The Effect of Artificial Surface
David W. Smith
June 17, 1995
The effect of natural grass vs artificial turf on batting performance has been debated since the ersatz material was first installed in the Astrodome in 1966. Many different kinds of arguments have been advanced, supported by various types of data.
The present analysis is of all the 22806 games played from 1984 through 1994, covering over 1.5 million at bats. The play by play data used here came from the Baseball Workshop in Philadelphia, from which anyone can order them.
The most common argument put forth is that batting averages are inflated by artificial turf because the ball moves so much faster and therefore ground balls get through the infield more easily.
An examination of this simplest assertion appears in Table 1.
Table 1.
Batting average by league and playing surface, 1984-1994.
League Surface Games At bats Hits BA
AL Grass 8664 590960 154545 .262
Turf 3466 237665 62888 .265
NL Grass 5474 372243 95367 .256
Turf 5202 352519 89421 .254
Both Grass 14138 963203 249912 .259
Turf 8668 590184 152309 .258
Several interesting conclusions may be drawn from this table. First of all the two leagues play very different numbers of games on the two surfaces. In the AL there are 2.50 games on grass for every one on the carpet; in the NL the figure is 1.05. The AL has an overall higher batting average of .263 to .255. (Ed. note: This, of course, is due to the effects of the DH.) However the differential as a function of surface is quite small, with the AL having a three point increase on turf and the NL a perhaps unexpected two point drop on the artificial surface. The league differences cancel out, leaving a net of .259 in all games on any surface.
The information in table 1 is really much too superficial, however, since different parks differ in many respects other than the playing surface. For example, all domed stadia have artificial surfaces and domes have been suggested as significant factors in offensive statistics. Some other factors that affect batting average besides playing surface are: symmetry of the playing field, extent of foul ground, height of fences, prevailing wind conditions, average temperature, and altitude.
The analysis was therefore refined in an attempt to minimize these other effects (the operative word is clearly "minimize"; since complete elimination of all effects besides playing surface is a very elusive objective).
The refinement done here is to modify and expand the batting average calculation. Batting average may be defined as the number of successes per opportunity, where the successes are safe hits and the opportunities are at bats. These two basic parameters were adjusted in four ways:
1. Subtract home runs from hits and at bats. Balls hit over the fence presumably are affected very little by the nature of the playing surface.
2. Subtract strikeouts from at bats. Since strikeouts are plays in which the ball is not contacted, the effect of the playing surface can be safely ignored.
3. Add sacrifice flies to at bats. Sacrifice flies are currently not included in at bats (for many years they were), although they are essentially regular fly balls that are not hits.
4. Subtract non-sacrifice bunts from hits and at bats.
Although there is very likely an effect of the playing surface on the chance of a successful bunt, the primary analysis is concerned with balls put into play on full swings. Sacrifice bunts are not at bats, but the data used does identify bunts, so bunt hits and bunt outs that are not sacrifices are removed.
Net Effect: Adjusted At Bats = At Bats + Sacrifice Flies - Home Runs - Strikeouts - Bunt Hits - Bunt Outs
These adjustments can be summarized simply: all balls put into play on full swings are considered and only balls put into play on full swings are considered. In this context "in play" means that a fielder could conceivably make a play, therefore the exclusion of home runs. The small number of inside the park home runs (10-15 per year) is disregarded. In addition each type of hit: single, double, or triple, is considered separately, since there are good reasons to believe that they will be affected differently by the playing surface.
The analysis will be presented in three phases: 1) all games for the 11 years in aggregate; 2) all games for individual seasons; 3) all games for individual teams for each season. Moving to smaller subsets of the data, such as the performance of individual batters, tremendously increases the statistical variability and is more likely to confuse the analysis than to enhance it.
All Games for 11 Seasons
By making the adjustments described above and separating each hit type, we obtain the information presented in Table 2. For both leagues individually and combined the patterns are unmistakable: the rate of singles goes down on turf about 3.9% while the rates of doubles and triples go sharply up, 13% and 36%, respectively.
