Projecting the 2021 Blue Jays: team distribution

In late March before the season began, I did a series of posts projecting the 2021 Blue Jays regulars. The main idea was to create and show a distribution of outcomes based rather than the single “point projection” that are published by various projection systems representing an expected value of median scenario.

Projecting the 2021 Blue Jays regulars: Catchers/DH | Infield | Outfield | Starting Pitchers

The logical extension to such an exercise is then rolling those projections up to the team level, and I intended to do so near the beginning of the year but didn’t get around to putting it all together. Even though it’s now a month into the season, I thought it would be interesting to take a look at what things look like, and then subsequently some of the best and worst outcome.

Having created 49 different outcomes for of 17 regulars according to a 7x7 matrix of playing time, I then did an abbreviated version for the eight relievers who figured to be most significant: Ryan Borucki, Tyler Chatwood, Rafael Dolis, Thomas Hatch, Tim Mayza, Julian Merryweather, David Phelps and Jordan Romano (Kirby Yates was already known to be out for the year at this point). Since relievers pitch fewer innings and accordingly have a much narrower range of outcomes, five innings scenarios and a 7x5 matrix of 35 outcomes sufficed.

That gave a total of 25 players projected (11 position, 14 pitchers), who cumulatively will not come close to covering all the other playing time. Here my approach was a different than the last time I did this for 2018. Because the hitters enjoy considerable positional flexibility to be moved around as necessary to fill out a lineup, and the main reserves are are not particularly differentiable in terms of talent, I forecast the bench as one unit with 25 possible levels of productivity ranging from atrocious to competent, but centred on essentially replacement level. Playing time is what spills over from the regulars, the difference between the total for the 11 regulars in a given scenario.

Likewise, the same thing is done for starting pitcher depth. Their innings are whatever flows down from the six forecast starters, with productivity forecast as a group on a 25 point scale. Here the possible outcomes incorporate the possibility for someone like Alex Manoah to come up and make a midseason impact, so there’s scenarios where this group adds real value.

Finally, the remainder of the innings are filled from by bullpen depth. This would be guys like Anthony Castro, Joel Payamps, A.J. Cole, etc; the waivers claims and minor league free agents who fill out pitching staffs and in today’s games cumulatively log a significant amount of innings. Again, it makes more sense to treat them as a group rather than trying to get ridiculiously granular.

All in all then, with 25 players and three groups of reserves, that’s 28 independent variables for each simulated season. I ran 100 trials of those simulated seasons, to get the following team level distribution. Keep in mind, this is what things looked like on Opening Day, unadjusted for anything that’s happened since:

Given that team replacement level is about 48 wins, the bulk of the distribution lying between 33 and 39 WAR suggests a low- to mid-80s win total, which roughly fits with where the Jays were forecast in the pre-season.

What’s more interesting is the shape of the distribution. On the low end, it rises quickly and steadily from the high 20-WAR totals into that flattish midsection, suggesting a relatively high floor largely due to the sheer amount of positional talent around the diamond. In the absolute worst of the 100 scenarios, those 11 regulars combine for just under 17 WAR, which in most years would rank just a little below average.

On the high end, there’s a much longer and thinner tail of positive outcomes at 40+ WAR. This reflects the number of younger position players entering their primes who could consolidate or breakout in 2021, as well as the potential for bouncebacks seasons from a number of starting pitchers. In both cases, there’s significant upside if things break right, somewhat akin to what happened for the 1985 Blue Jays.

That’s WAR, but what about wins directly? Here more uncertainty has to be built in. WAR is just production of individual events, and that translates imperfectly into runs. Likewise, while runs and run differential are related to wins, it’s not always linearly. Hence, two more random variables to approximate the possibility for outliers (though in most cases the combined effects offset or are small). That results in the following forecast win distribution:

Fundamentally, it’s the same distribution, just with a wider range of outcomes to reflect the possibility of good and bad “luck”. The red line is a (smoothed) cumulative probability distribution, which really picks up starting in the high-70s and flattens out around 90 wins (the 20th percentile is 78 wins and 80th at 89 wins). The median lies at 83-84 wins, with worst and best case scenarios at 68 and 98 wins.

One final caveat is that this considers only players within the organization. It explicitly doesn’t contemplate the possibility of midseason additions if the Jays are in contention or fall out and sell off expiring contracts.

To wrap this up, later this week I’ll break down the best possible 2021 simulated seasons, and the worst case simulated seasons in terms of what goes right and wrong. And then at the end of 2021, hopefully circle back to see what this analysis got right and wrong.