The Value of Consistency

In a recent post, I developed a statistic to quantify the consistency of a player's performance over a season. Using the consistency statistic (the coefficient of variation of GPA), I determined the top 10 most consistent and the 10 streakiest (or least consistent) hitters in 2007. Looking at the two tables, one thing I noticed was that consistent players tended to have better seasons than the streaky players.

To examine the difference between consistent and streaky hitters further, I split the qualifying major leaguers into three equally sized groups according to the the consistency stat. I placed the 65 most consistent hitters into the high consistency group, the next 65 into to the medium group and the rest into the low consistency group.

The results in Table 1 below show that the high consistency group performed better, on average, than those in the other groups. In fact, they were statistically significantly better on the following stats: batting average, on base percentage, slugging percentage, OPS, GPA and K/600 PA. The differences between groups in isolated power and BB/600 PA were not statistically significant.

Table 1: Performance by Consistency Category - 2007

Consistency	BA	OBP	SLG	OPS	GPA	ISO	BB/600 PA	k/600 PA
high	0.293	0.364	0.474	0.838	0.282	0.181	51	88
medium	0.280	0.346	0.443	0.789	0.266	0.163	48	95
low	0.266	0.339	0.435	0.774	0.261	0.168	51	101

I did a similar analysis for 2006 and came to the same conclusion. That is, in each year, players who performed most consistently in GPA had better numbers at the end of the season than streak hitters.

As a next step, I looked to see whether consistency within a season was likely to be repeated from one season to the next. I calculated the consistency in GPA for each of the 137 players with sufficient playing time in each season and found very little correlation (r=.15) between the two seasons. Thus, it appears that consistency in overall performance (GPA or OPS) is not a repeatable skill and is actually fairly random.

My next thought was that changes in performance could be linked to changes in consistency. To test this theory, I took the difference in GPA for each player from 2006 to 2007. I then took the difference in the consistency for each player from 2006 to 2007. Finally, I looked at the correlation between the two differences (r=.30) and found it to be statistically significant.

In other words, players who improved their consistency from 2006 to 2007 tended to do better in 2007 than 2006. Conversely, players who became more streaky from 2006 to 2007 tended to see a decline in performance. This can be better illustrated by looking at some lists of players. Table 2 lists the 10 players who improved the most in consistency from 2006 to 2007. You can see that 7 of 10 also improved his GPA. Conversely, Table 3 shows the players who regressed the most in consistency from 2006 to 2007. In this case, all 10 also saw their GPAs decline.

So, it does look as if performance is tied to consistency even though consistency is not repeatable from year to year. That is, consistency describes something about a player's season more than it describes a player. If a player has an inconsistent season by the consistency statistic, that might be an indicator that he was injured or had some unusually bad luck which prevented him from hitting his best all year long. It also might mean that he's a candidate to improve in the following season. I'll look more into that later.

Table 2: Players who became more consistent from 2006 to 2007

first name	last name	Consisency change	GPA 2006	GPA 2007	GPA change
Brad	Ausmus	-18.0	.207	.223	0.015
Nick	Markakis	-14.6	.269	.284	0.015
Victor	Martinez	-13.1	.292	.298	0.006
Hanley	Ramirez	-12.2	.278	.313	0.036
Prince	Fielder	-11.5	.277	.332	0.055
Carlos	Delgado	-11.3	.299	.262	-0.038
Edgar	Renteria	-11.2	.270	.292	0.023
Mark	De Rosa	-11.0	.274	.271	-0.003
Chase	Utley	-10.4	.302	.326	0.023
Miguel	Tejada	-9.5	.295	.271	-0.024

Table 3: Players who became less consistent from 2006 to 2007

first name	last name	Consisency change	GPA 2006	GPA 2007	GPA change
Melky	Cabrera	17.6	.258	.243	-0.015
Jason	Kendall	16.5	.249	.211	-0.038
Ronny	Paulino	15.1	.260	.239	-0.022
Andruw	Jones	13.8	.296	.243	-0.053
Brandon	Inge	13.7	.256	.233	-0.023
Carlos	Beltran	13.1	.323	.290	-0.033
Jermaine	Dye	12.8	.328	.264	-0.064
Tadahito	Iguchi	12.7	.262	.255	-0.007
Ivan	Rodriguez	12.7	.257	.237	-0.020
Mike	Cameron	11.2	.280	.255	-0.025

The information used here was obtained free of charge from and is copyrighted by
Retrosheet.  Interested parties may contact Retrosheet at "www.retrosheet.org".