Sportswritersand fans love players who are said to do the "little things." While "statheads" grouse about this a fair amount (and usually rightfully so), it must be conceded that at least on the immediate level of playing the game, things like "moving a guy over," a sacrifice fly by a player on a team who is down one, or stretching out a double into a triple can make a difference, especially in a certain game state. The stereotypical sabermetric "nerd" (insert jokes re: mom's basement, never kissed a girl, etc.) dismiss such events as being non-repeatable or statistically insignificant. But even the guy who gave sabermetrics its name, Bill James, uses such an instance of "game awareness" in the New Historical Abstract entry on one of his favorite players, Amos Otis:

One time when George Brett was had been hitting about .600 for a week he came up with runners on first and second, two out in the ninth, game tied. Earl Weaver intentionally walked Brett, bring A. O. to the plate with the based loaded. Otis was magnificent. He took two close pitches to make it 2-0, worked the count full, then started fouling off pitches. Knowing that a walk would win the game, he just spoiled pitches until the Oriole pitcher (Tim Stoddard) missed with ball four. The next morning the Kansas City Star report led off with the statement that "Amos Otis won the game for the Kansas City Royals last night by doing nothing more than the 31,184 paying customers."*

Is there a way to measure such contributions to the game? This would be a method that would indicates that a player went beyond being a good (or bad) hitter in the ordinary sense, yet something different from "clutch" (although such a skill could contribute to "clutchiness") -- something apart from the situation's leverage that indicates a general game awareness? Is there a repeatable skill involved in doing these "little things?"

This is a stab at quantifying such efforts.

* I verified that this actually happened, by the way, on August 12, 1980, in Kansas City. I love that James is still an unrepentant fan to this day. Click here and do a search to read James calling Otis "perhaps the greatest player who ever lived.' Awesome.

Setting the Stage: wRAA and WPA/LI

In this post, I'll be exploring some rather uncharted ground (for myself, anyway), so bear with me, as this post may very well raise more questions than it answers... I do think that there is something here, but I make no pretensions to have all (or perhaps any) of the answers. At the very least, I've come up with another cheesy award.

In a recent Q & A thread at Fangraphs, the-source-of-so-much-internet-sabermetrics, Tom Tango, answered questions from all comers about the various unusual and often confusing stats contained at the site that has quickly become the one-stop shop for "stats geeks" everywhere (if only some teams knew about it...). The thread has a lot of good stuff in it, from Tango and others. For my purposes here, I just want to focus on two of Fangraphs' statistics: wRAA and WPA/LI.

wRAA is wOBA Runs Above Average. wOBA (weighted On-Base Average), as you may or may not know by now, is the on-base percentage-scaled rate-stat invented by Tom Tango to give a stat that "looks like" a traditional rate stat but that, unlike BA, OBP, SLG, or even OPS, summarizes, in one number, a player's offensive contribution per plate appearance in a way that accurately reflects that player's contribution. GIven its origins, it is also handy in that it can be converted to a players runs above or below average over a season -- which is what wRAA is at Fangraphs. (Stat Corner also carries wOBA, along with park-adjusted versions, and includes conversions to hitting wins above replacement [hWAR]). Fangraphs' version of wOBA uses linear weights that are specific to the year in question.There is much to read about wOBA all around the web at links provided here or by google searching, so I won't get into details. What is important here is to note that one gets a player's wRAA (runs created above what a league-average player would produce given an equal number of plate appearances) by comparing his wOBA with the league-average, then multiplying it by the number of plate appaerances. Like batting average, on-base percentage, and so on, this is a context-free measure of of the player's offensive performance -- it completely ignores the games scores, win expectancy before and after the player's plate appearance, how many baserunners there are, how many outs there are, and so on.

But while wOBA is easy enough to get a grasp of, the other stats that are more commonly associated (at least in my mind) with Fangraphs, such as WPA/LI, are a bit more difficult to grasp for the regular baseball fan. In a sense, they are even more difficult for the fan who knows the basics of sabermetrics, since we are (rightly, for the most part) told that "clutch" is not a repeatable skill, etc. So what is with all these context-sensitive numbers all the sudden?

Once again I recommend reading around the usual sites (some links are provided below). Here, we are using WPA/LI to get to another point ( or perhaps, as we'll see, using another point to come back to WPA/LI). The short-long version is that it is derived from WPA -- Win Probability Added. Again, you can find more about WPA in various places, such as this excellent THT article by Dave Studeman. In short, it takes account of each team's chances of winning before and after a plate appearance, and assigns a +/- score to the hitter and batter respectively after the play based on how that win probability changed (This is the basis, by the way, for the cool in-game Win Probability graphs that give FanGraphs its name.).

As one might expect, from the win probability perspective, some plate appearances are more important than others. To illustrate this, let's look briefly at #1 shift in WPA this season, for example, happened when David DeJesus blasted a two-run, walk-off homer off of Brandon Morrow that provided at 90% shift in win expectancy. This was a "clutch bomb," of course. But notice that much of the immediate situation was not in DeJesus' control. He didn't control the events leading up to a runner being on base. There were many factors more important than his earlier performance that led to him being in that base/out/game state. Other people were involved in the scoring and lack of scoring earlier in the game.

