Steve Cohen Said The Probabilistic Part Out Loud

When the Mets' owner spoke in the language of probabilities, he opened a new dialogue in baseball — and pulled back the curtain on the front office.

Since liquidating an enormous amount of talent at last week’s MLB trade deadline, the New York Mets have lost six straight, dropping their record to 50-61 and falling 21 games back in the NL East standings. When they woke up Monday morning after getting swept by both the Royals (36 wins) and Orioles (MLB-best 71 wins), the Mets were 7.5 games out of the wild-card picture with 51 games left to play. 

Unsurprisingly, the team’s chances of making the postseason are now 1% and its probability of winning the World Series is less than 0.1%. With those odds, even journeyman optimist Lloyd Christmas would find it damn near impossible to muster a grin and lean into the idea of, “So you’re telling me there’s a chance, yeah!” By design, the Mets’ season is now done.

But in the days leading up to the trade deadline, Mets center fielder Brandon Nimmo was still a believer — so much so that he lost sleep as the team was in the process of becoming MLB’s biggest deadline seller. Three-time Cy Young winner Max Scherzer reacted to the fire sale with a mix of disappointment and thinly veiled discontent toward the front office. 

After Scherzer waived his no-trade clause and ended up with the Texas Rangers, he put the Mets on blast by telling reporters — and expressing his disbelief — that New York hadn’t just waved the white flag on this season but maybe even next summer as well.  

These were expected responses from big leaguers, who are wired like Han Solo: Don’t tell them the odds. The fact that they’re in the majors — the chance of any U.S. high school player getting to the highest level is 1-in-829 — means they’ve already overcome far greater improbability than a team staring down FanGraphs’ projections in this year’s playoff race. You don’t get to The Show without having immense self-belief, which isn’t easily turned off.

Conflicting visions between owners and players (and by extension fans) have long been part of the game. What’s unique about this year’s trade deadline is the public-facing language that Mets owner Steve Cohen used to explain the process behind his roster moves. There was no coach-speak aimed at the clubhouse, no false hope pitched to paying customers. 

“When you look at the probabilities, what were we, 15%?” Cohen said. “And other teams were getting better — you have to take the odds down from that. If you’re going to have a 12% chance of winning — just getting into the playoffs — those are pretty crummy odds. I wouldn’t want to be betting any money on that. I don’t think anybody else would, either. I said before, hope is not a strategy.”

Moneyball has been a part of baseball’s vernacular ever since the eponymous book about the Oakland Athletics mining for data insights was published nearly two decades ago. Every roster decision is always guided by an analysis of cost versus benefit, risk versus reward, and short- versus long-term thinking. For an owner like Cohen, who has made billions crunching numbers, it makes sense that he relies on statistical models. 

What’s new to the game, however, is how Cohen used those probability models to speak directly to players and fans.

“At some point, we have to go win two-thirds of our games,” said Cohen, who made the decision to sell, sell and sell some more at the deadline. “We had shown no consistency along the way. So it really would have to take a stretch to believe that something would change now.”

Whether Cohen was talking about internal playoff probability models kept by his own (growing) analytics team or even the public forecasts available at websites such as FanGraphs and Baseball Reference, he was right that New York looked far more likely to miss the playoffs than to make them on the eve of the deadline.

But if an owner is going to speak the language of mathematical models to fans, it’s important to understand what goes into building them.

Most playoff forecasts use a version of this process: Start with a talent rating for every team, which can be used to generate predictions for any head-to-head matchup on the schedule. Then simulate all of those remaining contests a bunch of times, tracking how often each simulated season ends with each team making the playoffs, winning its division and winning the World Series. (You have to make sure you’ve programmed all of the fun tiebreaker rules into your simulation, so you can accurately say who would make the playoffs if certain specific scenarios play out.)

Based on his comments, out of every 100 simulations Cohen looked at, only 15 ended with the Mets mounting a stretch-run surge, making the playoffs, and ultimately justifying the record $353 million he’d sunk into this year’s Opening Day roster. In the other 85 simulations, they missed out — and it would have made no sense hanging on to the expiring contracts and expensive veterans the team was carrying before the trade deadline.

Of course, there is no one canonical answer to what the Mets’ playoff odds were last week, only estimates. And within our basic forecast framework, there are unique choices you can make to customize a model: For instance, you can base your ratings on a team’s recent run differential, or on a projection of its future roster talent; you can also handle the uncertainty around a team’s rating (and how it might change over time) in different ways. But the end result is always a probabilistic forecast that provides a realistic guide to help with a team’s decision-making.

The key word is realistic. 

When judging probabilities, it’s not only important that events with, say, a 15% likelihood (like the Mets making the playoffs) fail to happen the vast majority of the time, but also that they do happen some of the time. In other words, when a forecast is calibrated well, a 15% chance really means a 15% chance. And in recent seasons, MLB forecasts have been pretty good in that regard. 

FiveThirtyEight has an interactive chart that compares its (now-defunct) playoff forecast to the results that actually happened each season. Here’s what we see when we plot a team’s predicted playoff chances during the week leading up to the trade deadline against its actual rate of making the playoffs:

There are some anomalies along the ideal 45-degree-angle line. Teams with less than a 30% chance, for instance, tend to make the playoffs less often than predicted — perhaps the effect of selling off talent at the trade deadline and weakening their roster down the stretch. This is mirrored at the other end of the chart, where teams with a better than 70% chance mostly outperform their forecast, perhaps because they are the buyers who benefit from adding extra talent. But for the most part, there’s a tight relationship between what playoff forecast models predict and what ends up actually happening.

That’s why Cohen could feel confident that the 2023 Mets were never going to catch their rivals, and thus shifted the team’s focus to the future while playing out the string on a lost season.

But Cohen got one thing wrong in his post-deadline explanation: In some ways, hope is part of his strategy. The hedge fund titan tossed figurative buckets of ice-cold water on the players’ and fans’ hearts roughly a week ago, but by trading veterans and expiring contracts for young prospects, he ultimately boosted the longer-term chances of the Mets winning a World Series. 

To make it worthwhile to Mets fans, though, Cohen is betting the boost in odds was worth giving up on that 15% this season — a question that can only be answered in due time.

---

Filed under: Baseball