Given the return of Scotty Kazmir, the demotion of Sonny, and questions surrounding the starting pitching in general I figured I'd take a deeper look into our Super Six Starters.
By now we all know where each pitcher stands with their season long FIP and tRA as well as where they stand among nearly useless but popular metrics such as wins/losses and ERA. FIP and tRA are excellent, and they do paint most of the picture. However I do think they miss a key component: volatility.
Volatility is important because teams are not evaluation based upon cumulative 162 game statistics. Teams are judged based upon 9 inning incriments. There is a difference between two pitchers if one has a FIP/tRA in three games ranging from 0, 6, and 6 and another pitcher with a FIP/tRA of 4, 4, and 4. The team of the first pitcher likely goes 1-2 whereas the other team actually has a shot of winning three games. As we certainly all know W's/Ls in the public context in relation to pitchers is largely irrelevant, however, the number of times a pitcher pitches well enough to give his team a shot at winning is largely ignored and is ultimately very important.
For the time being all of this is another way at looking what has happened. I have no idea whether it has any predictive power so lets rather use at as another way to explain 2009 so far.
There are quite a few ways to look at volatility and I hope to do so in the future. In this piece we are just going to look at it in general with a simple introduction.First and foremost I took the FIP of each game started for each pitcher**. Obviously each pitcher has a different number of starts, but here are the FIPs for eachs of our starters for each start.
Firstly I apologize for the graph being tough to read. Basically some general highlights. Price has the best start. Sonny has the worst start. Sonny and Kaz both have two starts that are far worse than their others. Shields worst start was his first and has been solid ever since.
Here are some basic statistics:
All three of these metrics are fairly important. Firstly I calculated the "Yearly FIP" by using the cumulative season counting stats for each pitcher. I also average the FIP for each start for each pitcher. They certainly are different. Both are important in their own ways. However what is even more interesting is the standard deviations. Sonny and Kaz have by far the most volatility. Neimann has the least, which is quite a bit surprising. However is a high standard deviation good or bad? I'll glimpse into that in a bit but here is what I see. Neimann is consistently below average. That is going to result in poor losses (with our elite offense a bit more wins) since he has no volatility resulting in excellent games. Kazmir and Sonny both have higher FIPs, but they have far higher volatility. Therefore they'll have more games falling into the "winning" category due to their volatility. Generally if you have a great FIP you want low volatility, if you have a high FIP then more volatility is good. All of this with the overall goal in mind, maximing wins and losses for the team.
This next chart we are going to look at the individual volatility for each pitcher. 0% is set as the average FIP per start for each pitcher. Anything over 0% is something higher than their individual average. Anything lower is less than their average. Of course 0% for Kazmir is far higher base FIP than it is for Shields.
The few things I take out of this chart. Sonny clearly is the most volatile around his average. Neimann is very consistent. Garza is remarkably inconsistently consistent. He bounces above and below his mean slightly but very often. Kazmir and Sonny have two starts that really crush them. Shields has been improving as compared to his average (as has Neimann).
The last thing I want to share is something I find very interesting. As you've seen have had quite a bit of W and L theme to this post, and this is ultimately why. I cannot emphasize enough how wins and losses in the context of MLB pitchers are irrelevant. However my modifed W and L are actually a fairly good indicator of what each pitchers record SHOULD BE given the team that they are on. Typical wins and losses are based a lot on luck, run support, bullpen, and distribution of runs. I take all of those variables out here.
I assume that the Rays RPG has no standard deviation. We score our runs per game average every game. After 74 games this is 5.66
I assume that all 9 innings of the game is performed at the FIP of the starting pitcher. Therefore taking the bullpen out of the question
Therefore if the players FIP for that game is less than the Rays average RPG then they put the team in position to win. On average they should have won the game. If it is over the Rays RPG then they deserve to have lost. Clearly the variability of runs scored is important. A 6 FIP clearly gives the team a better chance of winning than a 15 FIP, and this modified W/L ignores that. I'm merely counting the times above and below the Rays RPG
According to this Sonny should have the second most Ws on the team. Why? Because of volatility. He has had 10 games where he has put up a lower FIP than what the Rays score. Sure some of those 5 games he "lost" were pretty bad, but they all count the same. Kazmir fits the same bill. Now lets take a look at Neimann and Shields. They've been our two most consistent pitchers. However Shields has been consistent on the good side and Neimann on the below average side. You see this in the hypothetical W/Ls. Even though the Rays have the highest RPG in the league Neimann has only truly pitched to a .500 record. This is due to his low volatility at a FIP right around our RPG. Shields**** has low volatility, but has a low FIP resulting in tons of good results.
Clearly a lot more work can be done with this type of analysis. I'm particularly interested in downside risk. However hopefully this gave some folks a different angle onto how to look at our pitchers.
**My FIP will be a bit different than what you see on FanGraphs. I'm using a different constant and only starting pitching innings. It won't truly matter as this is more of a comparison between starters and I used the same formula for each starter. However be very cautious if you try to compare one of these guys in this analysis to numbers you pull off of FanGraphs or another reference
****Shields start on Friday June 26th is not included