Est. Wins using Fip for and Against

by Sandy Kazmir on Jun 27, 2009 4:04 PM EDT Comment 10 comments

After reading Matthan's fanpost this morning, I wanted to see what my offseason analysis called Est. Wins looked like through this point in the season. This was a first look at how many wins a pitcher should have had. It was rough and rudimentary. Here is what it looks like updated through 75 games:

Pitcher	R	IP	R/IP	Starts	PROBw	Est. Wins
Shields	2.88	6.77	0.42	16	0.522	8.35
Garza	2.67	6.49	0.41	15	0.556	8.33
Niemann	2.71	5.17	0.53	14	0.477	6.68
Sonnanstine	4.07	5.44	0.75	15	0.354	5.31
Kazmir	4.56	5.07	0.90	9	0.314	2.83
Price	3.67	5.06	0.73	6	0.429	2.57
Team	3.31	5.81	0.57	75	0.454	34.07

After conversing back and forth with Matthan I wanted to see this in FIP format, first the data, then the explanation:

Pitcher	FIP	PROBw	Starts	Est. Wins
Shields	3.79	64.3%	16	10.29
Garza	4.67	50.8%	15	7.61
Sonnanstine	5.63	46.3%	15	6.95
Niemann	4.81	44.6%	14	6.24
Kazmir	6.11	41.9%	9	3.77
Price	5.82	34.2%	6	2.05
Team	4.96	49.2%	75	36.92

First off, check out the workbook on google docs or download the file FIP win% 75 games. I ask to do this so the process doesn't come across as complete gibberish. Using Baseball-reference's gamelogs I was able to come up with a game by game FIP for opposition pitchers (FIPa). It stands to reason that if our starter has a lower FIP (FIPrays) he should be credited with a win. From here I was able to create a matrix comparing each individual FIPa to the entire population of Rays starters FIPs. Using conditional formatting (green is >1, or that cell represents an event where the FIPrays > FIPa) I came up with a total number of events where our performance should be better than theirs. This yielded a probability of a win for each FIP that we have had this year (% of games on Matrix tab).

From here I moved on to the "Data" tab of the workbook. By comparing each actual game FIP to the probability of winning at that FIP (PROBw) I was able to get an idea of estimated wins. This was then totaled and put into the summary that you see before the jump. The team FIP is a weighted average based on Individual FIPs and games started. The PROBw is the likelihood of a starter winning any given starter, based on Est. Wins / Starts.

I used 3.15 as my FIP constant term. Our starter FIPs are based on actual Innings Pitched, while FIPa is based on all opposition pitchers, assuming 9 innings/game. Overall, I really like this, and in the future will make the matrix a little bit different so it is easier to update. It took a long time to do the matrix manually, so if anyone is any good at programming/macros it could be a lot easier in the future. Your thoughts?

2 recs | Comment 10 comments | Share

Story-email Email Printer Print

Comments

Very nice work

I think it all this work compliments each other very well. I think what the common theme is Sonnanstine has not been that bad, and to a lesser extent Kazmir. Their volatility in their starts is what has driven their metrics up, but on a game by game basis they have been fairly solid.

Looking at game by game data is very important. The total season number is a good rough estimate, but we aren’t playing roto fantasy baseball here. There is a huge difference in real results between a starter getting rocked in one game and throwing a cgso the next compared to throwing the average of the two games twice in a row. It is a very good starting point to try to capture this difference.

by matthan on Jun 27, 2009 4:20 PM EDT reply actions 0 recs

google doc referenced isn't currently shared

Follow Me on Twitter @FreeZorilla

by FreeZorilla on Jun 27, 2009 4:36 PM EDT reply actions 0 recs

Thanks for letting me know.

Rays Win!

by Sandy Kazmir on Jun 27, 2009 4:39 PM EDT up reply actions 0 recs

I also think my work combined with yours paints a certain picture with Neimann

We all have a history with Sonny, so to see that he really hasn’t been THAT bad shouldn’t be surprising. However I do Neimann is an interesting case. I’m going to dig deeper into that.

by matthan on Jun 27, 2009 4:37 PM EDT reply actions 0 recs

I also think the next step would be to use confidence intervals

For both FIPagainst and FIPraysstarters. This should give us more predictive power on how our pitchers should perform going forward. Of course this is somewhat similar to that type of angle so I doubt the differences would be too large.

by matthan on Jun 27, 2009 4:38 PM EDT reply actions 0 recs

My next move is to do this same thing for the Yanks, Sox, Jays, and maybe some of the extreme teams

example, Texas (great offense, so so pitching), Oakland (decent pitching, horrible offense), to see some of those effects.

Rays Win!

by Sandy Kazmir on Jun 27, 2009 4:40 PM EDT up reply actions 0 recs

that cell represents an event where the FIPrays > FIPa

This sign is backwards, in the matrix, anything greater than 1 indicates the Rays FIP was lower (better) than the opposition FIP

Rays Win!

by Sandy Kazmir on Jun 27, 2009 4:42 PM EDT reply actions 0 recs

This and matthan's stuff have been great

I am hopeful that Sonny’s demotion is a short term thing to set up something else. His command has been very inconsistent this year, yet even still he has the characteristics you want in your 4 or 5.

Follow Me on Twitter @FreeZorilla

by FreeZorilla on Jun 27, 2009 4:46 PM EDT reply actions 0 recs

He hasn't pitched well

However a few really bad games have really hurt him. It caused all his metrics to be really high. It is the same issue as Kaz. The bad games created some media led monster that that causes most people to overlook his otherwise decent games.

Of course I’m arguing that Sonny and Kaz have been better than what most people think that doesn’t mean that a demotion wasn’t in order. Both players were pitching under their talent level. The first priority needs to be to get them producing at their talent level.

by matthan on Jun 27, 2009 4:50 PM EDT up reply actions 0 recs

But using your study is a good indication to relative performance

underperforming true talent level does not necessarily mean a demotion is in order. THere were larger things in play, namely option availability. Kaz was a different case. He was getting progressively worse. He needed a step back for awhile to regroup and get back to business.

Follow Me on Twitter @FreeZorilla

by FreeZorilla on Jun 27, 2009 4:53 PM EDT up reply actions 0 recs