JotCast | Dan Szymborski FanGraphs Chat

Dan Szymborski FanGraphs Chat - 7/15/21

Prev Page View Full Chat

Dan Szymborski

1:05

758 actual wins

Erik

1:08

Fair. But even two consecutive years of overperforming as dramatically as the Royals did sure felt significant. Might just be humans inventing patterns though.

Dan Szymborski

1:08

It was only really the one year of dramatic overperforming

the 6 wins in 2014 is basically an average error

1:09

And 14 isn't even in the top 20 for overperforming projections during ZiPS era

Este

1:10

Do all teams have their own version of Zips, just presumably with more data?

Dan Szymborski

1:10

Oh, this piece talks about team projection errors

https://blogs.fangraphs.com/the-zips-projections-midpoint-roundup-of-t...

Josh R

1:10

What is your neural network? I want to play with it so bad!

Dan Szymborski

1:10

It's one implementation of GPT-3 that I'm good at manipulating with custom generators and such

1:11

https://app.inferkit.com/generate

Teddyt

1:11

Pretty sure it was the Orioles that always over-performed expectations. Winning 90+ with ~75 projections

Dan Szymborski

1:12

Misses since 2005

1:16

ZiPS per franchise, 2005-2020

Guest

1:17

So Commish is contemplating a shift rule. Isn't it counterproductive to reward lazy hitting rather than making them figure out a more successful approach? He talks about getting the game back to when he was 12 years old but when I was 12 (I am 51) hitters attempted to move runners along and get them home from 3rd with less than 2 outs.

Dan Szymborski

1:17

I'm of multiple minds of this

ON a philosophical level, I agree

But on a practical level, most players are so much worse with a drastically changed approach

that it's not worth the time to try to change the approach

SO I'm torn

1:18

On that note, time for me to head out for another week