Motion-blurred skater in a New Jersey Devils sweater

Breakdown · COL at NJD

Devils edge the Avalanche in overtime

Final / OTSun, Oct 26

COLColorado Avalanche3 NJDNew Jersey Devils4

The verdict · stolen

The Devils stole one. They lost the expected-goals battle 3.40–2.87 and won anyway.

COL 3 · NJD 4

◆ By the ModelJul 11, 1:19 PMxG: xgboost-0.758Edited for clarityDrafted from NHL play-by-play by Backcheck's model, worded for clarity by an AI editor that may not change a number. Nobody at Backcheck was at the rink. How this works

The Devils won 4-3, but the Avalanche's underlying numbers were better — expected goals favored Colorado 3.40 to 2.87, a 54-46 split. New Jersey overperformed their xG by 1.13 goals, converting low-danger chances that should have stayed unconverted.

The Devils scored twice from spots the opposing goalie needs to have. Their defense held shot attempts even (24-24) and high-danger chances nearly so (4-5), preventing Colorado's superior generation from translating into more goals.

Goaltending was decisive. T. Miner (COL) stopped 20 of 24 — a save rate 1.13 goals worse than expected — while J. Allen (NJD) posted 0.875 on 21 of 24.

Both teams found their edge on the rush: the Devils scored 3 goals on 28 chances, the Avalanche 2 on the same 28. New Jersey's rush efficiency proved decisive. Meanwhile, Colorado created 4 rebound opportunities but couldn't finish any, wasting the fruit of their net-crashing approach.

The game turned at 01:53 of overtime when J. Hughes fired a wrist shot from the slot — the moment that separated a team that generated less from one that finished more.

The breakdown in twelve frames

Swipe it

Every number is from this page

Final · OT · Oct 26

COL

NJD

Devils edge the Avalanche in overtime

Chances 3.4 – 2.9Did they deserve it? →

Did they deserve it?

The Avalanche deserved better.

Expected goals · chance value

COL

3.40

NJD

2.87

54% of the chances46%

The board read COL 3, NJD 4. The chances read COL 3.4, NJD 2.9.

Our xG model · NHL play-by-play · all situations

Luck check

NJD

+1.1

goals above expected

NJD2.87 expected · 4 scored

Expected

Scored

COL3.40 expected · 3 scored

Expected

Scored

Both sides finished close to what their chances were worth. This one was earned on the ice.

PDO = sh% + sv% · drifts to 1.000 over a season

Every goal, in order

P1 09:09NJDA. Gritsyuk0–1
P1 09:48NJDJ. Hughes0–2
P1 12:55COLV. Nichushkin1–2
P1 17:03COLN. MacKinnon2–2
P2 12:51NJDC. Brown2–3
P3 13:55COLB. Nelson3–3
OT 01:53NJDJ. Hughes3–4

NHL play-by-play · chance value from our xG model

Where the shots came from

COL

46 shots

Final

87 shots

NJD

41 shots

COL NJD GoalBigger dot · more dangerous look

shots · 3 in

shots · 4 in

Every unblocked shot · press play to replay it

Chances by danger

COLNJD

1 / 5

1.10 xG

High danger

1 / 4

0.98 xG

2 / 14

1.56 xG

Medium

1 / 10

1.25 xG

0 / 52

0.74 xG

Low

2 / 46

0.65 xG

High-danger looks: COL 1 of 5, NJD 1 of 4.

Danger tiers by our xG model · high ≥ 20% chance

In the net

J. Allen

+0.4

NJD · 21/24

D-

T. Miner

−1.1

COL · 20/24

The numbers, checked

BACKCHECK

Devils edge the Avalanche in overtime

A model, not a newsroom. Every number on these frames comes from the NHL's play-by-play and our own models. Nothing was watched, guessed or heard from anyone.

Read the full breakdown

backcheckreport.com

01 / 08 · Final

Every goal, in order

How it happened

NHL play-by-play

1stshots 10–10

09:09NJDA. Gritsyuk(T. Meier)0–1
09:48NJDJ. Hughes(S. Nemec, B. Dillon)0–2
12:55COLV. Nichushkin(C. Makar, V. Olofsson)1–2
17:03COLN. MacKinnon(A. Lehkonen, D. Toews)2–2

2ndshots 9–3

12:51NJDC. Brown(S. Nemec)2–3

3rdshots 5–9

13:55COLB. Nelson(R. Colton, S. Malinski)3–3

OTshots 0–2

01:53NJDJ. Hughes(S. Nemec)3–4

score reads COL–NJD

Source · NHL play-by-play (api-web.nhle.com)

