Once again, I’ve updated my Estimated Wins workbook for the entirety of 2010 for the Rays. You may recall that this is an attempt to normalize the day-to-day variation in run scoring that can lead to this kind of year-long run support (these seem high, but then again, who goes 9 innings these days?):
The runs that are scored for you are completely out of the pitcher’s control in the American League. Instead of using the record as established by the rules, what if we assigned wins (or portions of wins) based on a pitchers performance compared to the performance of every pitcher that we faced this year? For example, what if Shields allowed 3 runs and in 75% of our games the Rays score 3 or more runs? Well, we’d say that James Shields deserves to win 75% of the time or .75 wins for that performance. Instead of using runs scored for this what if we used the pitcher’s FIP or wOBA for the amount of time that he was in the game vs. 9 innings of whatever FIP or wOBA our bats were able to put on the opponent? If you can grasp these, then you can understand the basis of this workbook. If David Price has a 2.55 FIP for a start, as he did in his first of the year, we would expect him to win 85% of the time and assign him with .85 wins for that start. In another start he had an FIP of 4.40, which is pretty close to league average. Our offense outperforms this level 48% of the time so we would assign .48 wins for that performance. We can total these up for everyone and get this:
We use FIP to get a look at just the three things that a pitcher can control (Ks, BBs, & HRs) which completely removes defense from the equation for both teams. wOBA takes into account what actually happened off the bat. It properly scales (w means weighted in this case) each of the following outcomes per plate appearance: BB, HBP, 1B, ROE, 2B, 3B, HR Because it looks at the results of a plate appearance then we are not removing defense and only looking at the results. Good defenses can make a pitcher look better, but within a team that uses similar defensive alignments from game-to-game, then we can have an idea of how well a pitcher is doing his job compared to his peers.
One caveat to the above chart is that for the FIP and wOBA columns, these are averages of each start, so it’s not their FIP or wOBA for the entire year, but an average of each of these for every start. The idea behind this is if two guys have identical game-to-game averages perhaps we can find that one of the pitchers always has a 4.00 FIP, while the other will have 2.00 in half his starts and 6.00 in the other half.
Which guy is more valuable? On the Rays, we have a pretty similar case between Wade Davis and Matt Garza above. Davis had a game-to-game average FIP of 5.12, while Garza had an average of 5.24. Meanwhile, by this metric, Davis gave his team a chance to win 43% of the time, while Garza had a win probability of 50%! To help explain this, we can look at the standard deviations of each of these statistics. Davis had a standard deviation of 2.56. Meaning that most of the time he had an FIP between 2.56 and 7.69. An FIP of 2.56 gives a win expectancy of about 85% as we noted above. The win expectancy on 7.69 is around 4%. With me so far? At the same time, Garza had an average of 5.24 with a standard deviation of 3.91, much more variation from start-to-start. These yield a first standard deviation of 1.33 to 9.15. As we already noted, an FIP of 7.69 is only going to lead to a win about 4% of the time, well, an FIP of 9.15 isn’t all that much worse at a win roughly 1% of the time. On the other hand, an FIP of 1.33 is going to lead to a win around 98% to 99% of the time. That’s a guaranteed victory where the 85% still has a decent chance of being a hard-luck loser.
It would appear from this example, that it’s better to be a boom or bust pitcher, like Garza, than a steady guy, because the downsides are roughly about the same while the upsides are greater when you get them. We can look at this in another way. We can look at the likelihood of each starter giving us a 90% chance of winning a given game:
Pitcher//% of starts over 90% PROBw//% of starts under 10% PROBw
Niemann and Davis are good pitchers, but they are almost never great, and are more likely to have an awful start than our other pitchers (I’ve left off Sonny and Helly due to 8 combined starts skewing the data). We can do this same thing by wOBA to give some balance to guys that maybe don’t have great K/BB/HR rates, but are good at not giving up many extra base hits or base runners at all. Let’s take the same look at this:
Pitcher//% PROBw > 90%// % PROBw < 10%
Niemann//7%//17% (Includes starts shortened due to injury)
Shields//9%//18% (Including 5 starts with a win expectancy of 0%, ouch)
Take from that what you will as each stat looks at a different aspect of how to grade a pitcher. Garza had 10 different starts where we wouldn’t be expected to get a win more than 10% of the time. That just blows my mind that a guy with that good of stuff, who has proven on several occasions that he can carry a team on his back, is that hittable. On the other side, when he’s on, he’s one of our best pitchers matching Price’s likelihood of throwing a real gem.
Just for SnG, let’s take a look at this stuff through the lens of 80/20 instead of 90/10:
Pitcher//% of PROBw > 80%//% PROBw <20%
Pitcher//% of PROBw > 80%//% PROBw <20%
Take from this what you will, but I think it goes to show you that even guys that are generally regarded as good to great pitchers are going to have their fair share of brutal games. If you would like to download this workbook, click this LINK. If you click on “Open” or “Download” it should save to your desktop or whatever your downloads to depending on your OS. Please let me know any questions as this is kind of nerdy stuff. There are no bad questions, either e-mail me at sabresrule79080 at yahoo dot com or get at me on twitter @sandykazmir, also I encourage you to leave a comment as we can go back and forth for the benefit of the audience instead of a one on one exchange.