It's the Outs, Stupid!

Dave Paisley

As I'm on something of a statistics kick these last few weeks (a result of the Mariners plummeting out of contention for anything but the "who gets fired next" speculation lists) I thought I'd continue by delving into some more mysteries of player evaluations. Last week I took a look at some recent data on team runs scored. Today I'll be looking more at individual players.

One thing about looking at a whole season's worth of team stats is that we tend to think that teams get a certain amount of plate appearances a year, and that they get divvied up between players about evenly. Not so fast, my hasty friend. The most constant factor of annual team offensive performance is the number of outs. Face it -- every game, each team gets 27 outs to play with. The exceptions are in a home win when the home team is leading after eight innings, where the team only makes 24 outs, and extra inning and rain shortened games. The numbers invariably show that teams average 25.2 outs per game, with a tight variation around that. In the 1996 and 1997 seasons combined, the variation between the most and fewest outs per game was a measly 5%.

Plate appearances can be roughly summarized as outs plus hits plus walks, so the flip side to outs is hits plus walks, both of which are generally good things. Again, in 1996 and 1997, the average number of hits plus walks was 12.7 per game, but with a much larger variation than outs -- 30% difference between highest (14.5) and lowest (11.2). This is generally the difference between good and bad offensive teams. The more hits and walks you get per out you make, the more runs you score.

This is pretty significant, especially when we get down to individual players, where the differences are more marked. Recently, I've looked at the differences between OPS and Runs Created as measures of evaluating player performance. Overall, I'm still trying to understand what RC offers that OPS doesn't. I thought I'd take a look at how broad a variation in Runs Created I could get if I held OPS constant

The following table shows eleven variations on a theoretical 1.000 OPS hitter. The variation is in the way OBP and SLG are split up. They vary from an extreme low OBP, high SLG player (.300/.700, think extreme Jose Canseco) to an extreme high OBP, low SLG player (.500/.500, think extreme Edgar Martinez) I pegged batting average at 80% of OBP, so the .300 OBP guy is hitting .240, and the .500 OBP guy is hitting .400. Each player gets 600 plate appearances.

OPS	OBP	SLG	H	BB	TB	RC	OUTS	RC25	% Nominal
1.000	.300	.700	133	47	387	120	420	7.11	75%
1.000	.320	.680	140	52	373	124	408	7.58	80%
1.000	.340	.660	148	56	359	127	396	8.04	85%
1.000	.360	.640	155	61	345	131	384	8.50	90%
1.000	.380	.620	162	66	331	133	372	8.96	95%
1.000	.400	.600	169	71	317	136	360	9.42	100%
1.000	.420	.580	176	76	304	138	348	9.88	105%
1.000	.440	.560	183	81	290	139	336	10.34	110%
1.000	.460	.540	189	87	277	140	324	10.80	115%
1.000	.480	.520	194	94	263	141	312	11.27	120%
1.000	.500	.500	200	100	250	141	300	11.73	125%

For players in the 1.000 OPS class, the average split tends to be around .400/.600 (highlighted in bold red), with lows of .380/.620and highs of .440/.560 (shown bold), so reality isn't anywhere near as extreme as the spread in the table. From the table, you can see hits and walks increase markedly as OBP goes up, and outs goes down dramatically - from 420 to 300. The basic Runs Created formula shows that the higher OBP player at the bottom of the table creates a modest number of runs more than the .400/.600 guy. Interestingly, as OBP drops, the effect gets very non-linear, so the worst guy creates 16 runs less than average, while the best guy only creates only 5 more.

Runs Created within our normal player range (.380 to .440 OBP) only varies from 133 to 139, so a value derived from an OPS prediction somewhere in the middle would only be in error by about 2% at most. So, once again, I don't see the big hoo-ha over RC.

But all of this analysis is for 600 plate appearances. Is there a better way to evaluate what these players are creating? To get back to my original thought, the most precious commodity an offense has is outs, and these hypothetical players vary a great deal in how well they use outs, as a casual glance at the table shows. One thing the RC crowd has popularized is a little thing called RC25 (or it's bigger, uglier cousin, RC27). RC25 is basically Runs Created per 25 outs.

Imagine each player taking every plate appearance in a game. How many runs would they score given 25 outs? Rey Ordonez - not many; Mark McGwire - a whole bunch. Of course, there would have to be pinch runners galore to run the bases for this exercise, but it would be a lot of fun. Would you intentionally walk Mark McGwire to get to Mark McGwire?

As we've calculated RC already, itís easy to calculate RC25. Of course, you can do this "per 25 outs" deal with any number - Runs, RBI, Monkeys -- whatever. It's just a much smarter way to look at the relative value of players. Our RC25 column shows that within the normal band (.380 - .440 OBP) the difference in RC25 production is now 15% from top to bottom rather than the measly 4% difference in RC.

So, all 1.000 OPS hitters aren't made alike. But the key point to remember here is that even a well correlated constant OPS generated number would only be off by a maximum of 7% or so from the median, and when a minor mental compensation factor (think - Jose Canseco bad, Edgar Martinez good) is used, it really isn't going to lead you astray much.

Key
OPS  = On base percentage Plus Slugging average
OBP  = On Base Percentage
SLG  = Slugging Average
H    = Hits
BB   = Walks
TB   = Total Bases
RC   = Runs Created
RC25 = Runs Created per 25 Outs

about the author

It's Dave Paisley's lifelong ambition to create a Unified Field Theory based on OPS. Send transcripts of all your Albert Einstein channelling sessions to him at drdjp@strikethree.com.