May 14, 2012 posted by Patrick DiCaprio

The Closer Identifier Algorithm and Predicting Closers

Print Friendly

Mariano Rivera, NYY

My articles for this year have introduced a new way to look at closers. The Closer Identifier Algorithm is a ridiculously simple algorithm that I am betting will do better in the long run at predicting the future than even the most skilled fantasy baseball analyst.

This is an experiment. We do not know if this particular set of criteria is the best and we may tweak it as we go along. But if my general philosophy is right this might be the next big idea in this game, and will not bode well for the future of traditional fantasy analysis. This is nothing less than an attempt to completely change how fantasy baseball players think.

My philosophy is that rather than doing a full-on analysis of the two or three pitchers involved, we use a simple algorithm to make a decision. Research has shown that simple algorithms are far better at predicting outcomes than even the best experts in any field.  Fantasy baseball and picking closers should be no different.  So here, is how we look at the decision:

1.Who has the job now

2.Does the current closer have a BPV over 80

3.Is his K/BB rate over 2.5?

4. Is his K rate over 7.0?

These are only rough guides so feel free to give me input if you think they are too low or high. But the idea here stems from the simple premise that a minimally-qualified pitcher with the job is likely to hold it, even if a better-skilled pitcher exists. If the answer is yes to all of these questions then we will in every case prefer the current candidate to the challenger. The old trope of “draft skills, not roles,” appears to have it reversed.

The early returns are pretty good: 6 correct predictions and one “miss” (Jordan Walden) for closers that lost a job because of non-injury.

Here are links to the articles so far:


You must be logged in to post a comment.