Prior to the 2021 season, last winter I put together distributions of potential outcomes for the players expected to be Toronto Blue Jays regulars, which then served as the basis (along with some other factors) to create a team level expectation. The main idea to go beyond the ubiquitous single number “point” projections that represent an average expectation, but don’t consider the range of possibilities.
With the calendar on the verge of flipping, as a final lookback on 2021 and in the interest of accountability, I thought it would be interesting to see how those projections fared vs what actually occurred at the player level, and see how close some of those scenarios came at the team level.
Let’s start by running down the position regulars:
- Vladimir Guerrero Jr — at just under 7 WAR, Vladdy came in about 20% above the most optimistic scenario. Some of this was staying healthy and playing everyday to post 700 PA that was essentially the high end of the forecast, but the bigger element was his productivity blowing well past even the high end of what I thought reasonably foreseeable
- Marcus Semien — I see his 6.6 fWAR as more representative than the 7.3 bWAR, but even that was essentially the top percentile of scenarios. That was a combination of coming in right at the high end of productivity expectations plus exceeding the high end of playing time expectations by not just staying healthy but playing literally everyday
- Bo Bichette — Here too is a significant divergence between 4.9 fWAR and 5.9 bWAR, with the former being about the 90th percentile with the latter essentially the very top. They key factor here was staying healthy given some history of missing time to injury, with 694 PA representing a >95th percentile outcome. His productivity was more in line, favourably above base case at the 70th-90th percentile depending which production number
- Teoscar Hernandez — Playing time ended up right around expectation (143 GP/575 PA), with productivity at the extreme high end and thus ~4 WAR ended up at the 90th-95th percentile of the overall expectation
- Lourdes Gurriel Jr. — 541 PA was almost exactly right down the expectations fairway; productivity depends on the system. 1.5 fWAR was about the 35th percentile; 2.7 bWAR about the 75th.
- George Springer — His 2.4 WAR ended up around the 35th percentile due to playing time coming in around the 10th percentile. The productivity when was just fine when he did play (~60th percentile)
- Danny Jansen — 205 PA came in at the very low end of playing time expectations, thereby pinning overall production just below the median despite strong productivity that conversely was at the very high end
- Randal Grichuk — playing time ended up around the 70th percentile at 545 PA, but his productivity cratered to about the 20th percentile expectation and overall production thus came in around the 25th percentile.
- Alejandro Kirk — 189 PA came in around the 5th percentile as he missed almost three month, but when healthy his productivity was around the 60th percentile.
- Cavan Biggio — Easily the biggest disappointment, with playing time at the very bottom of the projected distribution (having no history of missing time professionally), while simultaneously productivity did the same and thus overall production was essentially the 1st or 2nd percentile of expectations
- Rowdy Tellez was another with a lost season, with productivity tanking to the very low end and playing time ending up near the bottom too (~20th percentile).
In total, that’s four players who significantly-to-massively outperformed, four who underperformed (including Springer based on missing half the season), and three who had mixed seasons. At a micro level, that’s reasonably balanced.
Overall, these 11 players who projected for significant/regular playing time totaled ~28 fWAR and ~32 bWAR. Cumulatively, that worked out to at least the 90th percentile of my projections, with the higher total close to the very high end (33 WAR in the best scenario). In hindsight, there were a few places I might have been a little light, but overall I think it was more a confluence of broadly fortuitous health and standout seasons for the ages than being systematically too light.
Let’s turn to the starting pitchers:
- Robbie Ray — Almost 20 more than his previous career high, 193.1 innings exceeded the very high end of the projected expectation. Run prevention is a matter of interpretation, with a 87 FIP- at the strong end while 63 ERA- blew the most optimistic of expectations out of the water. Consequently, even ~4 fWAR is the very high end of the projection, >6 bWAR blows is completely off the chart.
- Hyun-Jin Ryu — 169 inning was a strong total at the ~80th percentile of expectations; the 97 ERA- and 94 FIP- conversely are closer to the 20th. The net of that is that ~2 WAR comes in around the 40th percentile. Not a great season, not a disaster.
- Steven Matz — 150 innings came in around the 80th percentile, while his 85 ERA- and 89 FIP- made for roughly the 90th percentile, and with both components so strong the total of 2.8 WAR was ~95th percentile. It was as good a season as could be expected within reason.
