Stolen Bases and Caught Stealing by Catchers: Updating Total Player Rating
This article was published in the Spring 2014 Baseball Research Journal.
When I did The Hidden Game of Baseball with John Thorn, I came up with a total player rating involving batting, pitching, base running, and fielding. Some notes on total player rating:1
- Each player was rated on runs above or below average, with runs turned into wins, using a variable factor based on league data, which was usually around 10 runs per win. A rating of 3 wins above average was quite good, while some players could be as high as 5 or 6. Babe Ruth and Ted Williams could get close to 10 in their best years.
- These figures are comparable to WAR (wins above replacement). The only difference is I use a .500 team for the baseline, while WAR uses .294, which is 33 fewer wins. If you assume 18 rated full time players per team, that would be about 1.8 wins per player.
- Since I was rating players back to the beginning of baseball, I had to come up with something which used existing statistics.
- Batting and pitching were no problem, since the existing stats are sufficient to accurately calculate runs compared to the league average.
- For fielding I had to use putouts, assists, errors and double plays.
I thought the system worked quite well for infielders, and fairly well for outfielders. It wasn’t that great for first basemen. I did not count putouts for them, only assists, and some players pick up a lot of assists by throwing to the pitcher, while others run to the bag themselves. Still, the good ones tended to pick up a fair number of assists. But catchers were worse. If a catcher was effective against stolen bases, he would have fewer assists because players wouldn’t run on him. And handling of pitchers wouldn’t compute anyway. I tried to compensate for pitchers by giving the catcher credit for ten percent of the park-adjusted team runs allowed, which admittedly is inadequate.
Bill Deane undertook a private study in 1991 of Johnny Bench’s stolen bases and caught stealing allowed which encouraged me to try to incorporate that analysis into the catcher rating.2 At the time, Retrosheet had play by play of games back into the 1970s. For that period, I was able to use actual stolen bases and caught stealing from play by play accounts. For the earlier years, I went through the official averages and compiled stolen bases and caught stealing allowed by teams back to 1920. I used the team batting sheets and added those up by opponent. The only problem was the NL did not keep caught stealing from 1926 through 1950. Ernie Lanigan kept caught stealing records from game account for 1912 through 1919.
About half his data have been found in various newspapers. He generally just did base runners, but he did do caught stealing by catchers in some years. His data also allowed figuring the league totals.
I was able to estimate missing league caught stealing by combining three ways and taking a weighted average. We do have the league total of catcher assists. We can also estimate roughly the number of runners who were out on base. We know how many batters there were, how many outs there were, and how many runs there were. We can estimate the number of batters who reached base. The biggest problem is base on error. Luckily the NL actually kept base on error for a few years in the teens. I used 57 percent times errors to estimate times reached base. The other errors resulted in extra bases. You can calculate runners left on base as being total plate appearances minus runs and putouts.
So the runners out on base are equal to base runners minus runs and left on base. The major reason for outs on base are double plays. You also have pickoffs, runners out advancing and caught stealing. I also estimated caught stealing based on the number of stolen bases allowed.
I actually went back to 1890 for team stolen bases allowed. I also estimated total caught stealing, which was more of a stretch, since I didn't have actual data before the teens. Stolen bases had been compiled starting in 1886, but box scores before 1890 had too many missing stolen bases to be useful.
Once you have the team stolen bases and caught stealing allowed, you can estimate the breakdown among catchers. I prorated caught stealing by assists and stolen bases by putouts. Admittedly, the caught stealing data should be more accurate. I then rated each catcher using three factors, stolen bases/caught stealing allowed, putouts/assists/errors/double plays other than caught stealing, and the previously mentioned team runs allowed. So a catcher like Bench would have an excellent stolen base factor, although his total caught stealing would be low. This would result in improving his fielding rating because he would have more non-caught stealing assists.
Since then, Retrosheet has driven its play-by-play back to 1950, with a few games missing. Also Tom Ruane and his team there have a wonderful collection of box scores of every game back to 1914. This allowed calculating actual stolen bases allowed. For the 1914–49 period, if there was more than one catcher in a game, I prorated stolen bases by plate appearances and caught stealing by assists. I used fractions and rounded off at the end of the year. I also used the same method for those few games missing play by play in the 1950–73 period.
I was happy to see that my original estimates of caught stealing from before I had the box scores were pretty accurate. No catcher was off by more than three caught stealing in a year after 1927. The 1920s were a period of higher stolen bases. Luckily stolen base attempts were fairly low from 1926–50 when the NL was not keeping caught-stealing records. I also improved my inning estimate, which goes into the fielder rating. I found that using actual innings from the box scores gave me totals within about 10 innings from my earlier estimates for catchers and infielders in over 1,000 innings for the season.
