Scouting Value
How much better (+) or worse (−) than the scouting consensus each scout and NHL team drafted, judged against a position-fair value measure — a defenseman against defensemen, a forward against forwards, on their best sustained stretch — that credits scoring, even-strength defense, and award recognition, not just cumulative WAR. 0 is dead average.
What "value" counts (and what it doesn't). A player is judged position-fair (a defenseman against defensemen) and on his best sustained stretch, crediting scoring — against his position and in absolute terms — even-strength offense AND defense, and award recognition (Selke, Norris and the like, peak and sustained). That deliberately lifts the two-way and defensive players a box score alone misses, and docks the offense-only. The measure is fit to a reference tier list (held-out ρ=0.917): a disclosed target, not a single-metric endorsement — and the top-of-board forward/defense order is contested judgment that does not affect the finding.
Two lenses, one number. A bright cell is a descriptive grade of a single call; the "no scout and no franchise beats the field" verdict is the inferential test against a permutation null. A vivid right call can glow without its scout clearing the null — a real call, not a proven skill.
Every edge carries its p. Even on the class-cut most favourable to him, the top scout's raw edge reaches only +0.42 — and still doesn't clear the null (p = 0.20). A raised edge that's still inside chance is not skill.
| Scout | |||
|---|---|---|---|
| ▸Craig Button · TSN | 6/9 | +0.15 | +0% |
Year by year — each draft’s raw edge (6 of 9 above the field) 2013 −0.322014 +0.442015 +0.572016 +0.262017 +0.282018 +0.152019 +0.642020 −0.262021 −0.03 Each number is Craig Button’s edge in that one draft — green above the field, red below. The unshrunk average across them is +0.15; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Cam Robinson · Dobber Prospects / EliteProspects | 3/4 | +0.04 | +0% |
Year by year — each draft’s raw edge (3 of 4 above the field) 2017 +0.232018 +0.172019 +1.032020 −0.62 Each number is Cam Robinson’s edge in that one draft — green above the field, red below. The unshrunk average across them is +0.04; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸The Draft Analyst · The Draft Analyst | 2/4 | +0.04 | +0% |
Year by year — each draft’s raw edge (2 of 4 above the field) 2016 −0.072017 −0.092018 +0.072019 +0.23 Each number is The Draft Analyst’s edge in that one draft — green above the field, red below. The unshrunk average across them is +0.04; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Future Considerations · FC Hockey | 3/7 | −0.03 | +0% |
Year by year — each draft’s raw edge (3 of 7 above the field) 2013 −0.372014 −0.022015 +0.092016 +1.732017 +0.282018 −0.092019 −0.42 Each number is Future Considerations’s edge in that one draft — green above the field, red below. The unshrunk average across them is −0.03; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Hockey Prospect · Hockey Prospect (Black Book) | 4/6 | +0.07 | +0% |
Year by year — each draft’s raw edge (4 of 6 above the field) 2013 +0.182014 +0.792016 +0.252017 −0.302018 +0.042019 −0.21 Each number is Hockey Prospect’s edge in that one draft — green above the field, red below. The unshrunk average across them is +0.07; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸ISS Hockey · International Scouting Services | 2/7 | −0.28 | +0% |
Year by year — each draft’s raw edge (2 of 7 above the field) 2013 −0.182014 −0.752015 −0.352016 +0.082017 −0.532018 +0.042019 −0.22 Each number is ISS Hockey’s edge in that one draft — green above the field, red below. The unshrunk average across them is −0.28; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Bob McKenzie · TSN (Scout Consensus) | 4/9 | −0.09 | +0% |
Year by year — each draft’s raw edge (4 of 9 above the field) 2013 −0.582014 −0.372015 −0.902016 +0.212017 +0.052018 +0.362019 −0.712020 +0.242021 −0.17 Each number is Bob McKenzie’s edge in that one draft — green above the field, red below. The unshrunk average across them is −0.09; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸McKeen's Hockey · McKeen's Hockey | 3/8 | −0.05 | +0% |
