Welcome to the fifth edition of Banished to the Pen’s annual win prediction and projection extravaganza. In the name of making everyone accountable for predictions they usually didn’t want to make, as well as keeping tabs on the projection systems, I am back to conduct this exercise once again.
In this space, I usually remind people that this is not really how projections work, and using a single number to represent thousands of projection runs is merely a gross oversimplification of a very complex process, which projects a vast range of outcomes. You can bear that in mind while you read this, and again when I review the results at the end of the season, or you can just assume that PECOTA hates the Cubs, if that’s more entertaining.
Before the breakdown, here’s the annual reminder of what this feature is about for the new reader. I compile the win total predictions and projections from a number of sources, compare the range of different outcomes they provide, and return at the end of the season to find out which was the best set of win totals, or perhaps more accurately, the least wrong.
This is done using the mean absolute error (MAE) and root mean squared error (RMSE). MAE is the average difference between the predicted total and the actual, while RMSE is the square root of the average of the squares of all the differences. RMSE gives greater weight to large errors because they are squared, so if you think bigger misses should be punished more heavily, this is the more relevant number. Below is a recap of the contenders and the abbreviations I’ll use in the various tables. You can find out how they all fared last season in detail here.
PECOTA (PEC): The Baseball Prospectus projected win totals based on their in-house projection system. First place in both measures in 2018, by some margin, after finishing virtually level at the top with FanGraphs in 2017. The first two years weren’t so hot.
FanGraphs (FG): The FanGraphs Depth Charts projected totals, which are a combination of the Steamer and ZiPS projection systems, with an additional playing time adjustment applied by FanGraphs staff. Fell to middle of the pack last year after three consecutive years in the top two.
Davenport (Dav): Totals based on Clay Davenport’s projection system, with Clay’s own playing time estimates. Still yet to beat the other two projection systems.
FiveThirtyEight (538): Site projections from FiveThirtyEight.com, based on their Elo rating system. The first year they have been included.
Banished to the Pen writers (BttP): Predictions from each of our writers from our season preview series. Number one by RMSE in 2016, fell behind all but the EW guests in 2017 and finished last in 2018. The only way is up.
Effectively Wild guests (EW): Predictions from each of Effectively Wild‘s team preview podcast guests. Victorious in 2015, rebounded to third in both measures (and second by straight ranking) in 2018 after two years in the basement.
Composite (Comp): The average of the six projection/prediction sets above, with the BttP/EW sets adjusted down to add up to 2430 wins so they are not given extra weight. The most effective predictor in terms of overall rank in 2017 and the only set to get remotely close to PECOTA last year.
Public (Pub): The average of all responses to a preseason poll in which I asked people to predict win totals for every team. This has replaced the PECOTA over/under game from previous editions. Some people did note that they had also completed it after having some beers, so the abbreviation seems appropriate.
Level of Optimism
Before we delve into the specific predictions, I always take a look at how far the humans were from the actual possible total of 2430 wins (projection systems care not for your concepts of optimism or pessimism, and are always at or very close to 2430). This is appropriately titled ‘Level of Optimism’ because we are, collectively, always more optimistic than that.
The Effectively Wild guests went all-out to show how optimistic they were this year. At 2511 total wins, they were 81 over , beating their previous record of 2504. In the Yankees preview pod, Lindsey Adler explicitly acknowledged the pressure from fans to be optimistic that I discussed in last year’s review, a factor which clearly influenced her 105-win prediction. The EW guests were more negative than the average prediction for just seven teams.
While BttP’s writers weren’t quite so enthusiastic, they did still come in at 2497, an increase of 35 over last year’s total. Between them and the EW guests, our writers represented 23 of the 30 highest win totals, and just 8 of the lowest. FanGraphs, incidentally, did not have a single win total that ranked highest.
Opening this up to the public did create some less predictable results, however. At first, I thought they had been incredibly disciplined and made the effort to make their totals add up correctly to 2430, as the average of the 55 responses came to 2435. That was indeed the case for three of the 55 respondents. The other 52, however, ranged from a few wins either side of that to a stupendously optimistic 2593 from Jorge Montanez, and a staggeringly pessimistic 2223 from Christopher Szakacsi. Christopher’s opinion of the state of some of baseball’s basement-dwellers is so low that he predicted six 100-loss teams, including both the Orioles and Marlins winning just 45.
