Couple of New Things
I just wanted to point out a couple of new things on FanGraphs:
- In the leaderboard section you can now filter the pitch type by NL/AL and Starters/Relievers. This can be done for the new plate discipline stats for batters and the pitch type data for pitchers.
- The game graphs now have a sidebar with all of the games for that day. Hopefully this will make for easier navigation. There are also two new items in the sidebar: aLI and aWE. These stand for Average Leverage Index and Average Win Expectancy over the course of the entire game. The higher the aLI, the more “exciting” the game was, and the closer the aWE is to .5, the “closer” the game was (throughout the entire game).
- The scoreboard, with all the game graphs on it for one day should load much much faster. Hurray for caching!
That’s all I can think of that’s really new, except for a few changes on the homepage, but you should expect to see a few more seasons of Win Probability rolled out later this week.
The sidebar is a fantastic addition for those of us that watch/listen to games from the past. It’ll make my workday much more interesting being able to easily pick an “exciting” game from those available.
Thanks!
Hmmm… average WE: you seem to have it based on the average WE for the home team. But, if the game through 7 innings was .600 to .400 to .600 to .400, while another game was .550/.450 through 7 innings, would they really be the same to that point? I’m thinking that you want to do: the average of abs(WE-.500), and then take that total for the game, and add it to .500. (Similarly, take the WE of whichever team is highest at that point, and average it.)
I’m pretty sure that’s better, but not positive. Can you come up with some big differences in the current approach and this proposed approach, so we can see the impact?
Tango’s idea would work best — average WE of the team with the higher WE. But … in a way average WE measures the same thing as average LI — the closeness and tension of the game, and I think average LI does a better job of this. If you head into the bottom of the 8th inning of a tie game, and the home team puts runners on 2nd and 3rd with none out but don’t score, there’s a lot of tension which is captured by average LI. However … the WE of the home team goes up to about .850, and so by the average WE metric, the game is actually less interesting than it would’ve been if the bottom of the 8th went 3 up, 3 down.
Might be best to represent “closeness” with average LI, and then the second stat could measure a different dimension of the game, such as “wildness” (ie 15-14 wins with lots of lead changes), using sum of win advancements or standard deviation of WE.
Also, I hate to treat you like a public utility, but … any chance of seeing a list of the highest average LI games over a season (rather than just a day)?
I ran the numbers and the results are pretty similar, but as the games get “closer” the results start to disagree more. They seem equally good at determining which games aren’t close at all.
Here’s a couple examples of games that completely disagree:
Brewers @ Astros - avg(WEH): 0.50, abs(WE): .78
Dodgers @ Rockies - avg(WEH): .47, abs(WE): .79
Here’s one where they agree:
Blue Jays @ Red Sox - avg(WEH): .55, abs(WE): .54
So the games that are close all along they tend to agree are close but the ones with huge swings they’re going to tend to disagree.
Dackle brings up a great point.
One thought is that the WE part that David is presenting is only supposed to represent “closeness” and not “excitement”. So, 3-up 3-down would keep things closer than putting men on base with 0 outs.
Another thought is that perhaps we should only look at the (highest) WE at the start of each half inning. This way, you get something “real” (the actual score of the game, rather than some theoretical expectation of runs scoring).
This might better capture the “closeness” of the game.
The leverage index does the best in capturing “excitement” of the game.
Tom