Year by year — each draft’s raw edge (3 of 8 above the field) 2013 −0.372014 +0.072015 −0.202016 −0.092017 −0.432018 +0.352019 −0.182020 +0.38 Each number is McKeen's Hockey’s edge in that one draft — green above the field, red below. The unshrunk average across them is −0.05; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Chris Peters · ESPN / FloHockey / TSN | 3/4 | +0.34 | +0% |
Year by year — each draft’s raw edge (3 of 4 above the field) 2013 −0.712018 +0.452019 +0.932021 +0.44 Each number is Chris Peters’s edge in that one draft — green above the field, red below. The unshrunk average across them is +0.34; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Corey Pronman · ESPN / The Athletic | 5/9 | +0.18 | +0% |
Year by year — each draft’s raw edge (5 of 9 above the field) 2013 +1.112014 +0.182015 +0.242016 −0.042017 −0.092018 +0.002019 +0.262020 −0.042021 +0.27 Each number is Corey Pronman’s edge in that one draft — green above the field, red below. The unshrunk average across them is +0.18; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Ryan Kennedy · The Hockey News | 4/9 | −0.06 | +0% |
Year by year — each draft’s raw edge (4 of 9 above the field) 2013 +0.112014 −0.262015 +0.142016 −0.072017 −0.412018 +0.262019 −0.442020 +0.042021 −0.10 Each number is Ryan Kennedy’s edge in that one draft — green above the field, red below. The unshrunk average across them is −0.06; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Sam Cosentino · Sportsnet | 1/5 | +0.04 | +0% |
Year by year — each draft’s raw edge (1 of 5 above the field) 2013 −0.072018 −0.342019 −0.422020 +0.622021 −0.17 Each number is Sam Cosentino’s edge in that one draft — green above the field, red below. The unshrunk average across them is +0.04; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
| ▸Scott Wheeler · The Athletic | 2/3 | +0.02 | +0% |
Year by year — each draft’s raw edge (2 of 3 above the field) 2018 +0.282019 +0.042020 −0.20 Each number is Scott Wheeler’s edge in that one draft — green above the field, red below. The unshrunk average across them is +0.02; held out of sample it is +0%. A single class is a tiny sample, so the years swing hard — which is exactly why the lumpy ones get shrunk in the verdict. | |||
Raw edge is the unshrunk average across drafts — vivid, but over-optimistic: a couple of lucky drafts inflate it, which is exactly why the shrunk Score (verdict tab) pulls the lumpy ones below the steady ones. Held-out is that edge tested out of sample, and the list is sorted by it. Both still carry wide margins — nobody's separation from the field is certain. Open a scout for their full interval, or switch to can we prove it?
| Team | picks | ||
|---|---|---|---|
| PHI | 13 | +1.43 | +0.86 |
| SJS | 8 | +2.14 | +0.80 |
| WSH | 8 | +2.06 | +0.78 |
| CAR | 13 | +1.70 | +0.68 |
| ANA | 14 | +1.54 | +0.52 |
| OTT | 14 | +2.51 | +0.52 |
| DET | 14 | +0.75 | +0.34 |
| WPG | 12 | +0.45 | +0.21 |
| STL | 11 | +0.33 | +0.07 |
| COL | 11 | +0.18 | +0.03 |
| NJD | 13 | +0.07 | +0.02 |
| NYI | 11 | +0.00 | +0.00 |
| FLA | 12 | −0.02 | −0.01 |
| NSH | 12 | −0.08 | −0.03 |
| ARI | 14 | −0.15 | −0.04 |
| LAK | 12 | −0.40 | −0.14 |
| BUF | 17 | −0.35 | −0.15 |
| BOS | 12 | −0.35 | −0.19 |
| CBJ | 10 | −0.59 | −0.21 |
| EDM | 10 | −1.00 | −0.25 |
| TOR | 10 | −1.39 | −0.28 |
| TBL | 11 | −0.62 | −0.30 |
| VAN | 10 | −1.13 | −0.33 |
| DAL | 12 | −1.09 | −0.40 |
| MTL | 14 | −1.01 | −0.58 |
| MIN | 8 | −2.24 | −0.67 |
| CGY | 12 | −1.64 | −0.77 |
| NYR | 12 | −1.59 | −1.26 |
| CHI | 10 | −2.16 | −1.30 |
No sparkline, no held-out for teams. A franchise drafts a pick or two per class — too few to split into per-draft bars or an out-of-sample test. So a team gets one pooled number: its raw edge (rate-WAR, unshrunk) and the Score after the luck-correction. Mind the gap — thin samples shrink hard, the same reason lumpy scouts do. No franchise clears the field → can we prove it?
