A teenager's bad season should count for less than we let it

Draft models fold age into a single adjustment when the real effect is about how fast a young player is changing.

Age adjustment in draft models is done backwards. Not wrongly, exactly. Backwards.

The standard approach takes production and applies a correction for how old the player was relative to his peers, so a nineteen year old producing at the same rate as a twenty two year old grades out higher. That is sound. Young players are competing against bigger, stronger, more experienced opponents and holding their own, and the models that ignored this got embarrassed for a decade.

But the adjustment treats age as a level effect when the important part is a rate effect. The reason to prefer the teenager is not only that his output is more impressive today. It is that his output is a snapshot of something moving quickly, and a snapshot of a moving thing tells you less than a snapshot of a still one. Which means a young player's bad season should be discounted harder than his good season is inflated, and almost no public model does that asymmetrically.

Think about what changes between eighteen and twenty two. Strength, obviously, but also role, coaching quality, minutes, whether he is asked to defend the opposing team's best player, whether anybody has taught him to shoot off movement. A twenty two year old who shoots badly has had four years of people trying to fix it. A nineteen year old who shoots badly has had, in many cases, one season of anyone paying real attention.

The practical consequence is that the variance on a young prospect is not only wider in both directions. It is skewed, and the downside is less informative than it looks. I would rather draft the eighteen year old with two obvious holes than the twenty two year old with none, and I would do it even when the older player's model output is higher, because I know where the older player's ceiling is and I do not know where the younger one's is.

There is a limit to this, and it is worth naming. Some flaws do not move with age. Feel for the game, the ability to see a pass before it exists, the willingness to be in the right place when nothing is happening. Those are close to fixed. Shooting moves. Strength moves. Feel mostly does not, and a model that discounts a bad shooting season should not extend the same mercy to a season of bad decisions.