While the Mets-Braves bonus doubleheader kept the computer idling for a bit yesterday, the numbers have been crunched, the models have been updated, and the initial run of the ZiPS Game-by-Game Postseason Odds is now live. These will no doubt change quite a bit as teams make their plans more explicit and our understanding of which players are available improves.
Since you shouldn’t be devoting any part of your brain to remembering Things Dan Szymborski does, let me do a quick review of how the playoff model differs from the generalized in-season model you see during the regular season.
Simply put, given the more compact nature of a three-, five-, or seven-game series and the additional scattering of rest days, the macro-level projections used to forecast a full 162-game season are less suitable when applied to the postseason. When looking a week into the future rather than half a year, we’re able to take a more ground-level view of the relative strength of playoff teams. We can make educated guesses as to who is starting each game, what the lineups will look like, who is healthy and who isn’t, and so on. The ZiPS game matchup tool has a built-in lineup estimator that projects the line of every pitcher and hitter against every other hitter and pitcher, so there is no need to look at a team’s generalized offensive strength. The postseason also comes with the benefit of being able to run the full, more robust model of ZiPS rather than the simpler model used during the regular season, a compromise necessitated by the fact that projecting the entire league takes more than a day.
Is the model perfect? Of course not. All models are wrong, but some are useful. The approach I take has benefits in some very specific situations where a team’s seasonal record isn’t a good predictor of performance. For instance, ZiPS saw the Nationals as being a far more dangerous opponent for the Dodgers in 2019 than conventional wisdom did by virtue of the team’s ability to concentrate a very large percentage of their innings in a few excellent pitchers. Still, there’s more work to be done. For example, I’d like to be able to more accurately predict bullpen workload consequences on a daily basis rather than going series to series.
If a particular game’s starters haven’t been officially confirmed, the pitcher listed is my most educated guess. Projections for the Division Series will populate after the Wild Card round has concluded, with the Championship Series projections coming after the Division Series wraps, and so on. For the moment, you’ll see those tabs grayed out. In addition to game-by-game projections, you’ll also find series odds, which will update as each series progresses.
Since FanGraphs offers a few different projection tools, here’s a quick summary of the features of the three — the FanGraphs Game Odds, the FanGraphs Playoff Odds, and the ZiPS Game-By-Game — that are most relevant for the postseason:
FanGraphs Postseason Projection Tools Comparison
| Feature | FG Game Odds | FG Playoff Odds | ZiPS Game-By-Game |
|---|---|---|---|
| Projects Game Results | YES | NO | YES |
| Projects Postseason Series Results | NO | YES | YES |
| Projects Future Postseason Rounds | NO | YES | NO |
| Lineup/Roster | Game Lineup | Projected Overall Playing Time |
Projected Game Lineup |
| Accounts for Starting Pitcher | YES | NO | YES |
| Base Projection System | Steamer/ZiPS | Steamer/ZiPS | ZiPS Only |
| Updates | Before/during game | Shortly after game ends | Before/Shortly after game ends |
All that’s left now is to enjoy the actual games, which are, of course, the best part of the postseason! As a reminder, all erroneous projections are due to either flaws in reality or villainy from Carson Cistulli.