Showing the 13 scouts with 3+ years of published boards (of 30 tracked). Click a row for each step's margin.
| Scout | |||
|---|---|---|---|
| ▸Corey Pronman · ESPN / The Athletic | +0.09 | +0%within margin | +0.9 pts |
Score (γ̂)+0.0995% CI 0.02 to 0.31 · excludes 0 (in-sample, shrunk) % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points+0.9 ptsCI +0.2 to +3.1 pts · the Score on a readable scale (×10) Corey Pronman's calls beat the consensus by ~0% in-window — about +0.9 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Craig Button · TSN | +0.08 | +0%within margin | +0.8 pts |
Score (γ̂)+0.0895% CI 0.02 to 0.29 · excludes 0 (in-sample, shrunk) % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points+0.8 ptsCI +0.2 to +2.9 pts · the Score on a readable scale (×10) Craig Button's calls beat the consensus by ~0% in-window — about +0.8 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Chris Peters · ESPN / FloHockey / TSN | +0.06 | +0%within margin | +0.6 pts |
Score (γ̂)+0.0695% CI 0.03 to 0.57 · excludes 0 (in-sample, shrunk) % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points+0.6 ptsCI +0.3 to +5.7 pts · the Score on a readable scale (×10) Chris Peters's calls beat the consensus by ~0% in-window — about +0.6 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Sam Cosentino · Sportsnet | +0.01 | +0%within margin | +0.1 pts |
Score (γ̂)+0.0195% CI -0.20 to 0.27 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points+0.1 ptsCI −2.0 to +2.7 pts · the Score on a readable scale (×10) Sam Cosentino's calls beat the consensus by ~0% in-window — about +0.1 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Scott Wheeler · The Athletic | +0.01 | +0%within margin | +0.1 pts |
Score (γ̂)+0.0195% CI -0.15 to 0.21 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points+0.1 ptsCI −1.5 to +2.1 pts · the Score on a readable scale (×10) Scott Wheeler's calls beat the consensus by ~0% in-window — about +0.1 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Hockey Prospect · Hockey Prospect (Black Book) | +0.01 | +0%within margin | +0.1 pts |
Score (γ̂)+0.0195% CI -0.16 to 0.20 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points+0.1 ptsCI −1.6 to +2.0 pts · the Score on a readable scale (×10) Hockey Prospect's calls beat the consensus by ~0% in-window — about +0.1 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸The Draft Analyst · The Draft Analyst | +0.01 | +0%within margin | +0.1 pts |
Score (γ̂)+0.0195% CI -0.16 to 0.20 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points+0.1 ptsCI −1.6 to +2.0 pts · the Score on a readable scale (×10) The Draft Analyst's calls beat the consensus by ~0% in-window — about +0.1 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Cam Robinson · Dobber Prospects / EliteProspects | +0.00 | +0%within margin | +0.0 pts |
Score (γ̂)+0.0095% CI -0.19 to 0.21 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points+0.0 ptsCI −1.9 to +2.1 pts · the Score on a readable scale (×10) Cam Robinson's calls beat the consensus by ~0% in-window — about +0.0 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Future Considerations · FC Hockey | −0.01 | +0%within margin | −0.1 pts |
Score (γ̂)−0.0195% CI -0.24 to 0.20 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points−0.1 ptsCI −2.4 to +2.0 pts · the Score on a readable scale (×10) Future Considerations's calls trailed the consensus by ~0% in-window — about −0.1 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸McKeen's Hockey · McKeen's Hockey | −0.01 | +0%within margin | −0.1 pts |
Score (γ̂)−0.0195% CI -0.22 to 0.17 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points−0.1 ptsCI −2.2 to +1.7 pts · the Score on a readable scale (×10) McKeen's Hockey's calls trailed the consensus by ~0% in-window — about −0.1 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Bob McKenzie · TSN (Scout Consensus) | −0.02 | +0%within margin | −0.2 pts |
