NFT: MLB analytics regular season vs playoffs

There should be flexibility in the approach in the playoffs. I know most will disagree but with Cole on the mound and a struggling Gardner at the plate, I think they should have bunted in the first inning.

In comment 14630947 Danny Kanell said:

Quote:

There should be flexibility in the approach in the playoffs. I know most will disagree but with Cole on the mound and a struggling Gardner at the plate, I think they should have bunted in the first inning.

him batting third was idiotic. Him not bunting was almost as dumb

The A's is a prime example.

I do believe over the past few years sabermetrics has evolved to take into consideration and value more recent performance but in the end, there's only so much a number can tell you to help you in such a small amount of games per series.

In the playoffs, the numbers are more like guides than actual answers. Making the gutsy move while understanding your players' clutch genes are much bigger factors.

You also have to consider that during 162 games, you're playing mostly mediocre talent, while in the playoffs you're competing against the cream of the crop. That in itself should make most data more or less useless.

Personally I think analytics can help you develop a playoff roster when budget is an issue, but come the playoffs that goes out the window. Where analytics teams tend to succeed is that they are able to stockpile an underpaid commodity like relievers, hitters that see a lot of pitches, etc. That works over a 162 game season where a team could go 2-3 weeks without a day off. In the playoffs you are going at most 2 games before a day off. That means a team that can consistently get 7 innings out of a starting pitcher only really needs 2 to 3 relievers for the playoffs.

like a hammer and a wrench, they can help you build an engine, but they're not the engine itself, they don't hit the gas pedal (or the brakes), and they certainly aren't driving.

But playoffs in general have pretty random outcomes. Short format, results are unpredictable. I think the best approach is probably just to stick with what got you there and hope it works while acknowledging it probably won’t work most of the time.

That analytics can put together the best regular season team but that the playoffs are basically a different animal

We developed mathematical models to determine the best course of action against a threat that was under surveillance. The longer the surveillance went, the more we could integrate information to our benefit. We did the "analytics" to get the best approach.

Then we had the problem of working the problem where there wasn't any time for surveillance. Instead of 1000 "looks", we got 7. So we had to do the best we could. What we ended up with was completely different (and in some ways) contradictory from our original approach. But it still moved the needle to our benefit.

When it comes to the A's, I think judgment should be tempered somewhat. If I worked for the A's and told them they aren't going to get it done in a short series and the best course of action for winning it all is to sign Max Scherzer and Steven Strasburg, that wouldn't be very useful to them.

In comment 14630994 Oscar said:

Quote:

But playoffs in general have pretty random outcomes. Short format, results are unpredictable. I think the best approach is probably just to stick with what got you there and hope it works while acknowledging it probably won’t work most of the time.

this is the correct answer. A playoff series is a small sample. 162 is a large sample. As fans we create narratives that people perform under pressure or about a money player.. it's all BS. I look at players and teams as random generators. If a team is truly better than another team in a series they are what? a 70% chance at best.. well 30% winners some in all the time.. even back to back to back. It doesn't mean the 30% team is BETTER than the 70% team. But fans and sportswriters fall all over themselves writing stories that they were money or better or clutch..