Updated Playoff Odds
I’ll update this post with the current odds daily.
Through games of 9/3:
Warning: If you don’t like numbers, you won’t find much in this post (or the next) worth reading.
This is an update to my playoff odds post from earlier this week. I’ve corrected a few misconceptions regarding tie-breakers and makeup games and made my model a bit more robust.
The biggest error I had was regarding makeup games for early-season rainouts. For some reason, I was under the impression that rained-out games would be replayed at the end of the season if they affected the pennant race. That is not the case. The ValleyCats’ rained-out games with Jamestown from July and with Aberdeen today will not be played, nor will Connecticut’s game with Staten Island today. Vermont has missed three-plus games so far, but can make up the Brooklyn and Hudson Valley games, plus yesterday’s suspended extra-inning Brooklyn game, because they play those teams again this year. Its rained-out game against Batavia, however, will not be replayed.
The other place I errored was with tie-breakers. I assumed that ties would be broken with a head-to-head game, but that is not the case. Instead, the tiebreakers go as follow: winning percentage, then divisional record, then run differential. It is rare that a tie will go even that far – I predict only a 0.5% chance that run differential comes into play.
The last update is an improvement to my model: an adjustment for home-field advantage. Home teams this year are 226-191 (.542), making home-field advantage a fairly significant factor. Thus, I gave teams playing in their home park a 4% boost in each game*. Note that I said “playing in their home park” – some of the makeup games (i.e., Vermont vs Brooklyn) will not be played at the same place they were initially scheduled; therefore, Vermont will be playing as the “home” team in Brooklyn. In such a case, Brooklyn would get the home-field advantage bump – research has shown that it is playing in a familiar park, not having the last at-bat, that provides the home team with an advantage.
*VERY technical note: In an ideal world, this home-field boost would not be linear – it has a smaller effect with a more lopsided matchup. To illustrate with an extreme example, if a team had a 4% chance of winning on a neutral field, we would not expect it to have a 0% chance of winning on the road. But I couldn’t figure out an easy way to make this effect non-linear, and I expect that all realistic matchups – certainly the ones that I am predicting here – are evenly-matched enough that it doesn’t make much of a difference.
(If anybody is following my work closely, I gave Vermont a 35% chance of winning its suspended game against Brooklyn, down 8-7 with two on and one out in extras. I got that number from this win probability table.)
The ValleyCats continue to improve their playoff hopes. Keep in mind that these numbers still don’t take into account momentum – June games count as much as August games do. If you think recent results should carry more weight, you should give the ValleyCats a somewhat better chance than listed here. (For what it’s worth, I do think recent results should count more, given how much rosters and players change in this league, but I haven’t come up with a good way to separate recent performance from schedule effects.)