Untitled Document

[ Home ] [ Library ] [ Bookstore ] [ Contact ] [ Search ]

RBI's are meaningless?

by Keith Woolner


This essay was originally posted to the Red Sox mailing list on November 1st, 1995. 
The message is a response to an earlier posting, the relevant parts of which have been
reproduced at the beginning of the file.

To join the list send an email with the word "subscribe bosox" in the body of the message
to majordomo@world.std.com

Peter Kaczmarczyk (pkaczmar@indiana.edu) wrote: 
> 	I am so tired of hearing that RBI are a meaninless stat.  Yes, they 
> are team dependant and therefore maybe not as valuable as some of the less 
> raw stats that the stat people like, but hardly meaningless. 
>
>The team with the most Runs wins.  Logic therefore concludes that more RBI  
>will usually mean more wins.  Sounds very meaningful to me. 
>
>	What frustrates me the most is the assumption that raw stats of 
>this kind should be ignored, thrown out or otherwise discarded as 
>completly useless.   
> 
>Do your numbers on Joe Carter [showing Carter to be above average, but
>not remarkable at driving in runs despite his lofty RBI totals]
>cover the San Diego and Cleveland years, when I  
>would assume that he had fewer men on base in front of him than in in  
>Toronto?   
>[...]
>
>	All of this also ignores that simple, easy to understand numbers 
>are extremly meaningfull and usefull to the casual fan who wants a quick 
>read on the general value of a player.  [...] I don't want to sit there and  
>try to plug his numbers into some formula, (from RBI % to PkPsoPUPU....)  
> 
>Player got important RBI, team won.  Those RBI are awfully meaningfull to me  
>in that situation.

From:     Keith Woolner <KWOOLNER.US.ORACLE.COM> 
To:       bosox @world.std.com
Subject:  RBI(s) and the meaning of meaningless

 pkaczmar@indiana.edu wrote: 
> 	I am so tired of hearing that RBI are a meaninless stat.  Yes, they 
> are team dependant and therefore maybe not as valuable as some of the less 
> raw stats that the stat people like, but hardly meaningless.

I agree with you (probably to the shock of many on the list).  The  
statement "RBI's are meaningless". is pretty demonstrably false.

As with many controversies on this list (and elsewhere), much of the problem  
here is a misinterpretation of the terms being used.  Let's focus on 
"meaningless" in this context.

Suppose I told you that Joe was a baseball player who appeared in the major 
leagues last year, but I told you nothing else about him.  If I asked you 
how good a hitter he was (or more completely, how likely you thought he 
was to be a bad, average, or good player relative to the league), you'd 
use your background knowledge of how talent is distributed among major league 
players, and probably come to some conclusion that he was pretty likely to be 
bad, a small chance of being good, with average somewhere in the middle. 
For argument's sake, let's say that the likelihoods were 50% bad, 30%  
average, and 20% good (ignoring for this argument how we define those  
categories).  This is your base case for how good a major leaguer 
is as a hitter, given no other information.

Now suppose I also told you how many RBI's he had last year.  In order for 
that figure to be "meaningless", no RBI total that I could possibly tell 
you would influence your beliefs about how good a hitter he is.  If any 
knowledge about how many RBI's he had changes your assessments of his 
offensive abilities, then the RBI information is *not* meaningless.

For example, if I told you that Joe had 130 RBI's last year, you'd probably 
change your thinking about him a little bit.  First of all, we know that 
130 is a lot of RBI's for a single season, and so, at the very least, 
Joe must have had a lot of playing time.  We also know that teams rarely 
play a player they know to be a bad hitter regularly (ignoring positional  
considerations) for obvious reasons.  Even given a full season of playing 
time, a total of 130 RBI's is a unusually high total, which means the 
hitter was probably significantly better than average at driving runners 
in, and therefore might be a good hitter overall.  Note that there 
is an implicit "rate of production" here, namely RBI's/season.  So, in 
addition to knowing that the player probably had a lot of playing time, 
he also was probably above league average at driving runners in.

Taking all of this into account, we might update our beliefs about the 
likelihood that Joe is a bad, average or good hitter.  For example, we 
might now believe that there's only a 10% chance that he's a bad hitter, 
a 60% chance that he's good, and 30% that he's average.

The fact that we could update our beliefs about Joe given information about 
RBI's means that that information is relevant to offensive ability.  Hence, 
RBI's are not meaningless!

As an aside, this doesn't mean that *every* RBI total will be useful in 
updating our beliefs about Joe.  If I told you that Joe had, say, 40 RBI's, 
that might not be as clear.  You can't tell, based on that number alone, 
whether Joe had limited playing time (and perhaps batted in runners well), 
or played all year (but was a poor hitter).  He might have been a defensive 
replacement, or a AAA callup for injured regular.  You might decide that, 
even with the knowledge of his 40 RBI's, you aren't going to change your 
beliefs about his ability as a hitter from your base case (50/30/20).  But 
the fact remains that since some possible RBI totals can influence your  
beliefs directly implies that RBI's are meaningful.

