Does baseball believe in analytics?
On luck, hubris, and the idea that the Dodgers are the best hitting team in baseball.
The most analytic-happy sport—baseball—is still not that good at analytics. I recently heard that Tim Kurkjian, ESPN’s esteemed baseball analyst, thought this Dodgers lineup was the best he could remember. Kurkjian has been covering baseball since 1981.
And it’s not even obvious the Dodgers have the best lineup this year.
Atlanta and the Yankees both have excellent rosters. The addition of Juan Soto to the Yankees means the Yankees could reasonably have the best two hitters on the season, if things go in their direction. And Atlanta has an improving, MVP-calibre outfield. Most of Los Angeles best players are past their prime.
In fact, Ohtani, who is Los Angeles most prized offseason pickup, doesn’t even feature in the top four most impactful players on their roster now that he’s not pitching.
What is going on? How come baseball—the sport that started the analytics revolution—is still so blind to the numbers?
Part of this is that predictions are really bad, so that analysts can get away with it. Tom Tango, who literally wrote The Book, on baseball analytics, has a good-enough player projection system: the Monkey System.
There are many, many problems with this. That’s why I call it a monkey system. I developed the worst forecasting system that you could possibly accept. And it takes minutes to run. Any other forecasting system out there, that spends countless hours to design, develop and execute, better be able to beat this system.
It’s not good. He admits it’s not good. But when put head-to-head against the best forecasting systems, it doesn’t look “not good”, it looks “average”. Which means on average, forecasting systems are not good.
It’s also because of luck. Luck is a big factory in baseball. It can account for as much as a 20-game spread between two teams. That’s roughly the difference between the NL Central’s first-place Brewers (2023: 92-70) and last-place Cardinals (71-91).
FanGraphs has a good, old post on this, and some of the sources of luck in hitting. At the team level, there are even more factors.
But part of the problem is, we mostly only ever focus on bad luck. We rarely focus on good luck. For every team or player that underperforms and should be expected to rebound the next year, there is also a team or player who is expected to regress because they over-performed.
We have a sense for this in football through fumbles. We know that fumbles are luck-dependent. Teams that win the fumble-ratio are lucky, and will have better outcomes than we expect.
The Kansas City Chiefs are a great case in point here. They won fumbles at a 4-1 ratio in their final two games. Lucky. But because the luck swung their way, we don’t talk about it.
The idea of modesty—admitting that there is a lot of luck and uncertainty in baseball—also goes against the baseball culture. Baseball seasons are long, so they must be good at discovering who the best team is (not true). Baseball has a rich history of newspapers and boxscores, so baseball writers are good judges of talent (not true). Baseball is really hard, so it would be impossible for someone to have a good year just by getting lucky (not true.)
There are a lot of reasons that we should be less-confident in baseball predictions than we are.
But we should also use the tools we have, e.g., the Marcel system, to make claims about who the best players are. And, unfortunately for Los Angeles—it looks like the best hitters are all on the East Coast.