Score (γ̂)−0.0295% CI -0.24 to 0.12 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points−0.2 ptsCI −2.4 to +1.2 pts · the Score on a readable scale (×10) Bob McKenzie's calls trailed the consensus by ~0% in-window — about −0.2 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸Ryan Kennedy · The Hockey News | −0.02 | +0%within margin | −0.2 pts |
Score (γ̂)−0.0295% CI -0.20 to 0.10 · crosses 0 % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points−0.2 ptsCI −2.0 to +1.1 pts · the Score on a readable scale (×10) Ryan Kennedy's calls trailed the consensus by ~0% in-window — about −0.2 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| ▸ISS Hockey · International Scouting Services | −0.07 | +0%within margin | −0.7 pts |
Score (γ̂)−0.0795% CI -0.46 to -0.01 · excludes 0 (in-sample, shrunk) % over consensus+0%95% CI +0% to +0% · crosses 0 — the held-out test Draft-value points−0.7 ptsCI −4.6 to −0.1 pts · the Score on a readable scale (×10) ISS Hockey's calls trailed the consensus by ~0% in-window — about −0.7 points of draft value per pick — but held out of sample the range runs through zero, so it's not distinguishable from average. Held-out and capped by ~4 effectively-independent boards · career-long (≈0 per season) · a negative value = regression · scouts add ≈0 on the bigger question of who even reaches the NHL. | |||
| Scout | Score |
|---|---|
| Chris Peters · ESPN / FloHockey / TSN | +0.44 |
| Corey Pronman · ESPN / The Athletic | +0.27 |
| Craig Button · TSN | −0.03 |
| Ryan Kennedy · The Hockey News | −0.10 |
| Sam Cosentino · Sportsnet | −0.17 |
| Bob McKenzie · TSN (Scout Consensus) | −0.17 |
| Scout | Score |
|---|---|
| Sam Cosentino · Sportsnet | +0.62 |
| McKeen's Hockey · McKeen's Hockey | +0.38 |
| Bob McKenzie · TSN (Scout Consensus) | +0.24 |
| Ryan Kennedy · The Hockey News | +0.04 |
| Corey Pronman · ESPN / The Athletic | −0.04 |
| Scott Wheeler · The Athletic | −0.20 |
| Craig Button · TSN | −0.26 |
| Cam Robinson · Dobber Prospects / EliteProspects | −0.62 |
| Scout | Score |
|---|---|
| Cam Robinson · Dobber Prospects / EliteProspects | +1.03 |
| Chris Peters · ESPN / FloHockey / TSN | +0.93 |
| Craig Button · TSN | +0.64 |
| Corey Pronman · ESPN / The Athletic | +0.26 |
| The Draft Analyst · The Draft Analyst | +0.23 |
| Scott Wheeler · The Athletic | +0.04 |
| McKeen's Hockey · McKeen's Hockey | −0.18 |
| Hockey Prospect · Hockey Prospect (Black Book) | −0.21 |
| ISS Hockey · International Scouting Services | −0.22 |
| Sam Cosentino · Sportsnet | −0.42 |
| Future Considerations · FC Hockey | −0.42 |
| Ryan Kennedy · The Hockey News | −0.44 |
| Bob McKenzie · TSN (Scout Consensus) | −0.71 |
| Scout | Score |
|---|---|
| Chris Peters · ESPN / FloHockey / TSN | +0.45 |
| Bob McKenzie · TSN (Scout Consensus) | +0.36 |
| McKeen's Hockey · McKeen's Hockey | +0.35 |
| Scott Wheeler · The Athletic | +0.28 |
| Ryan Kennedy · The Hockey News | +0.26 |
| Cam Robinson · Dobber Prospects / EliteProspects | +0.17 |
| Craig Button · TSN | +0.15 |
| The Draft Analyst · The Draft Analyst | +0.07 |
| ISS Hockey · International Scouting Services | +0.04 |
| Hockey Prospect · Hockey Prospect (Black Book) | +0.04 |
| Corey Pronman · ESPN / The Athletic | +0.00 |
| Future Considerations · FC Hockey | −0.09 |
| Sam Cosentino · Sportsnet | −0.34 |
| Scout | Score |
|---|---|
| Future Considerations · FC Hockey | +0.28 |
| Craig Button · TSN | +0.28 |
| Cam Robinson · Dobber Prospects / EliteProspects | +0.23 |
| Bob McKenzie · TSN (Scout Consensus) | +0.05 |
| Corey Pronman · ESPN / The Athletic | −0.09 |
| The Draft Analyst · The Draft Analyst | −0.09 |
| Hockey Prospect · Hockey Prospect (Black Book) | −0.30 |
| Ryan Kennedy · The Hockey News | −0.41 |
| McKeen's Hockey · McKeen's Hockey | −0.43 |
| ISS Hockey · International Scouting Services | −0.53 |
| Scout | Score |
|---|---|
| Future Considerations · FC Hockey | +1.73 |
| Craig Button · TSN | +0.26 |
| Hockey Prospect · Hockey Prospect (Black Book) | +0.25 |
