Seeking discussion of 'Curve Ball' by Albert, Bennett

[reviewboy]

Please forgive me if I should have been able to find this, but the words 'curve', 'ball', 'albert', and 'bennett' are very common.

My request: if the stats book 'Curve Ball' has been discussed and critiqued in baseball-fever, could someone please give me a hint as to where I can find the discussion?

Thanks.

[reviewboy]

Looks like I’ll have to start the discussion.

By the way, I always feel bad when I post a question that gets a couple of dozen views but no responses, as though people expected more from my original post. Sorry about that!

I’m about 80% of the way through reading Curve Ball. I’ve learned a lot about baseball statistics, but something bothers me about a portion of the book.

Have any of the rest of you read this book?

In 1-2 places, the authors will open a chapter saying something like, “Let’s suppose for now that players exhibit a randomized performance, based on a certain base ability level…” (I don’t have the book in front of me, and I’m sure I’m not using the same verbiage that the authors did) and then go on to explore areas such as hot and cold streaks, how many performances (for example, at bats) are necessary to be statistically significant, etc.

They go on to describe experiments that model player performances with computers, where the simulation software has a randomization factor applied to a base level of performance, a computer version of this Cadaco All Star Baseball game spinner (you'll have imagine a spinner in the center): http://major-smolinski.com/RANTPIX/ASB1.jpg . The program performs many instances of the same simulated event, for example, at bats for a particular batter, and then reports the statistical results, for the authors calculate standard deviation, etc.

But, as far as I could tell, the authors never revisit their ‘randomization’ assumption as applied to real world baseball players.

I can see that baseball players’ performances have a certain element of randomness (upon a base level of ability), for example, in terms of pitchers and batters attempting to out-guess each other in terms of the type of pitch to be delivered.

Even so, it seems that players would have good days and bad days, unrelated to randomness. We all have good and bad performance days in our lives, right? A bad day can be based on not sleeping well, not eating well, low blood sugar, personal distractions, having to travel on the road and being away from one’s home, etc. The resulting performance may seem randomized if the reason for the bad day isn’t known, but the fact is that the causes aren’t really random. That is, a particular cause might affect our performance for weeks at a time, and then go away forever.

While I respect the authors’ judgment in not venturing into areas of expertise in which they are not qualified, such as physiology and sports psychology, the nature of their real world randomization assumption brings into question some of their conclusions in my mind, for example those involving hot and cold streaks.

I would be very interested in hearing the opinions of you greatly knowledgeable baseball and statistics experts on this subject.

Thanks.

[stevebogus]

I haven't read the book, but I'll comment on the random nature of baseball.

The outcome of any pitch is subject to both skill and chance. No pitcher, no matter how skillful, can always deliver the pitch of his choice exactly the way he wants to. No batter, no matter how skillful, can always hit any pitch. Even if he gets the exact pitch he is looking for he will not be able to always hit it the way he wants to. For these reasons (and others) there is a certain amount of "scatter" in the results. The logical way to model this scatter is by introducing a random element. Pitching and hitting baseballs are not random acts. The path of the ball is governed by the laws of physics. But, like quantum physics, the results of any interaction cannot be definitely predicted. The best we can do is to try to calculate the odds of different results based on what we know of past results.

If randomness can explain hot and cold streaks do we have to accept that explanation? No. A player's skill level is not static. It depends on innumerable things that we cannot know. Is he healthy or ill, tired or rested, confident or nervous? But by understanding how randomness can affect a player's stats we can better recognize when randomness ceases to be a likely explanation for an event.