A new religion is forming before our very eyes. Only this one doesn’t ask you to believe – it asks you to comply. Its deity is data. Its gospel is probability and, in a world, obsessed with being right, it offers something even more valuable: Protection from being wrong.
Billions have been poured into building this machine. The global sports analytics market is already in the billions and climbing fast, with teams spending millions annually on data infrastructure, analysts, and AI systems.
Most professional teams now rely on analytics departments, and entire leagues have built centralized data ecosystems to feed them. At that point, the numbers aren’t just helpful – they’re institutional.
You don’t invest that heavily in answers just to ignore them. And the more a sport allows time for decisions – baseball between pitches, football between downs – the more those answers begin to feel less like suggestions and more like instructions.
In the 2023 playoffs, the Toronto Blue Jays pulled Jose Berrios early in a scoreless game, despite the fact he was pitching incredibly well. He had thrown three shutout innings in just 47 pitches, yet he was removed largely due to upcoming left-handed matchups. The analytics said so.
The move was pre-planned and matchup-driven which is a classic analytics play. Bringing in the leftie, Yusei Kikuchi backfired badly. The Jays lost and the season was over.
Though the move was heavily debated there was this sense of validation from within the organization, as if their reliance on these probabilities based on the data was always the right move and, in a sense, it was.
That’s exactly what analytics do. They chart patterns and probabilities to recommend the most likely outcomes. Not all of the outcomes of course – which is impossible – but at least the likeliest ones.
If something happens 86 times out of 100, analytics can account for the 86 probable outcomes but not the 14 outliers. To capture those results you would have to hope that the opposite strategy was the correct one although sometimes the best “strategy” isn’t any strategy at all. It’s just dumb luck, or a great hunch, depending on which religion you follow.
If Kikuchi had come in and dominated, analytics would be hailed as genius. When he failed the process was correct, although the outcome was the exception. If Berrios had been left in, the results – either way – would have qualified as irresponsible or instinctual. Again, depending on the religion you follow.
What we’re left with on the data side is an infallible pseudo-religion where following AI scripture is never wrong, even when it is, and because of that, I think we’re but one brilliant renegade away from an alternative digital revolution. One where the strategies of the game basically ignore the numbers.
This Old School thinking was standard before the introduction of real-time analytics. Of course, it had to be because that’s all they had. Coaches developed tactics based on real world experience and these strategies put their personal signature on the game. Learning patterns and tendencies for them came only from doing.
When the data revolution arrived, the games changed dramatically. Probabilities and projected outcomes became the norm and Old School coaches fought for relevance. Over time the data would transform some coaches and devour others.
You either floated with the tide, or it swept you under. To survive you just needed to figure out your own team’s devotion to this newfound religion and whether it was based on revelation or financial commitment. You either got on board, or you hit the road.
Organizations were either slaves to the algorithm or enlightened followers of it, and I believe the exact stance is based almost entirely on a proximity to the team’s ledger. If you had in hand in justifying the fee, you’re very likely a big believer in The Church of The Numbers mainly because your job depends on it.
Let’s be clear. AI isn’t the enemy here. It’s nothing more than a helpful (and sophisticated) tool but one that comes with an inherent flaw: it can never predict an anomaly. When it’s tasked with charting past results to determine future outcomes how can it possibly factor in an outlier?
And sports are all about the outlier. Like pinch hitting someone on a hunch rather than on the data. That seems utterly impossible to quantify. Or is it?
When teams are less dependent on the sanctity of their data to make decisions – and if they’re successful doing so – how does the technology account for this in the equation?
I asked it directly. The response:
“In practice, it’s less mystical than it sounds. The system logs the decision, the context, and the outcome – then compares it against what was expected to happen. If that “wrong” choice keeps outperforming the model’s prediction in similar situations, the model adjusts its weighting. Maybe it was undervaluing a certain type of pitcher, a matchup nuance, or even a game-state variable it didn’t fully account for. The outlier doesn’t stay an outlier for long – it becomes a signal. Not because the system predicted it, but because it was forced to learn from it.”
“The model doesn’t predict the heretic. It studies him afterward.”
Basically, the algorithm can never be able to predict the outlier. And that’s fair. That’s what data modelling is. It makes choices based on things that have already happened and what is also likely to happen. It majors in history and minors in prediction.
That’s the irony within sports analytics. It’s just really good at keeping records with a predictive element that seems revolutionary but is little more than a pattern recognizer. And by following it devoutly, as if it was sacred religious doctrine, you’re doing nothing more than making safe bets based on established trends, even when the safest thing to do is actually the most disruptive – like going for it on fourth down.
There’s more irony. If everyone is basically reading from the same data-driven playbook then your edge is actually acting alternatively. Doing what the others aren’t doing, even when the “safe” play is the risky one. Suddenly punting becomes a risky play.
We then end up in a world divided by an aversion to risk where one side makes a choice counting on the data to back them up and the other side does whatever it’s told not to do.
Deviation, in any capacity, thus becomes an advantage.
Innovation then becomes a question of would you rather be wrong, knowing you should have been right, or would you rather be right knowing you were probably wrong?
Turns out the heretic is only one willing to take that risk.