Process vs result

The numbers

XGBoost xG · NHL play-by-play

2.87

NJD chances

Goals their shots were worth

+1.1

Finishing

Goals above what those chances were worth

1.024

NJD puck luck

Above 1.000 — the bounces went their way

45.8%

Shot attempts

Under half — they were outshot

Split by club

Who controlled the game

3.40Expected goals2.87

COL 54% of the chance quality

24Shots on goal24

5High-danger chances4

71All attempts (Corsi)60

Adjusted for score: COL 56% · NJD 44%

Faceoffs · NJD won46%off50%neu58%def

Source · NHL play-by-play · XGBoost xG

How the game was won

The danger ladder

shotsgoalsexpected

High dangerslot & crease

COL1 / 5 · 1.10 xG

NJD1 / 4 · 0.98 xG

Medium dangermid-range

COL2 / 14 · 1.56 xG

NJD1 / 10 · 1.25 xG

Low dangerperimeter & point

COL0 / 52 · 0.74 xG

NJD2 / 46 · 0.65 xG

NJD went 2-for-46 from low danger on 0.65 expected goals — 1.4 above expected.

Source · Backcheck xG model · NHL play-by-play

The luck layer

Finishing vs expected

expectedactual

COL

3 G · 3.40 xG · -0.4

NJD

4 G · 2.87 xG · +1.1

012345

COL finished -0.4 against expected, NJD +1.1.

Source · Backcheck xG model

Shot types · rush and rebound chancesopen +

What they shot — and what went in

Shot diet

goalsshots

COL

WRIST0/23

SNAP3/7

TIP IN0/7

SLAP0/6

BACKHAND0/2

WRAP AROUND0/1

OTHER0/25

NJD

WRIST3/27

SNAP1/8

TIP IN0/1

SLAP0/2

BACKHAND0/3

WRAP AROUND0/0

OTHER0/19

COL

NJD

Off the rush

2 G on 28

2.40 xG

3 G on 28

2.08 xG

Off rebounds

0 G on 4

0.58 xG

0 G on 1

0.26 xG

Wrist shots carried the volume: COL went 0-for-23, NJD 3-for-27. NJD scored 3 of theirs off the rush.

Source · Backcheck xG model · NHL shot types

New to these numbers?

Two things happened in this game: who created the better chances, and who put the puck in the net. They are not the same thing, and this page separates them on purpose.

The one to watch is expected goals— the best single read on who deserved to win. When it disagrees with the scoreboard, that disagreement is usually the story. Nothing here is a projection; every figure is computed from this game's play-by-play. Full glossary →

xG: Expected goals. Every shot scored by how often that exact look goes in. 2.5 xG means the chances were worth about two and a half goals.
HD: High-danger chances — shots from the slot and the crease, where shots actually become goals.
Corsi: Every shot attempt, including the ones that missed or got blocked. A rough read on who had the puck.
Adj.: Score-adjusted. Teams that trail shoot more and teams that lead sit back, so this corrects the attempt count for the scoreboard.
GSAx: Goals saved above expected. A positive number means the goalie stopped more than an average goalie would have on those shots.
PDO: Shooting % plus save % — a luck gauge. It sits near 1.000 over a season, so a game far from it usually means the bounces decided it.

Shot by shot

In the net

Graded on goals saved above expected

T. Miner

COL · 20/24 · .833

-1.1

saved above expected

D-

the night

By dangerG/attxGsaved

High danger1/41.0-0.0

Medium1/101.3+0.3

Low2/460.7-1.4

A+ ≥ +2.5 saved above expected · A ≥ +1.5 · B+ ≥ +0.75 · B ≥ +0.25 · C ≥ −0.25 · D ≥ −0.75 · F below −1.5

J. Allen

NJD · 21/24 · .875

+0.4

saved above expected

the night

By dangerG/attxGsaved

High danger1/51.1+0.1

Medium2/141.6-0.4

Low0/520.7+0.7

A+ ≥ +2.5 saved above expected · A ≥ +1.5 · B+ ≥ +0.75 · B ≥ +0.25 · C ≥ −0.25 · D ≥ −0.75 · F below −1.5

87 shots · press play to watch the game back

Where the shots came from

NHL play-by-play coordinates

COL

46 shots

Final

87 shots

NJD

41 shots

COL NJD GoalBigger dot · more dangerous look

Each club shoots toward its own end — COL left, NJD right — the way you would watch it. A dot's size is how dangerous the look was, from distance and angle; it is a proxy for shot quality, not the xG model the article is written from. Hover a dot for the shooter and the clock.

Ranked from the box score

Three stars

Points first

Three stars · computed, not voted

S. Nemec

#17 · D · NJD

0 G · 3 A · 2 SOG · 23:35 on ice · +2

J. Hughes

#86 · C · NJD

2 G · 0 A · 2 SOG · 22:43 on ice · +1

C. Makar

#8 · D · COL

0 G · 1 A · 2 SOG · 27:38 on ice

Ranked from the box score — points first · Backcheck

What the model said beforehand

Before the game

Written Sun, Oct 26

Open the pre-game read

Written from the numbers

The read

Edited for clarity

◆ By the ModelJul 11, 1:16 PMEdited for clarityDrafted from NHL play-by-play by Backcheck's model, worded for clarity by an AI editor that may not change a number. Nobody at Backcheck was at the rink. How this works

The model projects Colorado at 55% win probability over New Jersey. The 2.8–2.5 expected-goals spread is tight enough to justify the coin-flip metaphor.

