THE BOOK--Playing The Percentages In Baseball

This is one area where MLBAM (and a lot of other people) are really behind the curve on. It seems pretty obvious to me that there’s money to be made in providing fans with this kind of information in the form of advertising revenue. But the people who have the most access - the leagues themselves - don’t seem to provide nearly the depth or quality of information that places like Sports Reference do.

(Nobody misses the boat, in my estimation, more than the Elias Sports Bureau. It’s no stretch to say that they have records of pretty much everything. And yet, look at their official website:

http://www.esb.com/

One book for sale! You’re kidding me, right?)

I think they position themselves like Warren Buffett: Just one Mutual Fund?

Tango, is there any merit to this Goals Created stat they are using. To me it would value assists to low, but I am not sure what it is trying to determine.

I seem to remember it having merit.  From what I remember, it normalizes all the goals to a common environment, and also aligns the assists down toward the goals.  Do you have a current reference for me to look at?

They have a definition in the Glossary.

Ok, it’s even simpler than I remembered.  It’s an ok definition.

Outstanding.  I have been waiting for this for a very long time.

For the Crosby example, his 84 assists come out to the equivalent of 6 goals scored.  Does that mean that a line that he is on will be expected to score, in addition to his goals he puts in the net himself, an extra 6 goals that can be credited to his inherent skill?

Right, that’s the implication.  I don’t believe this to be true however.  It must be more.

The problem with these “partitions” that plague Goals Created and Win Shares is trying to split things into absolutes, when really you are trying to split the value-added.

Take Crosby: he had 36 goals and 84 assists.  Let’s say that someone with his ice time (ev, PP, sh) would have scored 30 goals and 50 assists (just a reasonable guess).  Crosby is +6 in goals scored, and +34 in assists. 

(Because you get about 1.7 assists awarded per goals, I would turn that +34 into +20.)

His “Goals Created” was 42, based on what’s on b-r.com

Presuming that Crosby played with average players, the “absolute” method would credit him as being +12 goals above average (42 minus 30).

In my case, I would credit him with being +26 goals above average.

In short, the difference between my method and their method is that theirs reports numbers as half as much as mine.  This is the general complaint I have with all these Win Shares type methods.  I’ve gone through all this stuff when I was developing metrics as a teenager.  It simply doesn’t work.

One method takes half the goals and half the goal-adjusted-assists, and adds them up.  In my case, I take the full goals above the average and the full goal-adjusted-assists above the average, and add that to the average.  To be more specific, I do:

Tango = (G-avg) + (gA-avg) + avg = G + gA - avg

B-r.com = 0.5G + 0.5gA

As you can see, in my method, you can get into negative goals.

I can prove that my way works out better if I spend more time on it with the data I need.  I’ve done enough research to know that I’m right, and I simply haven’t done it completely enough so that others know it.