| Bob McKenzie · TSN (Scout Consensus) | +0.21 |
| ISS Hockey · International Scouting Services | +0.08 |
| Corey Pronman · ESPN / The Athletic | −0.04 |
| The Draft Analyst · The Draft Analyst | −0.07 |
| Ryan Kennedy · The Hockey News | −0.07 |
| McKeen's Hockey · McKeen's Hockey | −0.09 |
| Scout | Score |
|---|---|
| Craig Button · TSN | +0.57 |
| Corey Pronman · ESPN / The Athletic | +0.24 |
| Ryan Kennedy · The Hockey News | +0.14 |
| Future Considerations · FC Hockey | +0.09 |
| McKeen's Hockey · McKeen's Hockey | −0.20 |
| ISS Hockey · International Scouting Services | −0.35 |
| Bob McKenzie · TSN (Scout Consensus) | −0.90 |
| Scout | Score |
|---|---|
| Hockey Prospect · Hockey Prospect (Black Book) | +0.79 |
| Craig Button · TSN | +0.44 |
| Corey Pronman · ESPN / The Athletic | +0.18 |
| McKeen's Hockey · McKeen's Hockey | +0.07 |
| Future Considerations · FC Hockey | −0.02 |
| Ryan Kennedy · The Hockey News | −0.26 |
| Bob McKenzie · TSN (Scout Consensus) | −0.37 |
| ISS Hockey · International Scouting Services | −0.75 |
| Scout | Score |
|---|---|
| Corey Pronman · ESPN / The Athletic | +1.11 |
| Hockey Prospect · Hockey Prospect (Black Book) | +0.18 |
| Ryan Kennedy · The Hockey News | +0.11 |
| Sam Cosentino · Sportsnet | −0.07 |
| ISS Hockey · International Scouting Services | −0.18 |
| Craig Button · TSN | −0.32 |
| McKeen's Hockey · McKeen's Hockey | −0.37 |
| Future Considerations · FC Hockey | −0.37 |
| Bob McKenzie · TSN (Scout Consensus) | −0.58 |
| Chris Peters · ESPN / FloHockey / TSN | −0.71 |
Raw, unshrunk score for that one draft (no % or points chain — one class is too small a sample to translate). Each scout's whole board that year, judged on careers so far; 2019–2021 still unfolding.
| Team | ||
|---|---|---|
| ▸PHI | +0.86within margin | +8.6 pts |
Score (rate-WAR)+0.8695% CI 0.16 to 2.70 · excludes 0 individually — but not licensed after the two-sided team permutation (both tails, ~30 franchises) Draft-value points+8.6 ptsCI +1.6 to +27.0 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸SJS | +0.80within margin | +8.0 pts |
Score (rate-WAR)+0.8095% CI 0.12 to 4.16 · excludes 0 individually — but not licensed after the two-sided team permutation (both tails, ~30 franchises) Draft-value points+8.0 ptsCI +1.2 to +41.6 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸WSH | +0.78within margin | +7.8 pts |
Score (rate-WAR)+0.7895% CI 0.08 to 4.04 · excludes 0 individually — but not licensed after the two-sided team permutation (both tails, ~30 franchises) Draft-value points+7.8 ptsCI +0.8 to +40.4 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸CAR | +0.68within margin | +6.8 pts |
Score (rate-WAR)+0.6895% CI -0.21 to 3.61 Draft-value points+6.8 ptsCI −2.1 to +36.1 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸ANA | +0.52within margin | +5.2 pts |
Score (rate-WAR)+0.5295% CI -0.65 to 3.72 Draft-value points+5.2 ptsCI −6.5 to +37.2 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸OTT | +0.52within margin | +5.2 pts |
Score (rate-WAR)+0.5295% CI -0.55 to 5.57 Draft-value points+5.2 ptsCI −5.5 to +55.7 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸DET | +0.34within margin | +3.4 pts |
Score (rate-WAR)+0.3495% CI -0.94 to 2.43 Draft-value points+3.4 ptsCI −9.4 to +24.3 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸WPG | +0.21within margin | +2.1 pts |
Score (rate-WAR)+0.2195% CI -1.21 to 2.11 Draft-value points+2.1 ptsCI −12.1 to +21.1 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸STL | +0.07within margin | +0.7 pts |
Score (rate-WAR)+0.0795% CI -2.63 to 3.29 Draft-value points+0.7 ptsCI −26.3 to +32.9 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸COL | +0.03within margin | +0.3 pts |
Score (rate-WAR)+0.0395% CI -3.11 to 3.46 Draft-value points+0.3 ptsCI −31.1 to +34.6 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸NJD | +0.02within margin | +0.2 pts |
Score (rate-WAR)+0.0295% CI -2.28 to 2.43 Draft-value points+0.2 ptsCI −22.8 to +24.3 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸NYI | +0.00within margin | +0.0 pts |