Outfielders were a bit worse. They are more complicated because data are combined for three positions. The stolen bases figures were off a bit more from my original estimates. The substitute catchers were apt to allow more stolen bases per game than the regulars.
I am confident that the 1910-to-date figures for catchers are quite accurate. The 1890–1909 numbers are fairly good, since we have exact numbers for team stolen bases allowed and for catcher assists, but there is room for error in the estimations.
I rated every catcher in 500 or more games, almost 350 players. I compiled the number of stolen bases and caught stealing and also the number of expected caught stealing for the stolen bases allowed if the success rate had been the same as the league average. I rated catchers on caught stealing above the expected number. The league average was calculated each year for all opponent teams. There was a turf corrector in there because the actual success rate on turf is higher than on grass, since when you run on turf, the surface does not slide back as it does on dirt. Turf parks with dirt base paths were considered grass. Success rate is five percent higher on turf. Currently Toronto’s Rogers Centre is the only field using turf basepaths.
Table 1: Top 20 Catchers Who Played Mostly Since 1890
|Last name||First name||G||SB||CS||AVCS||DIFF|
Table 2: Bottom 4 Catchers Who Played Mostly Since 1890
|Last name||First name||G||SB||CS||AVCS||DIFF|
* Games played before 1890 not counted
Looking at the data, there is a clear winner for handling stolen base attempts and it is Ivan Rodriguez. I-Rod had a surplus of over 300 caught stealing. Hall of Famers Ray Schalk and Gabby Hartnett were over 200. Next came two nineteenth century catchers, Chief Zimmer and Duke Farrell, but since I used prorated team assists and stolen bases, the measurement error there could be high.
The only other two catchers who had a rating greater than half of I-Rod’s were Bob Boone and Jim Sundberg, both considered good fielders. Bench ended up 11th. Among actives, Yadier Molina is the best and could move into the top ten before he retires. Henry Blanco is also highly rated.
Bill Bergen is an interesting case. You had to figure he must have been a great catcher in order to survive for 11 years with a .170 batting average (.201 slugging) and an OPS that was only sixty percent of average. Bergen does show up with a decent rating, 39th all time, but you might have expected him to be higher.
At the opposite end of the spectrum is Deacon McGuire, who played on some very bad teams in the 1890s. Again, his rating is suspect because of data accuracy. Mike Piazza has the worst rating otherwise. Mike was never considered to be good at preventing stolen bases, and the fact that he was the best hitting catcher of all-time will deservedly put him in the Hall of Fame.
McGuire’s claim to fame was that he was the first person listed in The Sporting News Record Book. McGuire played for 26 years and the first record in the book was longevity. He got his 26th year by being one of the players hastily recruited back in 1912 in the famous Ty Cobb strike game. Cobb was suspended for punching a fan who had been heckling him, and the entire team refused to play. McGuire, a Tiger coach, was 48 years old at the time. He played nearly the whole game at catcher, watching 24 runs cross the plate, and even got a hit. In 1993 Nolan Ryan had his 27th season and McGuire was erased.
One major change over the years has been specialization. In the early years everyone was expected to steal, depending on the game situation. Now many players hardly ever steal, no matter what. From 1901 to 1910 only 13 percent of players in 100 games or more had fewer than 10 stolen bases, now it is a whopping 69 percent. The success rate today is a lot higher as a result. However, the catchers are rated compared to the league average, so the change in distribution doesn’t matter. The average success rate during Rodriguez’s career was 69 percent, compared to only 55 percent for Ray Schalk.
The pitcher is also an important part of defending the stealing game. For this study it was assumed that the overall mix of pitchers for any given catcher would be about the same as far as preventing stolen bases was concerned. This could be an additional factor that could change the ratings. However, it would be difficult to separate the pitcher and the catcher, since each could help or hurt the other.
PETE PALMER is the co-author with John Thorn of "The Hidden Game of Baseball" and co-editor with Gary Gillette of the "Barnes & Noble ESPN Baseball Encyclopedia" (five editions). Pete introduced on-base average as an official statistic for the American League in 1979 and invented on-base plus slugging, now universally used as a measure of batting strength. A member of SABR since 1973, Pete is also a contributor to "Who’s Who in Baseball," which will celebrate its 100th year in 2015. With John Thorn, Pete also edited seven editions of "Total Baseball."