When we say that from WPA's point-of-view some plate appearance are more important than others, it means that because of the various elements of the situation, that particular PA counted at more than others because of the leverage the situation had. So, while DeJesus' skills certainly played a big part in the play, his overall WPA score doesn't merely reflect how well he "executed"  per PA, because some count more than others due to the context in which they occur. In other words, a player's cumulative WPA doesn't tell us how good he was over the season of adjusting to the game/base/out state. The same is true of WPA with regard to the Amos Otis situation above -- how much of it reflects his skills in recognizing the game state itself is all mixed in with the leverage of the particular situation.

This issue gets at the heart of "situational hitting," the reason often given for some players being called 'professional hitters'  (click the link to go to a Pizza Cutter post that seems to anticipate much of the thinking being WPA/LI) who allegedly have special skills with moving runners over, making "productive outs," etc. that go beyond batting average, OBP, SLG, etc. In other words, they "play to the situation." WPA sort of gets at this, but as we have seen, it doesn't value all PAs equally. Thus, we have WPA/LI -- that is, Win Probability Added divided by Leverage Index (LI). This simply sets the LI for each PA at 1, so we can see how much the player advanced his team's cause in context per PA, without the inflation provided by relative game score. That is what WPA/LI measures.

[While this sounds like "clutch," and involves some of the same notions, it is distinct from it. The "clutch" stat at FanGraphs is calculated by subtracting WPA/LI from WPA in order to see how much win probability a player added in high leverage situations in particular, whereas WPA/LI leaves that issue aside, to measure the players response to game/out/base state.]

The thought, then, is that if some players really are "professional hitters," perhaps WPA/LI can get at something missed by OPS, wOBA, etc. to reveal how a player can increase his production by responding to the game context, i.e., to measure a player's situational hitting. Note, you can also measure Run Expectancy (hence RE24 and REW at FanGraphs), but the goal of the game is to win, as Herm Edwards might say. In the A. O. situation described above (tie game, bases loaded in the bottom of the ninth), for example a walk is exactly the same in terms of game impact as a grand slam, and that is what WPA/LI reflects.

Measuring the Little Things

Going back to our original project -- measuring the little things -- we can see that WPA/LI definitely includes small things not included even in wOBA/wRAA. Moving a runner over at the cost of an out does advance a team's cause (in some situations), even if wOBA just sees it as another out. On the negative side of the ledger, wOBA/wRAA also doesn't recognize that grounding into a double play makes two outs -- it's just another failed plaet apperance like any other. WPA/LI recognizes proportionally how much that hurts a team by making two outs and taking a baserunner off of the base.

wRAA is better at measuring a player's raw offensive value, since it eliminates the game-state, base-state, and out state from the equation, things that the player has no control over (in the immediate sense). WPA/LI on the other hand, is very context-sensitive, only equalizing the importance of each PA.

In the thread that originally sparked this discussion, a user suggested (and Tango approved) that wRAA could be subtracted from WPA/LI to generate a stat called "Little Things"** to how much of a player's contribution to winning was generated by things not measured by wOBA, as well as his awareness of the game/out/base stat -- while wOBA would see A.O.'s walk as just another walk, WPA/LI would see it, in the context in which it would occur, as just as valuable as as a hit-by-pitch, double, or grand slam.

** It's worth pointing out that In his discussion of Craig Biggio in the New Historical Abstract, Bill James discussed compiling a stat called "Little Things" that would include things such as hit-by-pitch, not grounding into double plays, etc. I don't know if he ever went anywhere with it. The following stat is good in that (if viable) it measures just how many wins a player contributes by way of Little Things.

I don't know if FanGraphs is going to implement this, and I thought it was an interesting idea... So I exported this years qualified hitters wOBA and Win Probability Stats, did some joins and... here we are.

First wRAA needs to be converted to a wins scale, since WPA/LI is about win probability. If we were dealing simply with run expectancy and thus not dealing with the game state, then in the A. O. situation, a walk would certainly not be the same as a grand slam. While the runs/wins ratio varies year-to-year, here I've simply used "10" since both the AL and NL were pretty close to that. And I've "mushed" the AL and NL together to reduce clutter. The "Little Things" number is scaled to wins, but keep in mind that it is a bit rough at this point, since FanGraphs wRAA is not park-adjusted and WPA/LI is. And also keep in mind that players don't have control over what situations they find themselves in, only how they respond to it. Just because someone has a negative score doesn't mean they "suck." You can always look at the WPA/LI Leaderboard on Fangraphs for that overall ranking if WPA/LI sounds like your offensive stat -- here, we're just trying to isolate how much of a players contribution is done by things not measured by context-neutral stats.