Colorado owns the underlying play. The Avalanche generate 57.0% of 5v5 expected goals to New Jersey's 50.0%, and they control possession at 57.0% Corsi. Territorial dominance is clear; the actual quality gap is thinner.

The regression ledger cuts both ways. New Jersey's 0.971 PDO means pucks are bouncing the wrong way—they're due for positive regression. Colorado's 1.021 PDO is the flip side: pucks are bouncing favorably, and regression downward is expected. The tightness of the model reflects the uncertainty around which team's mean reversion matters more.

Monitor S. Wedgewood's status. He's on a back-to-back; New Jersey may start a backup or deploy a fatigued Wedgewood, either outcome shifting the margin.

Colorado's attack centers on Artturi Lehkonen, who generates 1.11 expected goals per 60 minutes with 18 goals and 22 assists. He's scored 6 from high-danger chances at a sustainable 14.9% shooting rate. Ross Colton matches Lehkonen's 1.10 xG/60 but is underperforming: 7 goals on 15 assists with just 1 high-danger goal and a 4.9% shooting rate. Positive regression should follow if he keeps getting touches.

New Jersey counters with Timo Meier at 1.08 xG/60, slightly below Colorado's top two, with 13 goals and 15 assists. Meier has 3 high-danger goals. Jack Hughes carries the secondary scoring load at 0.88 xG/60 with 18 goals and 17 assists, but his 11.5% shooting rate suggests unsustainable luck baked into that total.

Defensively, Colorado's weak link is Nazem Kadri. New Jersey flags Johnathan Kovacevic in that role. In a game this close, which pair plays tighter may settle it.

The xG variance is the story. Colorado's edge is real but modest. New Jersey's pending PDO regression is the wildcard.

What the model said beforehand

The model's lean

COL 55%

55%

COL

Model favorite

55%

45%

COL50NJD

COL by 10 points of win probability — a modest lean. Projected score COL 2.8–2.5 NJD; 18% chance it's tied after 60.

Source · Backcheck Poisson projection · season scoring rates

Season rates · 5v5

Tale of the tape

NHL stats API · our xG

COL← edge

edge →NJD

57.0%50.0%

57.0%51.0%

3.032.45

2.302.50

2.631.66

1.562.23

1.0210.971

COL holds 6 of 6 process categories — and the luck row (1.021 vs 0.971) says COL has been getting the bounces, which tends not to last.

Source · NHL stats API + our xG · 5v5

Form · Last 10

COL

5-2-3

+7 GD · 1d rest

NJD

8-1-1

+13 GD · 2d rest

1.021

COL PDO

Luck gauge · drifts to 1.000

0.971

NJD PDO

Luck gauge · drifts to 1.000

What decides it

Keys to the game

Where the edges live

The Avalanche own the underlying play — 57.0% 5v5 xG vs 50.0%.
The Avalanche drive Corsi at 57.0% — expect territorial tilt.

Regression watch

The Devils' PDO is 0.971 — running cold, due for positive regression.
The Avalanche's PDO is 1.021 — running hot, expect some regression.

Things to watch

The Avalanche own the underlying play — 57.0% 5v5 xG vs 50.0%.
The Avalanche drive Corsi at 57.0% — expect territorial tilt.
S. Wedgewood is on a back-to-back — watch for a backup start or a tired starter.
The Devils' PDO is 0.971 — running cold, due for positive regression.

Last game's lineup · confirmed at warm-up

Lineups

NHL box score

COL

Expected in goal

S. Wedgewood

back-to-back.842 last start

Backup Scott Wedgewood · +23.1 GSAx this season

Back-to-back: the backup often gets this one. Treat the starter as unconfirmed.

No completed game on file for this club yet.

NJD

Expected in goal

J. Allen

2 days rest.941 last start

Backup Jacob Markstrom · -11.9 GSAx this season

No completed game on file for this club yet.

Lineups are the players who dressed in each club's last game, with their minutes; clubs confirm tonight's at warm-up. Every player this preview names was in that lineup.

Computed danger · scouting

Players to watch

NHL stats API · our xG

COL

Artturi Lehkonen18G 22A

1.11 xG/60 · 6 HD · 14.9% sh

Ross Colton7G 15A

1.10 xG/60 · 1 HD · 4.9% sh

NJD

Timo Meier13G 15A

1.08 xG/60 · 3 HD · 6.7% sh

Jack Hughes18G 17A

0.88 xG/60 · 1 HD · 11.5% sh

Keep going

Where to next

Back to the front page