I've thought about this kind of approach before but never tried to work out the math - this seems quite reasonable. Thanks Eric!

Applying this to Bob Caruthers, he has 56.8 WAR, but this should be multiplied by about 1.3 for seasons averaging around the 116 game range, which gives him 73.8, clearly in the range of a HOFer. Looks about right.

Many thanks to MWE for his help in getting this posted.

Applying this to Bob Caruthers, he has 56.8 WAR, but this should be multiplied by about 1.3 for seasons averaging around the 116 game range, which gives him 73.8, clearly in the range of a HOFer. Looks about right.

I'm torn on whether to use this method on 19th century pitchers, because their workloads were so much higher than those of their modern counterparts even though the schedules were shorter, and it seems like that may require an additional adjustment.

If anyone's interested, though, I can run a few more early position players and see what happens.

Caruthers' career was ended abnormally early by modern standards; presumably with modern pitching workloads he'd have gone on at least into his mid 30s. Swings and roundabouts.

Great work, Eric. I, for one, would love to see a few more position players run through this exercise. Maybe some of the HoM honorees. Or maybe someone that we haven't quite gotten to, like Ed Williamson.

Fantastic stuff, thanks for sharing this.

Let's see... the guy who probably benefits the most, even more than White, is Ross Barnes. B-R gives him 29.3 WAR; adjusting turns that into 62.1, with a peak of 12.3, 11.8, 11.2, 9.9.

Cap Anson doesn't really need the help, but he gets it anyway, going from 91.1 to 136.5. His counterparts, Brouthers and Connor, don't leap up as much because their debuts were later, but they both do nicely as well. Brouthers goes from 77 to 98, Connor from 81 to 102 (rounding to the nearest whole number for the sake of brevity).

Other HOMers who debuted before 1885 (let me know if I'm missing anyone):
Charlie Bennett 51 (up from 37)
Buck Ewing 59 (46)
Cal McVey 46 (22)
Joe Start 59 (32)
Bid McPhee 57 (48)
Hardy Richardson 52 (39)
Ezra Sutton 54 (32)
Jack Glasscock 77 (59)
Dickey Pearce 21 (10)
George Wright 51 (25)
Jim O'Rourke 77 (50)
Paul Hines 68 (43)
George Gore 53 (38)
King Kelly 59 (42)
Lip Pike 33 (15)
Charley Jones 41 (25)
Harry Stovey 54 (42)
Pete Browning 49 (38)
Sam Thompson 50 (42)
Monte Ward (non-pitching only) 44 (35)

Other guys from around the same time who I think either get votes sometimes or I'm at least slightly familiar with:
Jimmy Ryan 47 (41)
Ed Williamson 51 (34)
Fred Dunlap 50 (35)
Tip O'Neill 30 (26)

I won't swear by the WAR values themselves, of course. They seem to have gone through some enormous changes since I last entered them - I know B-R updated WAR since then, but I'm not sure which updates had the effect. My guess would be the runs-to-wins conversion.

What to do with Deacon White? Well, he played in an era before steroid testing, and was a more "durable" catcher than Craig Biggio.

Just sayin'.

This is a very nice piece of work. Thanks for sharing. Personally, I wouldn't do any kind of adjustments for pitchers. I tend to view most pitchers as having a fixed number of innings in their arms, so fewer innings per season tends to be offset by longer careers and vice-versa (I've made a similar argument with respect to the impact of WWII on Bob Feller's career).

This is a really, really interesting process and idea. I'll put my hat back in the ring and thank you for sharing it; I, personally, really like both the theory and the application very much.

Good article!

Fantastic work.

As you note, your adjustment methodology is based upon wins. Presumably, an analogous approach could be applied to other counting stats such as hits or home runs. To take a strike year as an example, suppose a player had 30 HR in a strike-shortened season of 81 games. Extrapolation would suggest he'd wind up with 60 HR in a full 162 game season. But as you point out above, extrapolation over-predicts high-performances (Reggie Jackson had 39 home runs at the 1969 All-Star break and wound up with only 47 at season's end). By using full season data, we could find all players who had 30 HR after 81 games and see how many homers they wound up with in the full season. Of course, we'd need to try to account for environment (era, park, etc.) in selecting the players to include as best we could.

Anyway, just a thought.

Excellent piece of work, and a fairly easy to understand explanation of what you did and why.

