There's a lot of strong dynasty analysis out there, especially when compared to five or ten years ago. But most of it is so dang practical-- Player X is undervalued, Player Y's workload is troubling, the market at this position is irrational, and take this specific action to win your league. Dynasty, in Theory is meant as a corrective, offering insights and takeaways into the strategic and structural nature of the game that might not lead to an immediate benefit but which should help us become better players over time.

You Don't Know What You Don't Know

Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don't know we don't know. And if one looks throughout the history of our country and other free countries, it is the latter category that tends to be the difficult ones.

-Donald Rumsfeld 

Few offhand remarks have captured the public imagination like Rumsfeld's casual categorization of risks in 2002— so much so that Rumsfeld chose it as the title of his memoirs. But Rumsfeld didn't invent the schema; it was drawn from the Johari window, a tool designed in 1955 for use in therapy and self-help which had become a popular framework everywhere from the intelligence community to NASA. (I first encountered it in theater classes, where it was commonly used for building a backstory and developing a character.)

The Johari window showed up in so many places and Rumsfeld's comment elicited so much discussion because the framework is a truly useful tool for considering uncertainty. Noting the distinction between known unknowns and unknown unknowns is not necessarily intuitive (or at least it's not usually reflexive), but the two categories present different risks and must be accounted for in different ways.

At the risk of messing with a classic, I would like to paraphrase it for a moment to draw another distinction. We know there are knowable unknowns; that is to say we know there are some things we do not know, but are capable of knowing. But there are also unknowable unknowns-- the ones where it's not that we don't know, but that we can't know.

(Like Rumsfeld, I'm not proposing a brand new schema; this idea already exists in the field of uncertainty quantification, which distinguishes between so-called epistemic uncertainty— things that could be known— and aleatoric uncertainty— things that are beyond our ability to know.)

Two Types Of Uncertainty

The word "epistemic" means "relating to knowledge", while the word "aleatoric" means "characterized by chance", definitions which go a long way toward highlighting the differences between them.

Let's say I hold up a coin and ask you about the chances it will come up heads when I flip it. Your first concern should be whether or not this is a "true", or unweighted, coin. The fact that you don't know represents epistemic uncertainty. The odds are very different if the coin has heads on both sides, after all.

But even if you know whether the coin is true or weighted, you still don't know how the flip is going to come up. The best you can do is provide an estimate of how frequently one side or the other will result. That's because coinflips are random processes and, therefore, subject to aleatoric uncertainty.

If I fire an arrow from a bow, you could calculate with a fair degree of accuracy where that arrow would land... if you knew the power with which I drew the bow, the weight and aerodynamic properties of the arrow, the prevailing wind conditions, and all other relevant factors. Most of the uncertainty in the arrow's landing spot is, therefore, epistemic uncertainty.

But even if you knew and precisely calculated all of those variables, if I launched fifty arrows with the exact same inputs each time, there would still be some variance in where they landed. Every arrow has subtly different mechanical properties; it will vibrate and flex in subtly different ways as it flies, and these variations will result in slight deviations in their path. This is aleatoric uncertainty.

(Some would quibble that this isn't technically aleatoric; in theory, if we knew the precise mechanical properties of each arrow and had suitably complex models and computers with sufficient processing power, we could precisely calculate all of those deviations. Perhaps! Though for our purposes I will note that "practically unknowable" is the same as "theoretically unknowable".)

 Most uncertainty is similarly a combination of knowable unknowns and unknowable unknowns. What are the chances I get cancer in my life? There's a lot of epistemic uncertainty there, things that we could theoretically know that would allow a better estimate. Am I a smoker? (I am not.) Do I have a family history of cancer? (I do.)

But at the end of the day, there's a lot of aleatory uncertainty there, too. Sometimes, smokers with extensive family histories of cancer remain cancer-free. Sometimes, non-smokers with clean family trees get cancer. We can't ask enough questions and get enough answers to reduce the uncertainty here all the way to zero.

Why Does This Matter For Fantasy?

Uncertainty is one of the most important things when building a dynasty team. If everyone knew exactly how every career would turn out, there'd be no room to gain an edge; fantasy football would be a solved game. I love to say that dynasty (and fantasy football in general) is a game of pricing risks.

Uncertainty is good and should be embraced. If players are available in free agency, it is because everyone in your league thinks they are bad and not worth rostering; the only way they hold any value for you is if everyone in your league is wrong, so a large amount of uncertainty is helpful. But different types of uncertainty operate differently and work best in different roles.

Consider Rome Odunze, selected 9th overall in the 2024 draft. He is unlikely to help you win much of anything in 2024; Footballguys' most recent projections have him as WR50 for the rest of the year. But if he delivers on his potential, he could contribute to many championship runs to come. There is a lot of uncertainty about how his career plays out, but this is largely epistemic uncertainty. It hinges on a question that is knowable but unknown: how good is Rome Odunze?

Now consider Justice Hill, who was selected 113th overall in the 2019 class. There's far less epistemic uncertainty surrounding Hill; after four years in the league, we have a much better idea of who he is as a player. He is far less likely to prove our initial read wrong. He is also quite unlikely to help you win much of anything in 2024... unless Derrick Henry, the man ahead of him on Baltimore's depth chart, gets hurt.

If Henry gets hurt, Hill becomes the only experienced back on a dominant team that plays with a lot of leads. He certainly won't replicate Henry's production, but there's a great chance that he provides enough value to break into starting lineups. Will Henry get hurt? Probably not, but we can't know. It's aleatoric uncertainty.

In the short term, aleatoric uncertainty dominates. Despite being significantly cheaper to acquire, Justice Hill is probably more likely to contribute to a championship this year than Rome Odunze. Every year, we see backup running backs benefit from massive volume spikes after an injury to the starter. Last year, Zamir White had 20 carries for 54 yards through 13 games. An injury to Josh Jacobs led to White getting 84 carries for 397 yards over the final four games. He was the #8 running back during the fantasy playoffs.

In the long run, though, epistemic uncertainty is what matters the most. Jacobs left the Raiders in free agency and White hasn't continued to produce in his absence because the relevant question was "how good is Zamir White?" and the answer to this point seems to be "not very".

Rome Odunze is more expensive than Justice Hill because even if the aleatoric uncertainty breaks Hill's way and he contributes to a championship this year, long-term production is largely a function of talent and there's simply not that much uncertainty about how talented Hill is at this point.

Recognizing the difference between aleatoric and epistemic uncertainty and knowing the different roles they play allows you to make better decisions to maximize your limited roster space. Contenders should make an effort to add some aleatoric uncertainty so they have exposure to potential short-term upside to help their pursuit (though they shouldn't completely ignore epistemic uncertainty unless they want their title window to be short-lived). Rebuilders should eschew aleatoric uncertainty and focus more on epistemic uncertainty so they can build long-term value.

And both types of teams should prefer either type of uncertainty at the end of their roster over the known knowns— players with little doubt as to how good they are and little path to a larger role.