Talent vs Production: Why the Box Score Misleads
2026-08-16 · analytics · methodology · expected goals · finishing
Two wingers score 20 goals; only one of them will do it again. The gap between what a player produced and the process underneath it is the most useful idea in hockey analytics - and it is the split our expected-goals and finishing models are built to make.
- A player's production (goals and points) and their talent (the process that generates chances) are not the same number — production bakes in finishing variance, goaltending, and plain luck.
- Underlying metrics predict future goals better than past goals do. Chance quality and shot volume are far more repeatable season to season than the goals themselves.
- Finishing is a real skill, but a noisy one. Most shooters regress toward their expected goals; only a small set of elite finishers beat it year after year.
- This is why production can mislead and process is the signal — and it is exactly the split our expected-goals model and the finishing component of GAR are built to make.
Two wingers each score 20 goals. One did it on the back of a torrid shooting streak that will not repeat; the other generated so many quality chances that 20 was almost the floor. The box score records the same number for both. Everything that matters about what happens next is in the part the box score throws away.
Production is the outcome; talent is the process
That gap — between what a player produced and what a player is actually doing to produce it — is the single most useful idea in hockey analytics. Production is the outcome. Talent, in the measurable sense we mean here, is the repeatable process underneath it: the chances created, how dangerous they were, and how much of the play a skater drives when he is on the ice.
The practical test is repeatability. If a number measures durable skill, a player who is good at it this season should be good at it next season. Shot generation and chance quality clear that bar comfortably. Shooting percentage largely does not.
Why this changes how you read a stat line
Be suspicious of any evaluation that leans hard on last year's goal total. Goals are downstream of finishing, and finishing barely repeats. The player who scored 40 on a 19% shooting rate is very likely to score fewer next year — not because he got worse, but because the bounces regress. Meanwhile the player quietly generating chances at even strength is doing the thing that actually carries forward.
None of this makes finishing fake. Some players really are better shooters, and across a career it shows. But within a single season the signal is buried under variance, which is why our model keeps only half of a player's finishing and regresses the rest toward what his chances say he should have scored.
Frequently Asked Questions
What is the difference between production and talent in hockey?
Production is the outcome - goals and points. Talent, in the measurable sense, is the repeatable process underneath: the chances created, their quality, and the share of play a skater drives. Production bakes in finishing variance, goaltending faced, and luck; the process does not.
Do underlying metrics predict future goals better than past goals do?
Yes. Chance quality and shot volume are markedly more repeatable season to season than the goals themselves, so a player’s expected goals are a better guide to next season’s scoring than his actual goals were.
Is finishing a real skill?
It is real but noisy. Most shooters regress toward their expected goals; only a small set of elite finishers beat it year after year. Within one season the signal is buried under variance, which is why Hockey Alchemy keeps only half of a player’s finishing in his GAR total.
More from The Lab
Which Hockey Stats Are Skill, and Which Are Luck?
Line up every player's value in a category this season against next season and you get a brutal skill-vs-luck test. Staying out of the penalty box repeats more than four times better than finishing - and it changes how you should read a stat line.
Why Public Hockey Analytics Is Imperfect
An honest accounting of where public hockey analytics hits a wall — the data ceiling (the feed records no passes, tracking drifts, private camera models see more) and the bigger problem now: a discourse that shares cards without meaning, rewards certainty over nuance, and stopped teaching. With the Zach Hyman case as the tell.
How Good Is Our WAR Model? We Ran Four Tests and Lost Two
We tested our WAR model against HockeyStats and Evolving Hockey on the four questions an honest player-value model has to answer, matched to each benchmark's published protocol. Summed to a team it loses both tests - it forecasts next season worse than simply reusing last season's standings. Measured per player it wins both, repeating at 0.77 against 0.62 and 0.46.