No Update Today; Changes Coming...
As stated previously, I'm using FanGraphs for tracking win probability added. For whatever reason, they have updated their daily tracking of yesterday's games, but not incorporated them into the seasonal stats, which is what I use. Since it's about 7pm here on the East Coast, and Friday to boot, I'm calling off today's update.
But when the next update comes, there will be more that's new than just the current stats. For the list of changes, hit the jump.
Tango gave some unexpected but much appreciated input yesterday, which I will be implementing in the next update.
OPS Wins Per PA are now calculated by the formula .025 * ( 1.7*OBP+SLG-1 ) for more precision.
Two new columns appear on the stat page: Leverage Index (LI) and Leverage-adjusted OPS Wins (LevOPS Wins).
Leverage Index is explained here and has been at the crux of clutchiness already. The reason a player's seasonal leverage is being included is to better elucidate LevOPS Wins
LevOPS Wins is simply OPS Wins multiplied by LI.
These two make their appearance because clutchiness is now calculated as WPA Wins - LevOPS Wins. The whole gist of the comparison has been that stats like OBP and SLG are examinations of a player's production over all at-bats equally, while WPA uses leverage to judge value by production in certain, more important (higher leverage) at-bats. But while OPS Wins looks at every at-bat as equal, it also sees them in a vacuum. If the intent of clutchiness is to determine extra value added over what we expect, as Tango points out, it makes more sense to incorporate a player's seasonal leverage total with his OBP and SLG totals to better determine what we value we should have expected him to add. That is to say, if a player has, over the course of the season, come to bat in low leverage situations, we should lower our expectations for how many wins above average he could contribute. Obviously the converse applies as well.
All of these changes are being made in order to better evaluate a player's clutchiness, which is of course the intent of this exercise. It's a bit more work, but I'm glad to do it; anything in the name of clutchiness.Read More.
¶ 6:55 PM0 commentslinks to this post
What Is Clutchiness, The Site?
So clutchiness exists. It may or may not tell us anything about the future, but it does describe the past. And like all good (or at least, interesting) statistics, it deserves to be tracked. That's where Clutchiness comes in. I was so taken with clutchiness when I first saw it, I wanted to check it for other players. What was Papi doing? Who completely lacked clutchiness? Who was the clutchiest of my beloved Marlins? I had to know. And since I figure I'm not particularly out of the ordinary, it stood to reason that others would want to know. It had to be tracked. I had to track it.
And thus I set out to put together what you see here. There aren't any Excel-able MLB stat pages available to the general public, or that I'm aware of anyway. So the arduous process of simply getting everything in order just to be tracked was begun. This, obviously, leads to a very important caveat: the numbers you see are a combination of MLB's official stats (good), Fangraph's WPA (good), by-hand data entry (not so good) and Excel calculations and storage (also not fantastic). This is not to say you should assume this is all wrong, but that the occasional hiccup in the process may occur. If you spot one, please let me know.
The other main point is that clutchiness depends on WPA, and WPA depends on FanGraphs, and FanGraphs updates many hours after games end. That, combined with the by-hand nature of the project, means that Clutchiness will also update many hours after games end, and probably closer to the next day's games beginning. I intend to update the stats daily, though on occasion I have been known to live an actual life, so there may be days when I am unable to update. Every time the stats are updated, I will make a post noting as such, and the link on the right will state through what date the stats encompass.
Some notes about what the stat pages look like. First of all, I made the completely aribtrary decision to track players with 10 or more plate attempts. As a fan of a team in the NL, I think it's worth knowing which of your pitchers has been clutchy. Not to mention that many, if not most, pinch hitting situations are high leverage; which of your bench guys has gotten it done when called upon? At the same time, this means you'll see numbers fluctuate wildly on nearly a game-to-game basis. Just in the few days I've been putting the database together, I've seen guys go from hero to zero and back again. One swing of the bat can give you a third of a win; it can take it away just as quickly. As with everything, the more data that comes in, the clearer the picture will be.
You will notice some formatting of names in the stats. I did this for some very basic at-a-glance presentation, rather than looking at just numbers. Names in green are players with a positive WPA, regardless of clutchiness. More than anything, this is to make sure people keep in mind that clutchiness is simply a measure of actual vs. perceived value, not a measure of value itself. Players can, and many do, have high clutchiness but negative value; it's simply that they're not as negative as we would expect. And by that same token, many exceptional players have negative clutchiness for simply being good in high leverage situations, rather than their usual great. I hope that by keeping positive WPA easily seen, we won't lose sight of what's really important. Bolded names, whether green or red, are players whose value in one stat is positive and negative in the other. Bolded green means the player has a positive WPA despite an expected negative value. Players bolded red have contributed negatively to their team's win probability, despite rate stats that say they should be helping. Again, this is just a way of showing multiple pieces of information at once. While the players are sorted by clutchiness, the performance of these bolded players is seemingly especially important, regardless of whether their clutchiness itself is especially high or low.
I am already thinking of additional ways of presenting the data, namely a league top and bottom ten. At this point, I can't say whether it would be a daily update or perhaps as part of a weekly round-up. We shall see what the future holds. But with Derek Jeter having an exceptionally clutchy season when I began this project, and David Ortiz contributing back-to-back walk-off hits between Clutchiness' inspiration and this, its initial execution, clutchiness seems to be all over the place. It is my hope that this can become a resource for any and all fans looking to track it.Read More.
¶ 4:18 AM0 commentslinks to this post
What Is Clutchiness, The Stat?
To say there is debate amongst the stat community as to whether clutch hitting exists as a repeatable skill is an understatement. For just the tip of the iceberg, peruse any or all of these works on the subject. Personally, I'm clutch agnostic. That so many smart people have looked at the same issue and come to so many different conclusions leads me to believe the answer, whatever it may be, is still lurking.
That said, one of the most respected of those smart persons is Tom Tango, who co-authored the highly recommended The Book. An offshoot of The Book has been what I call The Blog. It was a post on The Blog that was the inspiration for what you see here.
What struck me in the post was not so much its conclusion (Captain Intangibles is having a clutch season? Shockhorror.) but the almost absurdly simple methodology: by comparing a player's actual value added, in terms of the leverage-adjusted win probability added, to his expected value, in terms of straight OBP and SLG, we can see how much value he added to (or cost) his team by over- or under-performing his seasonal averages in higher leverage situations.
Tango's Jeter example may help illuminate the point better. Back on June 21, when the entry was posted, Jeter's OBP, SLG, and PA described a batter who should have been worth a bit more than 1.6 wins (using Tango's equation where .03 * ( 1.7*OBP+SLG-1 ) = wins above average per plate attempt [see the update below]). But according to win probability, he had added 3.2 wins. In short, if he had performed exactly along his seasonal (if three months can be considered seasonal) averages in every plate appearance, he would have added 1.6 wins over average; his nearly double WPA makes it clear that he was outperforming these numbers in his most important at-bats.
As I said, I'm clutch agnostic. I don't know if clutch hitting as a repeatable, sustainable skill exists or not. But it occurs to me that if it does, the Jeter example probably looks a lot like what clutch would be. At the very least, it's demonstrable proof of already-occurred performance in what we think of as clutch situations. It may not truly be clutch... but it definitely is clutchiness.
Clutch hitting may or may not exist. Clutchiness most certainly does. By comparing a player's value in terms of win probability to his projected value based on OBP and SLG, we see how much he has over- or under-performed expectations due to performance in higher leverage situations. If clutch exists, it might just look like this.