Score (rate-WAR)+0.0095% CI -1.76 to 1.76 Draft-value points+0.0 ptsCI −17.6 to +17.6 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸FLA | −0.01within margin | −0.1 pts |
Score (rate-WAR)−0.0195% CI -1.31 to 1.27 Draft-value points−0.1 ptsCI −13.1 to +12.7 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸NSH | −0.03within margin | −0.3 pts |
Score (rate-WAR)−0.0395% CI -1.94 to 1.78 Draft-value points−0.3 ptsCI −19.4 to +17.8 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸ARI | −0.04within margin | −0.4 pts |
Score (rate-WAR)−0.0495% CI -2.49 to 2.20 Draft-value points−0.4 ptsCI −24.9 to +22.0 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸LAK | −0.14within margin | −1.4 pts |
Score (rate-WAR)−0.1495% CI -2.56 to 1.75 Draft-value points−1.4 ptsCI −25.6 to +17.5 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸BUF | −0.15within margin | −1.5 pts |
Score (rate-WAR)−0.1595% CI -2.16 to 1.46 Draft-value points−1.5 ptsCI −21.6 to +14.6 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸BOS | −0.19within margin | −1.9 pts |
Score (rate-WAR)−0.1995% CI -1.77 to 1.06 Draft-value points−1.9 ptsCI −17.7 to +10.6 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸CBJ | −0.21within margin | −2.1 pts |
Score (rate-WAR)−0.2195% CI -2.70 to 1.53 Draft-value points−2.1 ptsCI −27.0 to +15.3 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸EDM | −0.25within margin | −2.5 pts |
Score (rate-WAR)−0.2595% CI -3.71 to 1.71 Draft-value points−2.5 ptsCI −37.1 to +17.1 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸TOR | −0.28within margin | −2.8 pts |
Score (rate-WAR)−0.2895% CI -4.51 to 1.74 Draft-value points−2.8 ptsCI −45.1 to +17.4 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸TBL | −0.30within margin | −3.0 pts |
Score (rate-WAR)−0.3095% CI -2.21 to 0.96 Draft-value points−3.0 ptsCI −22.1 to +9.6 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸VAN | −0.33within margin | −3.3 pts |
Score (rate-WAR)−0.3395% CI -3.56 to 1.30 Draft-value points−3.3 ptsCI −35.6 to +13.0 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸DAL | −0.40within margin | −4.0 pts |
Score (rate-WAR)−0.4095% CI -3.15 to 0.97 Draft-value points−4.0 ptsCI −31.5 to +9.7 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸MTL | −0.58within margin | −5.8 pts |
Score (rate-WAR)−0.5895% CI -2.34 to 0.33 Draft-value points−5.8 ptsCI −23.4 to +3.3 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸MIN | −0.67within margin | −6.7 pts |
Score (rate-WAR)−0.6795% CI -4.61 to 0.14 Draft-value points−6.7 ptsCI −46.1 to +1.4 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸CGY | −0.77within margin | −7.7 pts |
Score (rate-WAR)−0.7795% CI -3.31 to 0.02 Draft-value points−7.7 ptsCI −33.1 to +0.2 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸NYR | −1.26within margin | −12.6 pts |
Score (rate-WAR)−1.2695% CI -2.38 to -0.79 · excludes 0 individually — but not licensed after the two-sided team permutation (both tails, ~30 franchises) Draft-value points−12.6 ptsCI −23.8 to −7.9 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
| ▸CHI | −1.30within margin | −13.0 pts |
Score (rate-WAR)−1.3095% CI -3.43 to -0.89 · excludes 0 individually — but not licensed after the two-sided team permutation (both tails, ~30 franchises) Draft-value points−13.0 ptsCI −34.3 to −8.9 pts · the Score on a readable scale (×10) A pick or two per draft is too few to split this team into a held-out test of its own, so it stays within the margin of error — the whole-league team edge is just +14% (CI −33% to +86%, crosses 0). Rate-WAR (opportunity-adjusted), but ~20% of even the league-wide team edge is just drafting higher, not skill · the points are the Score on a readable scale · career-long. | ||
Teams make only a pick or two per draft — far too few to grade a single year, or to give any one team its own held-out %. Not scored: PIT, SEA, VGK.
