A Quick and Dirty Win Shares
For a very quick-and-dirty way to estimate Win Shares: I take the average of Runs and RBI, and divide that by 5. Obviously it's not great and I prefer to use the proper numbers that you all worked so hard to calculate. But it's maybe almost as good as using pitcher wins.
Asked by: PeteRidges
Answered: 2/17/2021
Thanks. Rather than taking the average and dividing by 5, it seems easier to take the total and divide by 10, since 10 is the easiest number to divide by. Let's see... .Roger Maris, 1961, had 132 Runs Scored and 141 or 142 RBI, so that would be 274, divided by 10 would be 27. He actually had 36.
I would bet you would be closer if you divided by 9, rather than 10, but. . . .it's worth a look anyway.
I decided to look at the issue. Using a subset of players, about 8,500 players, all position players since 1920 who had at least 5 Win Shares each. Using that set of players, if you follow Mr. Ridges method, you get an average error of 4.59 Win Shares, almost all of which is accounted for by the numbers being too low. The players in the study averaged 70 Runs Scored and 68 RBI, thus 13.77 Q&DWS (Quick and Dirty Win Shares). Their actual Win Shares averaged 17.61. so the expected average is 3.9 less than the actual Win Shares, thus almost every player was too low, and the average error was 4.59.
If we divide by 9, rather than 10, then the Q&DWS average increases to 15.30, and the average error drops to 3.80. But the actual ratio of runs&RBI to Win Shares is not NINE to one; it is actually a little less than EIGHT to one. So if we divide the R&RBI by 8, rather than 9 or 10, then the Q&DWS average increases to 17.22, and the average error drops to 3.38.
Now, 3.38 is still a fairly large error, and if you want, you can stop listening now because we’re not going to be able to improve it very much more. But I had the thought at that point that the ratio of Runs and RBI to Win Shares had to be different for a catcher than for a Designated Hitter, so maybe we could use that to improve the Q&DWS estimates.
I figured the ratio of R&RBI to Win Shares for players at each position:
C- 7.01 to 1
1B- 8.30 to 1
2B- 7.43 to 1
3B- 7.78 to 1
SS- 7.47 to 1
LF- 8.04 to 1
CF- 7.66 to 1
RF- 7.98 to 1
DH- 9.99 to 1
So let’s modify the Q&D estimates in consideration of the defensive position. For shortstops, second basemen and catchers, we’ll divide by 7. For Designated Hitters, we’ll divide by 10. For all other players, we’ll divide by eight.
That improves the estimates a little bit, reducing the average error (which was 3.38) to 3.33.
And then I had another thought. It must be true that Gold Glove Fielders outperform the average, right? Has to be true.
So I checked that out, and that turned out to be true. There were a little more than 700 Gold Glove fielders in the study. They had an average Q&DWS of 21.53, but an average actual Win Shares of 23.68. The Gold Glove was worth two Win Shares, on average. This was perhaps the most interesting thing to come out of this study. . . the fact that a Gold Glove scans out at about 2 Win Shares.
So I did one more run. The Q&DWS were:
Runs Scored + RBI
Divided by the Position Ratio Number
+2 if the player won the Gold Glove
That’s still a pretty simple method, but it still has an average error of 3.32.
Of course there are many other things that we could do to further refine the estimates, but the point of a Quick and Dirty Estimate is that it is Quick. We’re stretching the meaning of "Quick" here; we can’t really stretch it any further. I would guess that, if you just use "8" rather than "10", the estimate is about the same as the rule that Wins=Win Shares for pitchers. Over and out.