SIERA

[Shelton]

I haven't seen a thread on this posted elsewhere on the site. Please delete if redundant.

Baseball prospectus has created a new pitching stat called siera. Like fip, it takes a pitcher's component stats and converts them to an era equivalent.

Has anyone dug deep to post any additional thoughts? The first thing that sticks out to me is the credit it gives to groundball pitchers. It also penalizes pitchers that give up a lot of home runs if the pitcher also happens to allow a lot of baserunnets. A common example is Santana, who gives up a lot of bombs but isn't hurt as much by them because he allows so few baserunners. Other pitching predictor stats treat all home runs the same.

It's interesting to say the least.

[jaymo]

I hadn't read about QERA until I read about SIERA.

Since both are quadratic I think essentially the move from FIP to QERA/SIERA is like the move from linear weights to RC was for hitting statistics. It allows for interatction between alllowing baserunners and what comes next just like when scoring runs someone has to get on base and then someone has to hit the runner in.

Swartz and Seiddman seem to have been careful in eliminating terms that don't really interact.

On a side note I liked the use of calculus in part III to show that more strike outs help, but don't help as much as the first few. I haven't seen much calculus in baseball articles.

Finally, the next time we discuss BABIP and HR/FB being mostly luck and not skill I can refer someone to the series of articles and if it take them as long to read as it took me it will effectively and mercifully kill the thread.

[jaymo]

Another thought I had reading the article is that all the ERA estimators have an RMSE more than 1 when predicting next year's ERA.

That is, luck/randomness seems to play such a large role in ERA that no one will ever really get close.

[Scottwood]

Here is some good discussion on it.

http://www.insidethebook.com/ee/index.php/site/comments/siera/

http://www.insidethebook.com/ee/index.php/site/comments/what_does_siera_think_of_walks/#comments

[Scottwood]

I really would like to see a shift to RA, instead of putting everything on an ERA scale. ERA overrates extreme ground ball pitchers and guys that do not miss many bats. They are going to have more errors made behind them and they should be penalized for that. Brandon Webb was highlighted in their most recent article and I've often seen people mention the discrepancy between his ERA and FIP. However, ERA overrates him b/c of the amount of errors made behind him. His career RA is 3.80. If you put FIP on a RA scale, then his career FIP would also be 3.80. He really doesn't outperform his peripherals.

[Scottwood]

And, here is an excel spreadsheet for SIERA totals from last year. You will need to download it.

Interestingly, Verlander led all AL starters in this stat last season.

https://docs.google.com/fileview?id=0Bw77YMT0Wy2LNTg3OTdjZmQtMzNjOC00YTQ0L Tg5ZTEtNzdkNzE3YjliMWRk&hl=en

Verlander: 2.80
Greinke: 2.83
Lester: 2.99
Halladay: 3.09
Hernandez: 3.39
Edwin Jackson: 4.21
Porcello: 4.41
Galaragga: 4.95
Scherzer: 3.54 (10th in the NL last season among pitchers with 150+ innings)

[tiger337]

In theory, I like that SIERA includes interactions. For example, it's interesting to consider that a pitcher who walks a lot of batters will benefit from more from groundballs than one who doesn't walk so many. They key is how much of an improvement it will be over other component ERAs. THe formula for SIERA is really complicated and it's hard to justify using it on a regular basis unless it is significantly better than something like xFIP in predicting future ERA.

[sagnam]

THe formula for SIERA is really complicated and it's hard to justify using it on a regular basis unless it is significantly better than something like xFIP in predicting future ERA.

That's something that should be easily tested using historical data.

[tiger337]

ok, I now see they have tested it:

Baseball Prospectus | Introducing SIERA (http://www.baseballprospectus.com/article.php?articleid=10042#commentMessage)

[Who is the Drizzle?]

huh. John Smoltz 2.98. Pedro Martinez 3.59.

Soriano 2.22 (!), Street 2.48 (!), Valverde 3.37, Jenks 3.56, Capps 4.05 Rodney 4.27, Lyon 4.30

Jason Grilli 3.92
Cliff Lee 4.01 (?)
Jair Jurrjens 4.38

I don't pretend to follow all the math but the reasoning seems sound, and anything that includes GB is an improvement over FIP.

I've said that John Smoltz could be a bargain so now I'll say it again. And I didn't realize how good Rafael Soriano and Hutson Street appear to be. Nice move by the Rays.