So why do statheads insist that RBI's are meaningless?  Well, the problem 
is that they/we aren't being very clear in our usage.  Almost everything 
that RBI's can tell you about a player's ability can be found in other 
measures that we prefer to look at, for various reasons.  In particular, 
plate appearances (or at bats, or games played) provide more direct and 
unbiased information about playing time, and slugging average (SLG) provides 
a better measure related to the rate of baserunner advancement potential.  If 
you know these things (and, to a lesser degree, other preferred stathead 
measures, like OBP), you can update your beliefs about Joe to a degree of 
certainty that adding the additional knowledge about RBI totals doesn't, 
in any circumstances, noticeably change your assessments.

In particular, since there doesn't seem to be a clutch hitting ability (as 
has been demonstrated by numerous studies), SLG captures nearly all of 
the information about a hitter's ability to drive in runners on base.   
Alternatively, you could look at percentages of runners driven in by the 
batter in different base-occupied situations, though this info is rarely 
available, and SLG has been shown to correlate with it quite well.

As others have pointed out, the RBI measure has flaws in terms of 
teammate dependency.  When we can, we prefer relatively teammate independent 
measures like OBP and SLG to evaluate individual players.

>The team with the most Runs wins.  Logic therefore concludes that more RBI  
>will usually mean more wins.  Sounds very meaningful to me.

Yes, the team with the most runs (and usually the most RBI's) wins.  This 
does not mean that the player with the most RBI's on the team is the 
most responsible for the team's success.  The "accounting method" for RBI's 
doesn't accurately reflect contribution to winning, and therefore, a  
player maximizing his own RBI's isn't necessarily maximizing his team's runs.

>	What frustrates me the most is the assumption that raw stats of 
>this kind should be ignored, thrown out or otherwise discarded as 
>completly useless.

This certainly isn't my position.  The raw stats need to be considered 
in context (whether quantitatively or qualitatively), but they are important.

>Do your numbers on Joe Carter cover the San Diego and Cleveland years, when I  
>would assume that he had fewer men on base in front of him than in in  
>Toronto?

You assume incorrectly.  They do cover San Diego and Cleveland, where he also  
saw unusually high numbers of men on base during his plate appearances.  Bip  
Roberts, Tony Gwynn, Brett Butler, Julio Franco, et al, all ran excellent  
OBP's during Carter's years with them as teammates.

>As well, while his percentage may not be ahead of the league average, this 
>doesn't reflect the fact that he gets *enough* runs in to make his team a  
>winner.

How many pennants did Cleveland and San Diego win during his time with them? 
What was the difference with the Blue Jays?  Toronto won because they had a  
better cast around Carter, not because of Carter himself.

But the point is still that indications are that his team's would have 
scored *more* runs, with a player who got on base more frequently than 
Carter did, even it it meant fewer RBI's out of that slot in the lineup.

Again, maximizing individual RBI's != maximizing team scoring, and 
therefore, maximizing individual RBI's != maximizing team wins.

>	All of this also ignores that simple, easy to understand numbers 
>are extremly meaningfull and usefull to the casual fan who wants a quick 
>read on the general value of a player.  [...] I don't want to sit there and  
>try to plug his numbers into some formula, (from RBI % to PkPsoPUPU....)

Agreed, simple measures are better than complex ones, given equal accuracy. 
I must point out, however, that most baseball fans are comfortable with 
batting average, and OBP & SLG are no more complicated that AVG.  In fact, 
OBP is considerably simpler a concept than AVG.  The desire for simplicity 
is exactly why most statheads fall back on OBP & SLG to do quick evals of  
players, much as traditional fans would look at AVG, HR, and RBI.  It may not  
be what you grew up with, but they aren't really more difficult.

Of course, baseball is a complicated game, and there are lots of different 
factors that influence the outcomes of games.  To do a serious analysis, 
rather than a quick-n-dirty one requires careful consideration of these 
other factors, which is why the more esoteric measures (Linear Weights, 
MLV, VORP) were developed.  But for most casual uses, OBP & SLG will do 
just fine.

>Player got important RBI, team won.  Those RBI are awfully meaningfull to me  
>in that situation.

Player C may have gotten the RBI, but it could very well be that Player 
A drew a walk from the ace pitcher after fouling off numerous pitches.  He 
then advanced to 2nd on player B's single to left.  Player A might have 
then stole third on a double steal, and scored on C's sacrifice fly (which  
would have been a routine out if A & B hadn't done such a superb job getting  
themselves in scoring position).  C gets the RBI (and the big contract), but 
was really responsible for that run?

An RBI may be the icing on the cake, but it doesn't tell you who was the 
best baker.

[ Home ] [ Library ] [ Bookstore ] [ Contact ] [ Search ]

Last Updated: Contact webmaster@stathead.com for corrections or problems