Clearly, the pressure to be optimistic is not felt so keenly by those who are not responsible for previewing teams. Let’s move on to take a look at the divisional standings, and where the chief agreements and disagreements lie.
It’s tight at the top in the NL East. That said, the Nationals are in first place in every set, albeit tied with the Phillies in both PECOTA and Public, and the Mets in Davenport. The Mets, Phillies and Braves are all incredibly close too on average, with the Phillies just getting the edge for second, while the Braves have the most disagreement in the division, with a total of 13 wins separating the top prediction from the bottom. The closest the Marlins get to not finishing last is in Davenport, where they’re a mere 14 wins behind.
If you thought the NL East was close, you might want to sit down. Just seven wins separate first from last on average here. PECOTA, 538 and the Public have the Brewers winning the division, FanGraphs and BttP take the Cubs, and Davenport and EW are pro-Cardinals. To be fair, Davenport’s five-game gap from first to last is about as close as one can get to a whole-division coin flip. The EW guest positivity really comes in here, as they have the three highest totals across all sets from Will Leitch (Cards), Sahadev Sharma (Cubs) and Robert Murray (Brewers), the latter tied with BttP’s Cubs prediction from Brandon Lee, not to mention Stephen J. Nesbitt’s vote of confidence in the Pirates.
Why not calm down after that chaos with the Dodgers, who have the third-smallest spread of predictions and are winning the division for the seventh straight year in every set. The previewers think it will be a contest with the Rockies, thanks to BttP writer Alexander Perez and Jesse Spector on EW, both giving them a low-90s win total. Beyond that, it’s not so interesting, although there’s relative hope for Padres fans with the Machado signing and their influx of prospect talent: they finish third in all but 538’s estimation. The Giants assume San Diego’s mantle as the last-place team, unless you ask Julie Parker of SFBay News.
For World Series Champions, one can argue the Red Sox get short shrift here. The Yankees were a 100-win team too, of course. Pay attention to that 99 from Alex Speier, who has made notable predictions the last couple of years, first with a spot-on 93 in 2017 and then by predicting 105 last year. Lindsay takes the crown for highest win prediction this year, and is in fact the only person to predict triple digits in 2019. The poor Rays are stuck in third place in every set, although with a win total that should be just good enough to earn them a trip to Fenway Park in the Wild Card game. The Orioles are unsurprisingly considered a 100-loss team by all but Davenport. They’re arguably the worst-projected team in PECOTA history, and certainly since I started this exercise. Sadly, all of these totals would be an improvement on 2018.
Not much has changed here, with the same projected order as last season’s post and the same division hierarchy of Cleveland cruising, Minnesota stuck in no-man’s land and not quite good enough to take the second Wild Card, and the bottom three just trying not to be the worst (or not trying, depending on your perspective). Here, Zack Meisel of The Athletic predicting just 89 wins for Cleveland qualifies as bold, and Mike Moriarty of BttP agrees with Davenport that the White Sox might escape that bottom class and almost get to .500.
Houston couldn’t quite make it to 100 in any set, and it doesn’t matter one bit. Once again, they dominate this division in every set, and BttP’s nine-win gap qualifies as the closest any other team gets. There’s some disagreement over whether the Angels or the A’s have to crane their necks up slightly less to look at the Astros, and they also both fall short of the Rays for a Wild Card spot. Davenport likes the Mariners more than everything else while Texas, curiously, seems to draw some optimism that they might be a winning team from the previewers and then extreme consensus from the other five sets, where they are at either 70 or 71. I don’t know whether Levi Weaver and BttP’s Eric Robinson colluded on the total, but it’s still not enough to keep them out of last.
Composite & Overall Standings
Finally, here’s everything in both an interactive graph (also linked here) and the final, fully sortable table of results, including the standard deviation and the composite projections. See you again in October, when we’ll once again find out who managed to be the least wrong about baseball.
Previous post: Red Sox Fans Should Not Panic about the 2019 Bullpen