Interesting stuff, Mike. I'd be interested in seeing someone come up with a Leverage Index that places every game on a sliding scale based on potential pennant impact. I think that some of the flack that Rodriguez gets is stuff like the Mayrod label. Also, I think that Justin Morneau won the AL MVP last year because it was perceived that he hit in the games that counted.

I'd be interested in seeing someone come up with a Leverage Index that places every game on a sliding scale based on potential pennant impact.

A good way to do that would be to look at the baseball prospectus postseason odds. To calculate a LI-like statistic for a game between, say, the Phillies and the Qankees(*) you'd compare the probability of the Phillies making the playoffs at the beginning of the day, the probability of the Phillies making the playoffs if they win the game, and the probability of the Phillies making the playoffs if the Qankees win the game.

However, this has a problem: it's possible that a game could be important to one of the teams involved (because they're a contending team but just barely -- the actual Phillies in the last couple seasons are a good example) but not to the other. You don't have this problem with LI, and it's not immediately clear how to fix it; perhaps you've have to calculate separate statistics for both teams? Also, the postseason odds report seems hard to calculate; I'm wondering if there's a quicker way to calculate approximate playoff odds than actually simulating the entire schedule.

(*) Yes, the Qankees. I generally call the teams in my examples the Phillies and the Qankees, because I'm a Phils fan, and q is the letter after p so clearly the name of the other team has to start with a Q; Qankees is more fun to say than Qets or Qaves (Qraves?), and I went to college in Boston so I kind of got the hating-the-Yankees thing drilled into me there.

I think that some of the flack that Rodriguez gets is stuff like the Mayrod label. Also, I think that Justin Morneau won the AL MVP last year because it was perceived that he hit in the games that counted.
Think Chipper Jones in 1999. After the ASG, hit 328/464/693/1.157, y'know.

I have a dumb question and am too lazy to go through the original work...
What is the 'favorable' state assumed in LI?. A basehit with each runner advancing one base? Some variation where a runner scores from second with two outs?

If just one base, might LI overvalue a hitter who is in fact a bit more likely to produce extra bases with the leveraged AB? Just offhand, note that the unadjusted SLG of the overachievers above is 27 pts higher than the underachievers.

Agreed, though, the operative word is 'fun'.

What is the 'favorable' state assumed in LI?. A basehit with each runner advancing one base? Some variation where a runner scores from second with two outs?

***

No, the beauty of LI is that it takes into account ALL possibilities of what could happen in determining the leverage of the situation, instead of seeing what would happen given one particular event, like Woolner's Leverage or Drinen's "P". I would suggest reading all of Tango's articles on LI:

http://www.hardballtimes.com/main/article/crucial-situations/
http://www.hardballtimes.com/main/article/crucial-situations-part-2/
http://www.hardballtimes.com/main/article/crucial-situations-part-three/

DSG-
Thanks. I remember reading the THT stuff originally but probably just forgot how amazing the work product really was.

So a player with higher ISOp, say, in those situations will get more credit but with good reason - he's favorably affecting outcomes better than the weighted average of outcomes, not better than some fixed, deterministic outcome.

No, the beauty of LI is that it takes into account ALL possibilities of what could happen in determining the leverage of the situation

There are multiple ways to calculate it, as Tango's series notes, but the definition is:

Leverage is the swing in the possible change in win probability.

so you are looking at endpoints - e.g. with bases loaded and no outs, the swing in WP goes from hitting a grand slam to hitting into a triple play.

So a player with higher ISOp, say, in those situations will get more credit but with good reason - he's favorably affecting outcomes better than the weighted average of outcomes, not better than some fixed, deterministic outcome.

Well, the player will get more credit for being a clutch performer in the minds of the fans. I don't know that the player actually deserves the level of credit that weighting performance by LI gives him - it can be argued that players who perform well in lower-leverage situations early in the game actually help their teams more by reducing the frequency of higher-leverage situations where teams can see wild swings in their ability to win or lose later in the game. I do think there's some value in looking at the shape of performance based on leverage, but I look at this as more of an adjustment to the base evaluation rather than as an overall evaluation system.

-- MWE

But the 'endpoints' are just two of the (more unlikely) plus/minus outcomes being weighted, yes? That is, with bases loaded, no out, an LI of 3.0+ may have your extreme scenario(s) in there but there's also the double that clears the bases or the double play that scores one run with the 'next state' at two outs.

By 'more credit' I meant that, given that it is a high LI spot, to the extent he has a better weighted distribution of outcomes than the distribution used to get LI (e.g. always hits a double when he does hit), then that is legit credit.

I sort of thought the opposite - this IS the way to see if a guy really does contribute five nickels early in a game for every quarter another player might.

The bottom 5 aren’t particularly well-known as “chokers” - with the possible exception of Tejada

Well, Miggy won an MVP on the basis of a handful of high-leverage successes, so I doubt he has a rep as a choker; I've never heard him assailed as such.

Miggy won an MVP on the basis of a handful of high-leverage successes, so I doubt he has a rep as a choker; I've never heard him assailed as such.

Since he moved to Baltimore, there have been a handful of rumblings about his lack of "clutch" performance, and of course his postseason baserunning booboo with the A's is still remembered not-so-fondly. That's why I said "possibly".

I was surprised by how well the numbers, particularly at the top of the lists, supported the popular perceptions of players.

-- MWE

I know you only ran 2003-2006, Mike (and thank you for all the work and interesting thought process), but it would be interesting to see how a great season (A-Rod) and a personally average season (Ortiz) might impact the rankings. I doubt one season will change the perceptions of players established over a 4+ year span, but given A-Rod's terrific 9th inning success thus far, I'm wondering how far up the rankings he would move. Similarly, it appears that Jeter has been decidedly "unclutch" this season (from a purely observational perspetive), and Ortiz has been battling injury/slip in performance, so high might drop out of the top 5.

I guess the shorthand questions is: how volatile are these rankings, given the expected yearly variation + the small sample sizes?

I guess the shorthand questions is: how volatile are these rankings, given the expected yearly variation + the small sample sizes?

Hard to say. I would expect them to be pretty volatile, in as much as players get only about 60-70 high-leverage PAs (I call any PA with LI of 2.0 and above "high-leverage") per season. ARod, for example, had 262 PAs in high-leverage situations from 2003-2006. Ortiz had 271, Jeter just 226 (which is pretty low for the group of good hitters, but not unusual for a top-of-the-order guy).

-- MWE

Thanks for the reply, Mike. Do you think you might recalculate these after 2007 to get an idea of volatility? I know you've done a lot of work on this already, so "no" is a good answer as well...just curious if it will be something to look out for.

On a side note, did the proportion of 3.0:2.0 LI situations have any predictive value? Given the increased weight and the (presumably) lower frequency and small sample size, I could see someone's (relatively) poor performance in LI 3.0 situations skewing their overall pretty considerably, if they faced an inordinately high # of PAs in those situations. That said, the likelihood of that affecting the rankings of more than 2-3 guys is, I would assume, pretty low.