The combination of these values into an adjusted batting average (labeled "ABA" in the table) shows a small but consistent increase in both leagues while playing on artificial surface.
Table 2.
Singles, doubles and triples and adjusted batting average (ABA)
per adjusted at bat by league and playing surface, 1984-1994.
Adjusted batting average is the sum of singles, doubles, and
triples divided by adjusted at bats.
League Surface Games Singles Doubles Triples ABA
AL Grass 8664 .228 .055 .00652 .290
Turf 3466 .221 .064 .00933 .294
NL Grass 5474 .230 .053 .00718 .290
Turf 5202 .219 .060 .00909 .288
Both Grass 14138 .229 .054 .00677 .290
Turf 8668 .220 .061 .00919 .290
All games for individual seasons
The next approach is to examine each league for each individual season, giving us 11 comparisons for each league. Rather than report a massive list of actual rates, table 3 contains a list of the number of times that each category increased in games on artificial turf out of the 11 seasons for that league and how many times that category went down in games on turf. For example, in the 11 AL seasons, the actual batting average was higher on turf 8 times. In the NL the increase occurred in only 3 of the 11 seasons. The individual hit categories are completely consistent in the surface effect they show: each league in each season had a higher adjusted rate of singles on grass and higher adjusted rates of doubles and triples on turf. Combining these three into the adjusted batting average gives proportions similar to the actual batting average changes.
Table 3. Proportion of increases on artificial turf in rates of actual batting average, rates of singles, doubles, triples and adjusted batting average for all seasons, 1984-1994. League BA Singles Doubles Triples ABA AL 8/11 0/11 11/11 11/11 9/11 NL 4/11 0/11 11/11 11/11 4/11
Individual team-seasons
The last stage of the expansion is to consider the number of times each individual team had its rates increase on artificial surface in a given season. For the AL there were 154 team- seasons from 1984-1994 and for the NL there were 136. These results are presented in table 4, which is parallel to table 3 in its organization.
Table 4. Proportion of increases on artificial turf in rates of actual batting average, rates of singles, doubles, triples and adjusted batting average for individual teams for all seasons, 1984-1994. League BA Singles Doubles Triples ABA AL (all teams) 82/154 50/154 115/154 109/154 87/154 NL (all teams) 53/136 30/136 104/136 94/136 52/136
Although the results shown in table 4 are consistent with the earlier information, we see that increased variability is introduced when the data sets get smaller. One conclusion to be drawn here is that the response of a given team in a given season may be contrary to the overall trend without seriously compromising that larger picture.
Overall conclusions
1. There is no consistent change in the rate of hits per adjusted at bat due to artificial surface. There is a slight increase in the AL, which is balanced out by a slight decrease in the NL.
2. There is a striking change in the distribution of type of hit on artificial surface. Singles are consistently lower on the turf, but doubles and triples are dramatically increased.
Explanations
Any attempt to offer a mechanism to account for these results must directly consider observable player responses to the two different surfaces. Let us return to the initial premise: the ball moves faster on artificial surface. That assertion seems to be indisputable, not only by observation from the sidelines, but also by listening to the men who play the game. However, the consequences of this increased speed of ground balls is less obvious. The first thought might be that more balls will get through because of the greater speed. However, infielders clearly play several steps deeper on turf than they do on grass. This greater depth is justified even though it means they have a longer throw to first base because the greater speed with which the ball reaches them means that they have more time to make the throw. When an infielder plays deeper, his lateral range is significantly expanded, allowing him to get to more balls than would be possible on grass. Furthermore, when infielders play deeper, they are more able to catch bloopers over the infield (so-called "Texas-Leaguers"). I know of no study which charts the position of the fielders when each ball is put into play, but I am convinced, based on my own anecdotal observations, that the pattern suggested here is true. I have watched many games in Philadelphia (turf) and Memorial Stadium in Baltimore (grass), and the distance between the infielders and outfielders as they await a pitch at the two parks is noticeably different, with a much greater gap on the grass field, even though the overall park dimensions are similar and the outfielders are playing no closer to the fence in Baltimore.
