Guest Post: Projecting The Jays

Thursday, January 04 2018 @ 09:24 AM EST

Contributed by: Gerry

Bauxite rabbit has penned an article projecting the Jays performance. Let him know what you think in the comments.

With no shortage of opinion on the Jays’ direction this off-season, is it possible to develop a common view of the team and potential roster decisions? To further the conversation, I simulated the 2018 season using a model that accounts for variable player performance. Simulate next season 100 times and you get team win probabilities. I won’t oversell the results, but I think this type of tool helps evaluate where we stand and the key choices ahead. Here’s a simple explanation of what I did and what it means.

From 2015 to 2017, Toronto was great (93 wins), good (89 wins), then bad (76 wins). Did they underperform last year or are the current Jays just bad? To examine this, I built an excel spreadsheet to predict team wins based on individual player performance. Using our current roster, I projected playing time and production for each player (at bats & OPS for hitters, innings pitched and ERA for pitchers). The resulting team OPS and ERA numbers were translated into runs for/against using simple formulae, and wins/losses calculated with the pythagorean formula. My ‘best guess”, or base case, predicts 80 wins. To understand the model’s sensitivity, I created two other scenarios. Here are the results, along with the past 3 years for comparison:

base:           80 wins	  (706 OR, 723 DR, .729 OPS, 4.13 ERA)
optimistic:    102 wins   (806 OR, 617 DR, .776 OPS, 3,53 ERA)
pessimistic:    57 wins   (603 OR, 815 DR, .628 OPS, 4.66 ERA)

2017:         76 wins    (693 OR, 784 DR, .724 OPS, 4.42 ERA)
2016:         89 wins    (759 OR, 676 DR, .755 OPS, 3.79 ERA)
2015:         93 wins    (891 OR, 670 DR, .797 OPS, 3.81 ERA)

It’s interesting that the optimistic and pessimistic cases produce such widely different results when both are driven by quite reasonable assumptions (Smoak repeats, Estrada bounces back, Sanchez/Tulo/Travis are healthy VS Smoak reverts, old players get old/injured, …). This highlights how fans often see what they want and why static projections have limited value: it’s unlikely that key players all simultaneously outperform, under-perform, or meet expectations. Like calling “heads” on a coin flip 10 times in a row (1 in 1000 odds), it’s not going to happen. Instead, each player will fall somewhere within their own upper and lower bounds of performance, creating a mix of individual outcomes across the team. It follows that team predictions incorporating this player variability would better assess potential.

Because of this, I expanded the model to individually simulate each player’s performance, randomly selecting optimistic, base, or pessimistic outcomes using related probabilities. This way, each simulated “season” incorporates a different mix of players performing above or below expectations … just like it happens in a real season. Simulating a large number of games then produces something really useful: a range of probabilities for specific win levels. Over a 100 season simulation, the model predicted the Jays would win 85 or more games exactly 4 times (4%) and 88 or more games just once (1%). Matching these win probabilities to Wild Card cut-off levels shows the current roster has about a 1% chance of making the playoffs in 2018.

The calculations are simplistic (OPS & ERA instead of WAR) and the player projections are mine alone, but the model is revealing - too many things have to go right for the team to succeed - and helps us see the team for what it is. More importantly, this is just the start of what can be examined. With the current roster lacking, what trades and free agent signings are needed to reach a 50% playoff probability, and how would the associated cost in payroll and prospect talent impact 2019 and 2020? More generally, how do different asset management strategies compare, especially when assessed across multiple seasons, where the costs of “win now” versus “win later” can be properly evaluated? With baseball run more and more by MBA bean-counters, and talent increasingly valued in similar ways, I’m sure many teams have tools to quantify these exact types of trade-offs.

Thanks rabbit for the post. If anyone else wants to write an article for the site, just email roster at batters box dot ca.