And what are those things? Well, from what I know, the version of wOBA at FanGraphs includes the following: singles, doubles, triples, homers, walks, stealing, and caught stealing (the original version of wOBA includes reached base on error, but FanGraphs data sources don't for some reason).

What's left, both positively and negatively for "Little Things?": Reaching base on error, sacrifice hits, sacrifice flies, grouding into double plays (negative, of course), things like that, the value of hitting with runners on base as opposed to the context-free value of the same hit, and probably a bunch of other stuff I'm not thinking about at the moment.

Without any further ado, then, here is your Scrollable Little Things MLB Leaderboard for 2008 (here is the full EditGrid Spreadsheet -- could this be the untimely end of "fun with Google Spreadsheets?" An anxious nation awaits the decision...):

So there it is.. Jack Hannahan is YOUR MLB God of Little Things 2008. He has a big gap between his slightly below average WPA/LI and wWAA (woba Wins Above Average). What's the deal? Anybody out there watch a lot of As games this past year? He doesn't seem to have an especially large number of sacrifices or reaching base on errors, nor an especially low rate of grounding into double plays. His percentage of other runners batted in, either. What am I missing? In any case, he seems to have done the trick this year. His wOBA didn't quite do him justice (and he's a good defensive 3B, too).

Jack Cust was a surprising name to see second. Wait, I thought the As were just a bunch of statheads who didn't care about fundamentals? But seriously, what am I doing wrong? I love Cust as much as the next internet nerd, but he isn't exactly one of the first (hundred) names I think of when I think of "situational hitters." Good for him. As you can see, "Little Things" (if I've done it correctly) indicates that both Cust and Hannahan contributed more than a win each to their team by doing "little things" not taken into account by the most advanced context-neutral stats out there.

I have to say that, as someone who followed the Royals pretty closely in 2008,  seeing Jose Guillen with a "Little Things" number of 0.71 wins was pretty surprising, considering that he grounded into 23 double plays this past season (18.8%!). He also didn't have a high number of sacrifices. However, he was decent at getting runners in. He did hit cleanup pretty much all year long (believe me, he did, no matter hard far below replacement level he fell, he was there -- hey, he made it back by the end of the year!), which makes me wonder if players in certain positions (say, hitting behind the only guys on the team who got on base at a decent rate [David DeJesus and Alex Gordon] for most of the year) doesn't put the player in more contexts where "situational hitting" comes into play.

Ichiro Suzuki was one of the players that for whom was WPA/LI  was supposed to really point up, but he ends up with a -0.28 in 2008.

The internet versus Derek Jeter thing was old at least two years ago, but for the record, he scored a -0.58 (David Eckstein didn't qualify this year, but had a -0.32).

So, what did we find out? Well, Jack Hannahan's offense didn't hurt the As as much as his context-free numbers might suggest, for one thing. And also that we (I?) need to look into what all goes into WPA/LI that doesn't go into wOBA.

The burning question, of course, is to what extent "Little Things" (and WPA/LI in general) reflects a repeatable skill (or set of skills) or not. That is something that would be of real value, but this post is long enough as it is, and there are certain things to work out in measurement (and in how to organize and peruse the data).

In lieu of the knowledge and means (and reader patience) to conduct such a study at the moment, I'll go back to Ichiro!, for whom Tango says WPA/LI (and thus, by extension, "Little Things") is "ideal" because of Ichiro's abilitity to "rebalance his game to take advantage of the new balance required for a game state." I assume this means that this ability wouldn't show up in wOBA, but only in WPA/LI. WPA/LI-wWAA shoud give us a positive number (with the qualifications that some further league/park/other adjustments might need to be made to make the two commensurate).

While I've been using Fangraphs numbers for convenience of large blocks of data, for the sake of getting runs/wins conversion of wRAA to wWAA right (Fangraphs doesn't currenlty publish wWAA), in the following example I'll using my own database, which uses pretty  much the same wOBA SQL as Fangraph, although some of my numbers are a bit different, probably due to Baseball Databank's PA issues (having to do with ROE, if you must know), I believe.

Hmmm... well, Ichiro never seems to go very far below zero (perhaps in 2007). His WPA/LI is never negative, but it's not clear that he's doing much beyond what wOBA records, but he certainly isn't in Hannahan/Cust 2008 territory. Perhaps I'm missing something?

What would be needed for further research? Assuming that things are calibrated correctly (making sure that the wWAA and WPA/LI runs-wins conversion are on the same scale, the park factors are the same, the baselines are appropriately adjusted), we'd need to see for sure if "0" or some other number was the "Little Things" neutral spot -- that is, where wWAA expressed everything for a particuar player that WPA/LI did. We'd then need to figure out the size of the standard deviation (that's getting a bit beyond me...), and do studies over a series of years for large numbers of players, somehow neutralizing for game state opportunities (if it needs to be done beyond what WPA/LI already tries to do by setting the leverage index of all player PAs at 1.)

Yes, that's all... three dots.... I know that's an inconclusive ending, but for now, Jack  Hannahan, God of Little Things 2008, we salute you.