Connect with us

Toronto Blue Jays

A final autopsy of the 2018 Blue Jays and how things went wrong

With just a few hours left before 2018 draws to a close and the calendar flips to 2019, I thought it would be worth tying up one loose end.

Projections are useful to help frame expectations, but typically focus on one number — a point estimate — rather than the expected variation around those numbers and the distribution of outcomes. So last winter, I tried to remedy that by doing player projections that created a range of possible outcomes (position players / starting pitchers).

I then extended this to the team level, projecting 12 players and six pitchers with some simplifying assumptions are depth and the rest of the roster and ran 100 simulations for the 2018 Blue Jays (43 independent random variables for each simulation). About a week into the season in early April, I showed what the top five projected 2018 seasons looked like.

The plan was to immediately follow-up with what the worst five projected 2018 seasons looked like. But it got delayed and by the time I circled back the Blue Jays were off to a torrid start and it didn’t seem particularly relevant: whatever the ultimate outcome, postseason or not, there was no way the 2018 season was going to go completely to hell.

Such prescience. Anyway, before totally moving past 2018 and turning to 2019 a new set of projections, I thought it would be interesting to complete the exercise in retrospect and compare the nightmare scenarios with what ultimately came to pass.

One important thing to note is that these projections were for the Opening Day roster over the whole year, so the potential for a sell-off was explicitly not considered. That’s relevant, given that in the end the team won 73 games which ironically was the lowest expected outcome. That would normally imply bias in the projections, but in this case it means the Jays didn’t have the services of a number of players in the second half whose production was factored into the projection.

Worst simulated 2018 Blue Jays team: 73 wins

Though the wins match exactly, this simulated season actually doesn’t match up very well to what actually happened. On the pitching side, Marcus Stroman had an ace level 5-win season, with J.A. Happ close to what he actually did at 3 WAR. But none of the other starters were effective, maxing out at 1 WAR despite largely healthy seasons. Jaime Garcia was the exception, with Joe Biagini being servicable in picking up the innings. If one substitutes Borucki for Biagini, that’s closer to reality, but there wasn’t much for more than six starters.

On the position side, Russell Martin, Justin Smoak, Kevin Pillar and Curtis Granderson were broadly in line with their real seasons. Josh Donaldson posted 3-wins, but over a full season. Troy Tulowitzki came back in the second half and was decent in posting just under a win in 300 PA. On the flip side, Randal Grichuk was replacement level and Kendrys Morales repeated 2017.

So all in all, though similarly miserable, not really close to what actually happened.

1st percentile 2018 Blue Jays team: 76 wins

This scenario I think is closest to what actually transpired. On the pitching side, J.A. Happ was the best pitcher, turning in 2.5 wins in a shortened 125 innings. Aaron Sanchez was the next best pitcher with 150 decent innings, so that was off. None of the other starters did much, but they got almost 2 WAR from the 6th starter. That was forecast to be Biagini rather than Borucki, but in a similarly roundabout way to Happ, the simulation nailed what the Jays actually got.

Positionally, things were remarkably close. Tulo effectively missed the whole year, with some decent production off the bench replacing home (though the simulation had Diaz doing little and Solarte excelling). Pillar had a 4 WAR season, so that was high, but Smoak was replacement level so that’s a wash. Donaldson was okay (2.5 WAR) in 525 PA, so even that was optimistic.

The bullpen was very good in this scenario, providing about 3-4 more wins than the Jays actually got. Subtract that, and you basically tie right into what the actual 2018 Jays did.

2nd percentile 2018 Blue Jays: 76 wins

This is another simulation that ends up pretty divorced from reality. Marco Estrada returns to 2016-17 form and posted almost 4 wins, with Jaime Garcia adding 2.5 wins. Sanchez was decent, with Happ the worst starter. So that…didn’t happen.

Other than Tulo again missing almost all the year with Diaz stepping in serviceable as he did, very little lines up positionally either. Granderson edged out Pillar as the best outfielder at 1.2 WAR. Solarte and Travis had good years, everyone else was pretty miserable. Donaldson did turn in 3 wins in just 375 PA, so that would have been nice.

The next three: 77 wins

It gets repetitive to go into detail on each scenario, but there are some higher level trends and notes. Donaldson was again worth 2-3 wins through varying level of performance and playing time. This underlines how important he was — in all these “worst case” scenarios, he was way below his 2013-17 production, and there were very few below .500 seasons where he added 5-6 wins or better.

Ironically, in all of three of these scenarios, both Sanchez and Stroman were very good, with all season coming in the 3-5 win range. That obviously implies a lot of really bad offsetting performance elsewhere, but would certainly have left the Jays better positioned for the future in terms of being able to cash their last two control seasons this winter if they wanted to.

One thing I’ve thought about quite a bit over the last couple months is whether my projections were overly bullish on the 2018 Blue Jays, given that even notwithstanding the caveat mentioned at the outset, they finished at the very bottom end of the initial distribution. As a team, the underperformance was in the neighbourhood of two standard deviations. We’d expect that to happen to a handful of teams, so that is not a red flag in itself.

The bulk of that was on the starting pitcher side, with only Happ meeting or exceeding the 50th percentile outcome, and the others falling well flat. Looking over the individual forecasts, I struggle to see systemic bias in that the base forecasts were heavily based on objective systems such as Steamer and ZIPS. And then of course, there’s what happened to Donaldson. If there’s one thing I think has to be adjusted, it’s the potential for total collapse by a player — what happened with Solarte and Travis posting significantly negative WAR.

And now: bring on the New Year!

Source Link

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

More in Toronto Blue Jays