xJawn NHL expected goals

← Players

Tyson Kozak

C · Buffalo Sabres · career →

9 GP · 81 min 5v5
2026-27 · 5v5
On the ice
Adjusts with rink filters
on-ice for◀ 5v5 offense
on-ice against5v5 defense ▶
How to read this map

One rink, both ends while Tyson Kozak is on the ice: what the team generates on the left, what it concedes on the right. Every attempt is a Gaussian bump of height = its model xG, summed per 60 minutes of ice time (a player's exposure is time, not games). The controls switch the surface (xG heatmap ↔ vs league, the league baseline per team-hour at the same strength) and the markers: × goal · • on net · ◦ missed · small • blocked — and own shots narrows the left end's markers to attempts Tyson Kozak took.

The page opens at 5v5; empty-net attempts are excluded throughout. On-ice membership comes from the shift charts; attempts without usable shift coverage are absent from both ends. Blocked-shot membership cannot be cross-checked against the play-by-play, so treat blocked-marker detail as softer than the rest.

Observed production

5v5 · on-ice · season to date
for against
axis spans ±235% of the forward cohort average · centre = average · right = more
Goalson-ice
2 9th pctl3 9th pctl
Shotson-ice
27 4th pctl20 0th pctl
Attemptson-ice
64 4th pctl72 3rd pctl
xGon-ice
2.0 2nd pctl3.0 3rd pctl
xGF%share of on-ice xG
40.0%13th pctl
0% 50% = even split · gray = cohort middle 50% / 90% 100%

Percentiles rank against the other forwards over the display floor at the selected strength — a non-strength filter (rush, period, …) moves the bars but keeps that strength's spread as the backdrop.

Rolling on-ice xGF%

5v5 · last 10 games
Per-game chance share while on the ice, smoothed. drag the slider to set the window
The model

Model estimates

🔒 as of Jun 14, 2026
Last season's ratings and goals accounting (2025-26) — 2026-27 is not rated until it has enough games to refit. The observed rows above are 2026-27.
creating   preventing   creates both  ·  paler = special teams  ·  solid = on-ice RAPM · hatched = skill rating  ·  ▽ = the total
Shooting
Conversionall shooter stages
outside the PP -0.21 goals vs average
-0.2247th pctl
Offense
On-ice impactRAPM · xGF impact
-1.8415th pctl
Defense
On-ice impactRAPM · xGA, flipped
-0.2743rd pctl
Special teams
Power playST RAPM · 3-season window
-0.0548th pctl
Penalty killST RAPM · season-resolved
-0.1017th pctl
Disciplinedrawn − taken
+0.5860th pctl
GAAgoals above average, this season
Offense impact -1.84 goals vs averageDefense impact -0.27 goals vs averageConversion · outside the PP -0.21 goals vs averagePenalty kill -0.10 goals vs averageDiscipline +0.58 goals vs average
-1.9123rd pctl
grouped by game state · 5v5 → power play → kill
Not shown — Blocking -0.00 — included in the total.
−18.960 = league average+18.96

Ranked against forwards; the band behind each bar is that cohort's spread (p5–p95, middle half brighter, dashed median), so where a bar lands against its band is the rank — the ordinal beside it makes that exact. Zero is whichever baseline the toggle selects — the league-average skater, or replacement level — and every row, its band and its ordinal are all drawn against that one baseline, so the bar and the number can never describe different things. Each row's bar grows from zero, so its end IS its value; the composite is the rows stacked, running both ways from zero, so only it carries a ▽ to mark the net. The rows sum to the total with nothing left over. The per-shot rating is quality only: it says nothing about how many attempts a player generates or faces, which is what the attempts figure is for.

