by Clifford Blau


Bill James, in his 1985 Baseball Abstract, suggested that if one studied the rate of return for regulars at different levels of offensive production, one could determine the amount of hitting needed for players to keep their jobs. This required performance Mr. James calls the sustenance level, and I call here the offensive replacement level. Determining the offensive replacement level at each position is important for several reasons. First, it is valuable to use in estimating the value of a player to his team, which is simply how much better he is than whomever his team could replace him with without giving up another player. Since regulars who produce at less than the offensive replacement level lose their jobs, it is obvious that teams believe that there are players readily available who perform at least at that level. Thus, the replacement level can be used for evaluating players in general, rather than considering what specific person would take over for a given player if he were replaced. Second, it is useful for comparing the relative importance of fielding at each position, at least as perceived by managers. Theoretically, one could determine a player's defensive value by examining at what point he loses playing time due to weak hitting (see Eddie Brinkman for an example.)

The complete results of my study can be found in Chart #1. Both the replacement level and average production for each position have the expected relationship, in that first basemen and left fielders have the highest replacement levels and averages, and shortstops and second basemen the lowest. For those not familiar with runs created per game, Chart #2 lists some representative seasons showing conventional statistics along with the corresponding runs created per game.



Other Study



The first step was to determine the regular at each position (except pitcher) for each major league team from 1969 to 1989. I did this using the Baseball Encyclopedia, generally choosing the player listed there as the regular as long as he had made at least 243 outs (about 9 games (9 x 27) worth, a full season being 18 games worth (18 x 9 players = 162 games,)) although sometimes, especially with catchers, I might go as low as 189 outs (7 games.) If there was no regular, then that position was not included in the study for that team and year. Next, using the statistics in Total Baseball, I determined the runs created per 27 outs (RC/G) for each regular for the years 1969 to 1988. I then noted if the player was a regular the next season at any position (or DH), even if he played several positions and could not be considered a regular at one position, such as Dick Allen in 1970. If he was traded and played regularly for his new team, he was considered to have kept his job. If he did not play regularly the following season due to injury, but returned to regular status the year after that, I counted him as keeping his job. However, if he never played regularly again, his last season was not included in the study. After I had all these data, I set up ranges of one-half or one runs created per 27 outs for each position. Then it was a simple matter of counting up the number of players in each category and the number who returned the following season.


As you can see in Chart #1, the rate of return varies greatly from position to position. The rates did not generally vary much between leagues. I had supposed that the replacement level would be slightly higher in the National League, since the designated hitter allows American League teams to put a weak fielder at DH and put a good field/no hit player in his stead, but that wasn't the case. Bill James had theorized that there would be a range of one-half runs created per 27 outs where the replacement level would change sharply. However, that did not prove to be true. The rates tend to change slowly; in some cases, it was lower at a slightly higher level of production, e.g. at second base, players hitting between 3.5 and 3.9 RC/G returned 78% of the time while at 4.0-4.4 RC/G, the rate of return was 71%. Note that although I studied a total of 504 team/seasons, the sample size in each range is fairly small.

The replacement level for each position is about 1.5 to 2 runs below average production. I expected it to be one run below, since using the Pythagorean method of predicting winning average, at the average level of 4.2 RC/G, 1 run below gives a winning average of .367, 1.5 below yields an average of .292, and 2 runs below average gives a result of .215. In the past I have used one run above average for pitchers, which seems superficially valid. Thus, the level for hitters shouldn't be so low for an average fielder. It may be that it is artificially lowered by only good fielders being allowed to play regularly at low levels of offense (e.g. the young Ozzie Smith), so the offensive replacement level for average fielders is actually higher than is shown on the chart.

In considering the validity of this study, one must consider several factors. One is that a single season's statistics may not accurately represent a player's ability- indeed, several players who kept their jobs after a poor season were established players having an off year and their teams obviously believed that they could do better the next year. Another is that other factors than offense are more important than I have hypothesized. Further, the overall level of offense varied during the period. Also, a manager may overestimate a player's offense based on an incomplete understanding of statistics. Finally, a team may not have an adequate replacement available and/or doesn't want to take a chance on an unproven player. Any of these factors could result in a "true" offensive replacement level being higher or lower than I have calculated.


Another study of this question was done by Phil Birnbaum and published in By The Numbers, the newsletter of SABR's Statistical Analysis Committee in September 1994. Mr. Birnbaum used the entire 20th Century as the basis for his study. Some differences in methodology included counting a player as returning if he ever played regularly again and counting all outfield positions together. He also tested the results when a player's Total Baseball Fielding Runs were taken into account. Both times the results were generally consistent with my findings: no clear replacement level but rather a gradual drop in return rate as hitting decreased.


Based on these two studies, there is no evidence that a "sustenance level" as postulated by Bill James exists.  Apparently, too many other factors affect management's decision process, so that even light hitting players have a decent chance to retain their jobs.

Any comments or questions?  E-mail me at

Return to homepage