Individual Expected Goals, the ixG Hockey Stat

ixG is the original expected-goals idea pointed at one shooter. When Dawson Sprigings and Asmae Toumi published their model at Hockey-Graphs in October 2015, they tested it on individual players as well as teams, and found that a player's ixG per 60 minutes did better than his shot attempts or his shooting percentage at estimating his future scoring. The finding held at the player level as well as the team level, which is what made ixG worth tracking on its own.

The 'i' is the same prefix that marks a player's own numbers across the public sites. Evolving-Hockey's tables list iCF (every attempt he takes), iFF (his unblocked attempts) and iSF (his shots on goal), and define ixG as the total expected goals of his iFF shots. MoneyPuck carries the same number in its data files under the column I_F_xGoals.

Every figure on this page uses MoneyPuck's model, the one described on the expected goals page. A player's ixG here can differ from his total on Evolving-Hockey, which trains its own model on its own choice of features, so you shouldn't mix the two sources in one comparison.

Take every unblocked attempt a player takes himself, price each one with the expected-goals model, and add them up. MoneyPuck's model prices each shot on these inputs:

• Distance to the net and the angle of the shot
• The shot type
• The last event before the shot, where it happened and how many seconds earlier
• How many skaters each side had, and whether the net was empty
• Whether the shooter was on his off wing

Here's a worked example, with invented round numbers. A winger takes 200 unblocked attempts in a season and the model prices them at an average of 0.09 each, so his ixG is 18.0. Score 22 and he's four goals above expected. Score 14 and he's four below, on exactly the same chances, and his ixG hasn't moved at all.

Blocked attempts don't count, because the NHL logs a block where it happened and leaves no honest record of where the shot came from. Totals on this site are all situations unless a column says otherwise, and power-play time pads them, which is why the five-on-five rate per 60 is the version to compare players on.

ixG measures chance creation from the shooter's side, and it rewards both volume and location. Among the 327 forwards who played at least 60 games last season, the median generated 17.3 in all situations and the top 10% cleared 29.7. McDavid led at 43.6, ahead of Nathan MacKinnon and Kirill Kaprizov. Defensemen live on a different scale, since their shots carry less than half the expected value of a forward's at the median (0.039 per unblocked attempt against 0.087), and the median regular defenseman generated 5.1 while Jakob Chychrun led them at 18.8.

DeBrincat was the only Red Wing in the top 15, with Dylan Larkin 18th. The drop after that is steep. Copp, Lucas Raymond, James van Riemsdyk and Emmitt Finnie all landed between 19.2 and 19.8, a couple of goals above the median forward, which tells you Detroit's chance creation ran through two players. Everyone else on the roster sat at or below the forward median, defensemen included, and that's a thin base for a team that finished the season short of goals.

Why lead with ixG over goals? It holds still. Among the 260 forwards who played 60 games in both of the last two seasons, ixG correlated at 0.77 from one year to the next, goals at 0.73, and the gap between them at only 0.38. Per 60 minutes the difference widens, 0.76 for ixG against 0.62 for goals. Sprigings and Toumi's 2015 tests pointed the same way, with ixG rates beating shot attempts and shooting percentage at estimating what a player would score next.

Reading a slump

Is a scorer's slump about his chances or his finishing? ixG answers the first half. Copp scored 9 goals last season on 19.8 ixG, the fifth-largest shortfall among the league's 940 skaters. His chances went up. He'd generated 12.3 ixG the season before, so you can put last season's drop in goals on the finishing, which is the half of scoring that swings the most.

Separating volume from danger

ixG is volume times danger, and the two pieces tell different stories. DeBrincat's total came from 443 unblocked attempts at 0.083 expected goals each, below the 0.087 median for forwards with 150 or more attempts. Zach Hyman got 35.2 from 245 attempts at 0.144 each, the highest rate among those forwards. Nearly the same total, opposite routes. Both worked. Hyman scored 31 goals and DeBrincat scored 41, so neither shape is the wrong one, and the split tells you which kind of shooter you're projecting.

Comparing players on a level field

Totals reward ice time and power-play minutes, so analysts compare forwards on five-on-five ixG per 60. The median forward with 500 five-on-five minutes generated 0.70 last season. DeBrincat ranked 53rd of 385 at 0.88, and you'll find a surprise a little further down the list. Marco Kasper, whose goals fell from 19 to 9, still created five-on-five chances at a higher rate than Larkin did (0.81 per 60 against 0.66).

It ignores who's shooting

The model prices a McDavid wrister and a fourth-liner's from the same spot the same way, because MoneyPuck's base model doesn't factor in shooting skill. That's the job of goals above expected.

It can't see the pass

The NHL's play-by-play doesn't record passes, so a cross-crease feed looks to the model like any other shot from that spot. The player who made the pass gets nothing on this stat, which is why playmakers look better on on-ice expected goals than on ixG.

Rebounds stack up

A player who follows his own shot can collect two attempts' worth of ixG on one sequence. MoneyPuck publishes a flurry-adjusted version that gives less credit for rebound chances than for first shots.

Blocked shots vanish

Attempts that get blocked add nothing. A shooter who fires through traffic, defensemen above all, generates less ixG than his attempt totals suggest, and you'd need his individual shot attempts to see the full volume.

Ice time pads the totals

A first-line winger with power-play time piles up ixG that a third-liner never gets the minutes to match, so compare totals only between players in similar roles and lean on the per-60 rate for everyone else.