If the infielders do alter their positions as suggested here, is there are any way to examine this data set for evidence of it? The bulk of this analysis has involved safe hits, which are of course a minority event. I therefore went through every play of every game and tabulated the number of times balls were hit to infielders on outs, including errors and fielder's choices where no one was retired as well. Furthermore, this tabulation was subdivided into balls hit on the ground and balls fielded in the air, the latter category including pop ups and line drives.
Finally, since there is a chance that varying amounts of foul ground will have an effect, foul balls, either caught or dropped for errors, were also tabulated. The results of this analysis are in Table 5. Rates are calculated by dividing each value by the adjusted at bats with foul balls removed.
Table 5.
Rates of foul balls and outs fielded by infielders, separated by
ground balls and balls hit in the air.
League Surface Foul% Ground% Air% Infield%
AL Grass .032 .280 .078 .359
Turf .033 .284 .077 .361
NL Grass .034 .288 .077 .365
Turf .031 .286 .078 .364
Both Grass .033 .283 .077 .361
Turf .032 .285 .077 .363
These results show clearly that the surface does not have an effect on the number of balls fielded by infielders, which would appear to be in contradiction to the claim asserted above. However, it is more likely that these numbers show that the fielders have altered their position on the field to maintain an optimum chance of getting to a ball.
Infielder positioning was posited as an explanation for the decrease in singles on artificial turf. The extra base hits require a different answer. Since the ball moves so much more quickly on turf, a ball which gets through the infield will have a much greater chance of getting past the outfielders as well. Although it is highly anecdotal, I clearly recall a game in Philadelphia in the late 1970s when Larry Bowa, the Philadelphia shortstop, just missed a ground ball to his left, sprawling on his face in a futile dive. The ball rolled to the wall in left center and the batter got a triple. It is highly unlikely that such a ball, even if had gotten through the infield, would have become a three base hit on a grass field.
Conclusion
One additional category worthy of a brief consideration is bunts. All bunts (sacrifices plus bunt hits plus non-sacrifice bunt outs) occur about 17% more often on grass fields than on turf. Since a bunt is an intentional play on the part of the batter, one must consider that this significant difference in occurrence reflects player perceptions of the difficulty of bunting on turf. There is an expected, although slight, difference in the success of non-sacrifice bunts as well, with 39.9% of them being hits on grass vs 36.6% on artificial surfaces. Although artificial surface seems to have only a minor effect on batting average, there is a clear effect on the pattern of safe hits. One might expect that the increase in doubles and triples on artificial turf would offset the lower rate of singles and lead to overall higher offense on the carpet. However, as shown in Table 6, scoring on grass is actually slightly higher (8.82 runs per game for both teams) than on turf (8.62 runs per game for both teams) during the past 11 seasons, although the two leagues differ in the direction of the change.
Table 6. Runs scored as a function of league and surface.
League Surface Games Runs Runs/Game
AL Grass 8664 78629 9.08
Turf 3466 31590 9.11
NL Grass 5474 46053 8.41
Turf 5202 43081 8.28
Both Grass 14138 124682 8.82
Turf 8668 74671 8.62
Note added in April, 1996. The analysis in Table 6 has a methodological error in that home runs are not excluded. The proper adjustment would be to subtract the number of runs attributed to home runs so that the effect of balls hit in play on scoring could be seen more clearly.
(Ed. note: The increase in NL scoring on Grass is due in large part to Mile High Stadium, which is a grass field and one of the most extreme hitter's parks in baseball history. It's a hitter's park due to the effects of playing at high altitude and the effect of the playing surface is overwhelmed by the altitude effect. A complete analysis of the effect of surface on run-scoring would control for other factors such as this.)