Faceoff skill

🔒 as of Jun 14, 2026
Faceoff skillfitted rating · not in GAA/GAR
53.2%50th pctl
A per-draw skill rating from the faceoff model, shown as the win probability against a league-average taker on a neutral dot. It is not priced in goals and is not part of GAA or GAR: the value of a won draw is shared by everyone on the ice, so this stays a rating.
−0.690 = an average taker+0.69
Context

Usage & context

season to date
The job the coach handed Tyson Kozak, not how well it went. Not a modeled component: each is its own metric (minutes, a rate, a TOI share, or opponents'/teammates' net xG-RAPM), shown with a cohort percentile only where one exists, never added together. Not a grade. The deployment rows are measured over the same 5v5 stints the isolate is fit on (regular season + playoffs); ice time and the penalty rates are all situations, every game — penalties create non-5v5 time, so a 5v5 denominator would inflate every rate.
Time on iceall situations · min per game
11:11per game
Penalties takenminor penalties / 60
1.19per 60 · all situations
Penalties drawnminor penalties / 60
0.00per 60 · all situations
Zone startsOZS% · O-zone start share
8%offensive / (off + def)
Score stateTOI by game state
trailtied 49%lead 40%
often tied
Competition (QoC)opponents' net xG-RAPM
+0.021xG/60
Teammates (QoT)teammates' net xG-RAPM
-0.080xG/60
On-the-flyshare of 5v5 TOI
65%of 5v5 TOI
0cohort average100 pctl

A skater absent from the fit counts as exactly cohort average (0) in QoC/QoT — the same claim ridge shrinkage makes. Penalty rates are descriptive only: taken-rate includes defensively correct penalties, and neither side is opponent- or deployment-adjusted. Nothing here is priced in goals.

Full table · all numbers
FamilyStatValue (current slice)
Observed ◆ filtered Goals on-ice for / against 2 / 3
Observed ◆ filtered xG on-ice for / against 2.0 / 3.0
Observed ◆ filtered Shots on-ice for / against 27 / 20
Observed ◆ filtered Attempts on-ice for / against 64 / 72
Observed ◆ filteredxGF% 40.0%
Modeled 🔒 season Conversion — all shooter stages -0.22 47th pctl
Modeled 🔒 season ↳ Conversion — 54 outside the PP 14th pctl
Modeled 🔒 season ↳ Conversion — 0 on the PP 7th pctl
Modeled 🔒 season ↳ Conversion — per-shot skill 48th pctl
Modeled 🔒 season On-ice impact — RAPM · xGF impact -1.84 15th pctl
Modeled 🔒 season On-ice impact — RAPM · xGA, flipped -0.27 43rd pctl
Modeled 🔒 season Power play — ST RAPM · 3-season window -0.05 48th pctl
Modeled 🔒 season Penalty kill — ST RAPM · season-resolved -0.10 17th pctl
Modeled 🔒 season Discipline — drawn − taken +0.58 60th pctl
Modeled 🔒 season GAA -1.91 23rd pctl
ContextTime on ice — all situations · min per game 11:11 (12th pctl)
ContextPenalties taken — minor penalties / 60 1.19 (76th pctl)
ContextPenalties drawn — minor penalties / 60 0.00 (3rd pctl)
ContextZone starts — OZS% · O-zone start share 8% (0th pctl)
ContextScore state — TOI by game state trail 11%, tied 49%, lead 40%
ContextCompetition (QoC) — opponents' net xG-RAPM +0.021 (63rd pctl)
ContextTeammates (QoT) — teammates' net xG-RAPM -0.080 (14th pctl)
ContextOn-the-fly — share of 5v5 TOI 65%

Observed values reflect the active rink filters; modeled values are fixed at their full-season fit.

On-ice numbers pair the regular-season shot set with shift-chart membership; skaters and goalies under the display floors are fit but not shown. Modeled rows come from the value artifact keyed to RAPM fit rapm__0cd6d075 on nested_structure__nested_structure_xg__d2ec2662 (the joint-xg structural baseline), with ratings from joint__baseline_xg__f3e3111e (as of 2026-06-14) — separate surfaces from the box-score xG model badged in the footer.