug. I spend most of my energy writing for THT, but sometimes I like to write about more geeky stuff, baseball research and things like that. That's what this blog is for.
I'm very interested in graphics and other information display, so I'll probably post links to creative stuff like that. I also plan to write historical articles that complement the historical graphs. And I have inexplicably developed an interest in baseball win-related stats, such as Bill James's Win Shares and something called Win Probability Added. So you can expect a few posts about that stuff too.
If you're interested, I've created a complex spreadsheet full of WPA logic, which you can download at ftp.baseballgraphs.com/wpa. If you'd like to read more about Win Shares, try these links:
Win Shares Intro
In our last article, we reviewed total salaries paid and Win Shares achieved by position by league. There are a lot of interesting stories implied in that chart, such as:
- The American League is the better “value” league, despite the presence of one George Steinbrenner among its owners.
- Second basemen are a good deal in either league.
- AL shortstops and NL first basemen aren’t.
- Pitching is overvalued by major league GMs.
Pitchers created about 36% of all major league Win Shares, but were paid 39% of the money. What’s up with that? Do GMs know something Win Shares doesn’t know? Or vice versa?
Pitching and fielding are the next frontier of my offseason Win Shares quixotic adventure. The analysis is not nearly as clean as the batting analysis, cause pitching and fielding are hard to allocate. But these salary tables are an interesting place to start.
Of course, Bill James designed his system to allocate more Win Shares to pitching and fielding in general. Over an entire league, 48% of Win Shares are allocated to batting, and 52% to pitching and fielding. So is pitching still undervalued?
I ran a little study of 2003 Win Shares, in which I looked at all position players with at least 300 at bats, and all pitchers with at least 75 innings pitched. These two baselines account for about 5% of any team’s total. The sample included about 270 position players and 215 pitchers.
In that sample, ten pitchers created no Win Shares. Yet there was not a single position player with zero Win Shares. In fact, there was not a single position player with just one Win Share.
Of all the pitchers, ten had zero Win Shares, seven had one WS, 16 had two WS and 15 had three. Among position players, one had two WS, one had three WS and three had four WS. The rest had five or more.
Now, four of the position players did have zero batting Win Shares, but received at least two fielding Win Shares. Fourteen position players created less than two batting Win Shares. However, many of them were fielding specialists (or catchers) and they racked up a lot of fielding Win Shares.
No player, other than the DH (and pitcher) received zero fielding Win Shares.
I’ve thought about playing around with the 52%, just for kicks. What would things look like at 54%, for instance? But I wonder: does Win Shares overvalue fielding?
Hmm. Maybe I should change the title of the article.
The batting/pitching/fielding split feels a bit arbitrary, so one could just as well use splits implied by $/WS as James’ intuition. I’m not aware of any evidence-based argument in favor of the 52% figure, for instance.
A WS question: is there a threshhold value for fielding WS, as there is for batting?
Yes. There are minimum and maximum percentage of defensive win shares going to fielding (25-40%) and absolute minimums and maximum fielding win shares per game. I’d like to see the latter axed and the former loosened, since more and more, I think we can have confidence in the various formulae.
As for overvaluing fielding, I’d say the opposite is true, at least historically. Tangotiger came up with a 40/30/20/10 split for DIPS as luck/pitching/fielding/park, and since most runs are still created from in-the-park hits, that would suggest that fielding is underrated in terms of wins. In terms of supply and demand, however, that’s a different story.
In terms of a pitcher/fielding split, I think I worked it out to a 70/30 split.
For off/def, it absolutely is 50/50.
However, I don’t think it follows that James is correct that you first break it by off/def, and then split the def into pitching/fielding.
What you have are various distributions: hitting, baserunning, pitching, fielding. Each has its own spread. The hitting has the most, and baserunning the least.
So, I wouldn’t be surprised that the split, if you were to use the standard deviation would be hitting/45, pitching/35, fielding/15, baserunning/5. In this case, it adds up to 50/50, but I’m not certain that it has to.
Win Shares undervalues pitching because if you come up with Win Shares above average for pitching, and compare it to a simple measure like FIP or ERA, you will see that the pitching will be a little low. (You have to remember that the runs per win is lower for a good pitcher like Pedro, Brown, Prior.)
As well, I think that an average pitcher ends up with less total Win Shares than an average hitter, as I mentioned in an earlier post.
