Leveraged Index (LI) - by the 24 base-out states (July 30, 2003)
Just as we can create an LI by the game state (inning/score/base/out), we can create one for the subset of the (base/out), by assuming the typical distribution of inning/score. Essentially, which base/out states have the most swing impact, when an event occurs. The average LI is set to 1.0. As you'd expect, bases loaded situations lead the way. All I am doing here is quantifying what we intuitively knew already. There's some interesting things that go on when you've got guys in scoring position and 2 outs.
Base Outs LI
Empty 0 0.90
Empty 1 0.65
Empty 2 0.41
...
1st 0 1.38
1st 1 1.11
1st 2 0.77
...
2nd 0 1.24
2nd 1 1.13
2nd 2 1.01
...
3rd 0 1.02
3rd 1 1.01
3rd 2 1.19
...
1st_2nd 0 1.72
1st_2nd 1 1.90
1st_2nd 2 1.62
...
1st_3rd 0 1.54
1st_3rd 1 1.61
1st_3rd 2 1.61
...
2nd_3rd 0 1.53
2nd_3rd 1 1.37
2nd_3rd 2 1.80
...
Loaded 0 2.22
Loaded 1 2.56
Loaded 2 2.96
In essence, figure out every player's OBA and SLG by the 24 base-out states. Multiply that figure by his LI x lgPA for that base out state. Add up your totals and divide by the sum of lgPA.
Compare that figure to his overall figure. Voila. Clutch Index.
(Again, using OBA and SLG is problematic, especially since they don't have the same denominator. And I would include SF in the SLG calculation. But like I said, to do it right, use LWTS by the 24-baseout states.)
Posted 9:23 p.m.,
July 30, 2003
(#2) -
David Smyth
You can combine similar situations, according to their LIs and their frequencies. I only have frequency data from 1987, so I combined that with the LIs above:
0 outs, .37
1 out, .34
2 outs, .29
Bases empty, .38
Runners on, .62
Scoring pos., .40
not Scr. pos., .60
So fewer outs in the inning is more "important" than more outs in the inning. And Runners on is more important than bases empty. Scoring position/not scr position is about the same as empty/runners on, so there is no big reason to divide runners on/empty into sub-categories.
Essentially, this means that you need on-base with bases empty and slugging with runners on. The runners on part is more impactful in real baseball (but not necessarily in any possible context). This matches up nicely with a weighting of OBA/SLG which I saw about 15 years ago in a Bill James Analyst newsletter article. The author used the shortcut RC formula of OBA*SLG*AB. He found that OBA should be weighted at .7*OBA w/bases empty +.3*OBA w/runners on base. And SLG at .85*SLG w/runners on + .15*SLG w/ bases empty.
Posted 9:25 p.m.,
July 30, 2003
(#3) -
David Smyth
Forget my comment about empty/on base vs. scr pos/not scr pos.
Posted 9:29 p.m.,
July 30, 2003
(#4) -
David Smyth
Another breakdown:
Bases empty, .38
Scor pos., .40
1st base only, .22
Posted 6:52 a.m.,
August 1, 2003
(#5) -
studes
(homepage)
Why do you sometimes see funky progressions within out states? For instance, runner on third with two out is higher leverage than one or zero outs?
Posted 12:37 p.m.,
August 1, 2003
(#6) -
tangotiger
Probably because with a runner on 3b and 0 outs, he's always assured of scoring (not much leverage there), while with 2 outs, he's got a 30% chance of scoring. So, there's a big swing possibility there.
If a reliever comes in to the 9th inning with a 3 run lead, he's almost assured of winning the game (not much leverage there), but coming in with a 1 run lead, there's a big swing possibility of winning or losing the game.
Leverage is about swing possibilities, and nothing else.