Draft-value points are the Score on a readable scale (Score × 10, where ≈1 point is one tier of draft value moved) — a sense of scale, not a precise count. The value measure is fit to a reference tier list (held-out ρ=0.917) and credits scoring, even-strength offense/defense, and award recognition. The % is held-out and capped by ~4 effectively-independent boards. Distinguishability is the held-out test, never the tighter in-sample Score CI. Updated 2026-07-02.
Each real call carries its leave-one-out weight — drop that one pick, refit, and see how far the grade moves. Watch the same call shrink as you go down the chain: a single draft’s bar → raw edge → Score.
How much better (+) or worse (−) than the scouting consensus a scout drafted, on average — after correcting for luck. 0 is dead average.
How it connects: A Score is just your raw edge after a luck-correction (shrinkage). Multiply it by a fixed constant and it becomes draft-value points — same number, new units. The held-out % is a separate test, not derived from it.
How it’s calculated: The slope of players’ demonstrated value on how far each pick departed from the consensus, pooled across the scout’s drafts and shrunk toward zero by its uncertainty (empirical-Bayes, cluster-robust). Demonstrated value = a position-fair value measure fit to a reference tier list (held-out ρ=0.917): scoring (versus same-position peers and in absolute terms), even-strength offense and defense, and award recognition (peak and sustained), weighted toward a player’s best sustained stretch, with an establishment floor from games played. Standardized within draft class and position. Teams are judged on rate-WAR.
How much a scout’s disagreements with the field improve the prediction of real NHL careers, as a share of what the consensus already predicts. It is about being right, not bold — it can be negative.
How it connects: Unlike the Score (which fits the past), this predicts held-out drafts — and here it is 0% for everyone. A positive Score with a 0% held-out means the edge is not repeatable.
How it’s calculated: Out of sample (leave-one-draft-out): each draft is predicted from the scout’s other drafts, then (R²_full − R²_base) / R²_base. Capped by ~4 effectively-independent boards.
The Score on a tangible, readable scale — draft-value points per pick. Positive = the scout’s off-consensus bets add draft value versus the field; negative = they subtract it. A unit-conversion of the Score, not a second measurement.
How it connects: Just the Score × a constant — so it moves exactly with the Score and never with the held-out %. A high % with a small Score still means small points.
How it’s calculated: Score × 10, onto a points scale where a whole point of value ≈ moving a pick a full tier closer to (or further from) his deserved slot. An AVERAGE linear translation — a sense of scale, not a precise count. (Teams: points per pick on rate value.)
The scout’s average edge over the consensus across drafts — UNSHRUNK. Vivid, but over-optimistic: a couple of lucky drafts inflate it.
How it connects: Pools the per-draft bars in drafts vs the field; shrink it and you get the Score. It is always at least as big as the Score — the gap between them is how much was luck.
How it’s calculated: The precision-weighted mean of the per-draft coefficients, with no luck-correction. Compare it to the Score to see how much survives the shrink.
How the scout did versus the consensus in each individual draft — ▲ beat the field that year, ▽ trailed — and how many came in positive.
How it connects: Pool these bars → raw edge → shrink → Score. A steady column survives the shrink; one or two giant bars get pulled down — that gap is the whole point of the held-out %.
How it’s calculated: Each bar is that draft’s coefficient (outcome-on-departure within one class); bar height is the size of the edge. A steady column of ▲ means the edge repeats.