- Ross Stripling — 101.1 innings came in around the 30th percentile as he missed over two months, with a little split between 107 ERA- (~40th percentile) and 122 FIP- (~20th). Overall, it puts his 2021 at the 20th to 30th percentile with less than 1 WAR
- Tanner Roark was written off, with the -0.3 WAR coming in around the 15th percentile
- Nate Pearson ended up a non-factor for the purposes of the rotation
Overall, these pitchers totaled 11 BWAR (~90th percentile) and 9 fWAR (~65th). Most of that outperformance over the median of 8 WAR is Robbie Ray being so far off the chart, and while I’m biased I think it’s fair to say no one saw that coming. On an individual level, there’s a balanced range from huge outperformance to huge underperformance of expectations, so I don’t think there was systemic bias.
The two other major contributors were Alex Manoah, who was not explicitly forecast but considered part of depth starting, and Jose Berrios. One of the limitations of this exercise is it necessarily doesn’t consider players external to the organization. Manoah blew the best case scenario for best starting out of the water (~4.50 ERA), so maybe I was too conservative there but then he was well beyond any reasonable (or perhaps even plausible) expectation.
These individual projections as well as further projections for relievers and a multitude of other factors were then rolled up into a team level projection (with 55 different independent variables/factors in each projection). Interestingly, over the 100 simulated seasons, 91 wins was one of the four most common outcomes at 8% (along with 78, 82, and 87 wins). It effectively amounted to a ~85th percentile outcome, spanning the 81st to 89th.
Granted, about 3 WAR came from players who were acquired during the season, and thus not really part of that forecast. If we back out three wins, 88 wins puts the 2021 Blue Jays in the low-70th percentiles (71-74) of my projections. Again, I think that feels about right, and on the whole moe things went right than wrong for the Jays.
That said, given the massively outlying seasons from Vladdy, Semien and Ray; very strong seasons from the likes of Teoscar and Manoah; the conversely massive underperformance of their run expectation; impactful injury to Springer...there’s weren’t any simulated seasons that lined up especially strongly with how 2021 actually went.
Finding the projected seasons that were closest to the actual 2021 season required a more systematic approach of minimizing the error of each of the major projections (the 11 regulars and six starters above, the bullpen collectively, the rest of the team, and the translation factors from event production to runs, and runs to wins).
Curiously, the season with the least deviation to the actual 2021 Jays was one in which they ended up with just 78 wins. The positional regulars end up light across the board with 22 WAR, but the trends by player are directionally right. Vladdy and Semien were the leads, except at ~4 WAR rather than 6+, Bichette was the next best at 3 WAR. Springer was the only regular to materially beat his actual 2018.
The rotation was a little weirder. Ray was the leader, but with just under 3 WAR. Roark was the second best starter, with Matz instead having the terrible year. Flip those two and it would be even better, but Ryu and Stripling come in almost bang on. The bullpen came in about right overall, and the Jays got a boost off the bench tracking the contribution from Espinal. It misses Manoah’s (a big factor in why they end up worse), but a big factor that makes it match is the team underperforming its Pythagorean (run) expectation by 5 wins.
Trial 27 wasn’t quite as good a match on individual factors, but matched better in that it was an 87 win team (to which acquisitions such as Berrios would add a couple wins). Positionally, it’s similar with Vladdy, Semien, Bichette and Tesocar having strong season but well shot of whatthey actually did (14 WAR vs. 23 actual). Springer and Jansen end up stronger than actual by not missing time, but Biggio has a poor season.
The rotation is stronger, with four starters in the 2-3 WAR range but no ace (similar to the 2010 Blue Jays of Romero/Cecil/Marcum/Morrow). Thus Ryu and Matz end up about right, but way light on Ray. Curiously, again one of the biggest misses is Roark, who again posts the kind of quality season the Jays were expecting in signing him. Again, a critical factor in achieving the match is a 4 win underperformance of run differential.
Overall, what surprised me was the diversity in outcomes of the closest matching seasons on individual factors. The top five seasons ranged from 71 to 96 wins, with an average of 83. Given there were about 30 seasons of 85+ wins, I would have thought more of the closest matches would necessarily be drawn from there.