(It went something like this: 36% of win shares goes to pitchers. So, an average pitcher with 252 out of 1440 innings will get 15.3 win shares. A non-pitcher NL, with 535 out of 680x8 PAs will get 15.3 win shares. 127 games for a 3B against 36 starts of 7 innings for a starter being equals? I don’t think so. Do you?)
So: pitchers start off with a lower base, and pitchers don’t get to add to their totals at the right rate.
You have to remember that the runs per win is lower for a good pitcher like Pedro, Brown, Prior.
Tango, can you explain what you mean by that?
I think win shares definitely overvalues defense and undervalues pitching. There are a few ways that you can see this. First of all, according to win shares, there is almost never any pitcher that contributes as many wins as the top ten or even fifteen best hitters in each league. Second, if you look at the Braves in the 1990s, you’ll see that the win shares for their defense is almost always in the top two or three - if I remember correctly, there was nothing special about the defense, it was just the pitching. The problem, I think, is in the valuation of the best pitchers; it does not follow that if defense is generally worth 30% of a team’s pitching+ defense win shares, it is also responsible for 30% of the win shares when each individual pitcher is pitching. Given a top pitcher, its probably worth only a much smaller amount.
Studes,
Here are the RS/RA and win% for various levels:
5 - 2 - 0.8116
5 - 3 - 0.7007
5 - 4 - 0.5951
5 - 6 - 0.4173
5 - 7 - 0.3473
5 - 8 - 0.2891
Let me also add “run differential” and “wins above .500”
5 2 0.8116 3 0.3116
5 3 0.7007 2 0.2007
5 4 0.5951 1 0.0951
5 6 0.4173 -1 -0.0827
5 7 0.3473 -2 -0.1527
5 8 0.2891 -3 -0.2109
Finally, let’s divide the run diferential by the win differential to come up with the runs per win converter:
5 2 0.8116 3 0.3116 9.6
5 3 0.7007 2 0.2007 10.0
5 4 0.5951 1 0.0951 10.5
5 6 0.4173 -1 -0.0827 12.1
5 7 0.3473 -2 -0.1527 13.1
5 8 0.2891 -3 -0.2109 14.2
As you can see, the runs per win converter is not at all stable. It’s directly proportional to the run environment. Each run that Pedro saves more directly leads to a win than a bad pitcher who gives up a run leads to a loss.
*****
This goes to the heart of one of the problems with Win Shares. Instead of setting the boundaries at 2.5 runs scored and 7.5 runs allowed for a 5.0 runs per game environment, James should set it at 3.0 and 8.0. At those levels, the win% will be .70 and .30.
But, if you do that, I don’t think things will add up nicely. The problem is that not all marginal runs have the same impact. A RS does not equal a RA.
Here are the boundaries that would ensure that you have a .300 team on offense or defense:
RS / AVG / RA
3.0 / 5 / 7.8
2.3 / 4 / 6.4
1.64 / 3 / 5
As you can see, the boundary levels should not be 50%/150%, or 52%/152%, but rather at the Fibonacci levels of 61%/161%.
If we took the graph for RS/game over time, could we derive a useful, non-linear function to be used in determining WS?
That is, if we know the change in a team’s expected win% by RA, could we hand out WS accordingly?
For instance, I’d expect the drop-off in WS to be severe between an adjusted ERA of 3.00 to 5.00, but not that much of a drop-off between 1.00 and 2.00.
A side benefit of such a function would be the ability to create a seperate function for particular inning-score situations, if anyone has or is willing to collect such data. For instance, a reliever with a 0.00 ERA probably should end up with more WS if he’s pitched a bunch of ninth innings in close games, but not as many WS if he only pitched the ninth when his team was behind by ten.
Good questions, David. If Tango comes back, I’ll let him answer your questions. If you haven’t seen his research, you should definitely drop by www.baseballprimer.com and look in the Primate Studies section. He’s developing a new system, called “Win Advancement” that promises to be awesome.
I plan to plug the Fibonacci levels into Pete’s spreadsheets in the next day or two, to see what happens. I’ll let you guys know.
This conversation has really heated up over at Baseball Primer. If you want to follow along, the link is http://www.baseballprimer.com/studies/archives/00000238.shtml#comments
Thanks to Tangotiger, by the way, for giving this some space on his discussion board.
| <<Previous Article: Cleaning Up Salaries | Next Article: The Win Shares Baseline>> |