Pricing the intangibles

Most teams carry a forward whose numbers say third line and whose coach keeps sending him out in the last two minutes. He goes to the net, takes the cross-check, finishes the hit along the wall, and his linemates' numbers are a bit better with him than without him. His own points and expected goals are ordinary. The people who played with him tend to mention him first.

Grit & GF starts from the guess that this player exists in larger numbers than the box score suggests, and that he leaves traces in the play-by-play and shift charts if you look past the puck carrier.

The hunch

Public hockey analytics has gotten good at one question, which is who drives shots and goals. Corsi, expected goals and goals above replacement answer it well. They mostly ignore what happens away from the shooter, because that is hard to count and harder to put a value on.

We think some of it matters. A forward wins the wall battle three passes before the goal. A defenceman holds a tight enough gap that the other team dumps the puck in instead of carrying it. A winger's two linemates both score more when he's on the ice. None of it shows up next to anyone's name on the scoresheet.

Where we look

The rating has five parts, one for each place we expect that player to show up.

Physicality covers hits, blocks, takeaways and the penalties he draws from opponents who've had enough of him. GF is about where his goals come from, and gives credit for chances taken from the crease and the slot, where a shooter gets hit for his trouble. Intimidation compares how the other team plays with him on the ice and off it in the same games. Chemistry measures his linemates with him and without him. Clutch looks at his team in the last period of one-goal games against his own usual rate.

All five are built from the league's play-by-play and shift charts and recomputed each morning. Thin samples get pulled hard toward average, since a big week in October tells you very little. The output is a 0 to 100 rating against his position, plus a goals and wins value for the parts that have shown a link to scoring.

What would prove us wrong

There are three ways this could turn out to be wrong, and we're checking all of them.

The first is repeatability. A part that doesn't carry over from one season to the next is measuring noise. Physicality carries over about as well as anything in hockey. Clutch, so far, barely does, and we expect most of it to turn out to be luck. It stays in at a small weight because the idea that some players rise in big moments is all over the sport, and we would rather measure it than wave it off.

The second is that it could just be a talent ranking with extra steps. Good players make everything around them look better, so talent leaks into chemistry and intimidation, and we try to strip it back out. If the top of the board were all stars, the rating would be telling us nothing new.

The third is the scoreboard. If none of this ever shows up in goals, it stays a description of a style. That is why hits and blocks count toward the rating but are worth zero goals in the value column. We suspect they're worth something, and they'll get a number when the data supports one.

Who it likes

The players near the top tend to look alike: lots of minutes in the hard areas, lots of contact, chances from close in, and linemates who do better around them. A few are stars. Most are the kind of player opposing fans boo by name.

For a bettor it's context. When a coach breaks up a line, chemistry shows whose night got harder. When a team loses a player with a big intimidation score, it's worth watching the other side's shot totals for a few games. It doesn't pick winners by itself, and we'd be wary of anyone claiming a grit rating could.