Thanks for all the support and feedback.

Personally, I wouldn't do any kind of adjustments for pitchers. I tend to view most pitchers as having a fixed number of innings in their arms, so fewer innings per season tends to be offset by longer careers and vice-versa

I expect this is true to a point. Still, an inning in 2012 is not necessarily the same as an inning in 1962, let alone an inning in 1912 or 1892 or 1872.

As you note, your adjustment methodology is based upon wins. Presumably, an analogous approach could be applied to other counting stats such as hits or home runs.

This is a really interesting idea, but it would take far more legwork to do something like this for individual players than it did for teams, even if you were only doing one counting stat. To do the entire batting line, you'd want someone who either has more time than I do, or has better data acquisition skills (I entered most of the data for this project manually).

Excellent piece of work, and a fairly easy to understand explanation of what you did and why.

This was especially gratifying to read, because I was at least as concerned about how well I'd communicated the information as I was about the specifics of the method itself.

You made it very clear - really a nice job.

I also like the method and the explanation. I'm adding my approaches to this problem not to argue with the author, but to offer him extra tools if he can use them.

1. On pitchers up through 1892, the last year of the 50 foot pitching box, I do this: Taking 40 games as a reasonable measure of how many games pitchers have started per season in all of MLB history, I then do this: As long as the season played doesn't include a whole 40 starts for the given pitcher, just include it as it is unless there's a leftover portion from the previous year. However, if the season involves more than 40 starts, then take the first 40 and call them a season. That leaves you with a remainder of games to take to the next year, where you will have to pick a part of that year, to made a combined "season", and have still another leftover portion, and so on. If you do that, the 1800s pitchers come out with much more modern-looking careers. They don't pitch 500 innings in a season, but they have more seasons. That is, they start to look much more modern, and can be compared to modern pitchers in that way.

2. Many years ago, I came up with this for 1800s catchers. You can plot a curve by taking, for each year of MLB play, the third-highest percentage of schedule played by any catcher. Using third-highest gets rid of outlier data, and there are always at least three catchers who have full, healthy seasons in any given year. If you do plot this out, you get a nice, graceful curve, starting on the right, in modern years, with data points close to 100%, but not making it quite that far. The curve serves you well, slowly dropping down as you go back in history, when medicine and equipment were primitive. That lasts until you go back to the 1870s, when the schedules get so small that catchers can play almost every league game, skipping only the occasional exhibition game, which suddenly wrenches your curve way up and unrealistic. You can fix this anomaly by simply continuing the historical curve with a French curve, setting a limit on how much of a 162-game schedule early catchers could catch, rather than paying any attention to the actual percentages in 29-game "schedules." When you want to look at a catcher in a given year, you attribute to him the higher of his actual games played and the games that the curve would give the catcher in a full, 162-game season.

So, when Deacon White plays all 22 games of his team's 22-game schedule in 1871 or 1872 or whenever, and your extended curve says 60% (which is about what it does say), then he gets credit for 60% of 162 games, or 97 games. More than 22, but a small enough number to fit in with the decline in playing time as you go back to weaker and weaker equipment. This gives you a reasonable base of playing time for these catchers, which you can then use with your favorite uberstat to estimate his value. You do still have the problem of a small sample size being inflated to a larger one, but you appear to have ways of dealing with that.

Fair Warning #1: People with credentials in formal math will tell you that extending the curve in this way is not allowed in formal math because the method lacks rigor, which it does. I, however, am trained as an applied mathematician, which means I think like an engineer. Engineers do this sort of stuff all the time; they are constantly running into data points that don't fit on any existing curve. That's why they also always want to build a model and test it - they know their math isn't rigorous. Well, sabermetrics is, IMO, a branch of applied math, not theoretical math. Rigor is impossible. Think like an engineer.

Fair Warning #2: If you do this, people will howl about Deacon White, because this method strips from him too many of his games played at catcher. He becomes a career third baseman. Deacon White fans think of him as a catcher. You will catch some heat. I know, I tried this in the Hall of Merit and got plenty of scorching. But it's still the best method I know of for dealing with very early catchers. - Brock Hanke

Brock--I'm currently reconstructing my HOM ballot, preparing for next year. I like the method you're using for catcher adjustments--can you send me the data you used, so that I don't have to go through every year on BR to get catcher games played data?