Pythia

Hits & Misses

Pythia's public track record. Every pick graded out-of-sample — the model trained on other seasons and scored the one it had never seen.

Updated June 2026 · leave-one-season-out · breakout and regression kept separate

Breakout calls

Young forwards (age ≤ 26) flagged before becoming a top-six scorer.

1.7×

Top-5 picks break out at 30% vs the 18% base rate.

Precision @530%1.7× base
Precision @1021%1.2× base
Top decile33%1.9× · n=24
Base rate18%43 of 245

10 seasons (20092023) · 245 eligible · 43 actual breakouts · out-of-sample

Highest-confidence hits

Pythia Score
Jonathan Huberdeau
FLA · 2013-14 · 21
94✓ Hit
Brendan Gallagher
MTL · 2016-17 · 25
84✓ Hit
Jeff Skinner
CAR · 2014-15 · 23
76✓ Hit
Brandon Dubinsky
NYR · 2011-12 · 26
68✓ Hit
Viktor Stalberg
CHI · 2010-11 · 25
67✓ Hit
Cam Atkinson
CBJ · 2014-15 · 26
57✓ Hit
Brett Connolly
BOS · 2015-16 · 24
56✓ Hit
Tomas Hertl
S.J · 2016-17 · 23
54✓ Hit
Kyle Okposo
NYI · 2012-13 · 25
54✓ Hit
Craig Smith
NSH · 2012-13 · 23
50✓ Hit
Jason Zucker
MIN · 2015-16 · 24
43✓ Hit
Mikael Backlund
CGY · 2014-15 · 26
42✓ Hit
Joonas Donskoi
S.J · 2016-17 · 25
28✓ Hit
Nick Foligno
CBJ · 2012-13 · 25
25✓ Hit
Kyle Turris
OTT · 2012-13 · 24
12✓ Hit

Honest misses — top-flagged, didn't pan out

Pythia Score
Jeff Skinner
CAR · 2012-13 · 21
99✕ Miss
Bryan Little
ATL · 2009-10 · 22
84✕ Miss
Nikolay Kulemin
TOR · 2011-12 · 26
81✕ Miss
Jordan Staal
PIT · 2009-10 · 21
72✕ Miss
Nail Yakupov
EDM · 2013-14 · 20
67✕ Miss
Andrew Shaw
CHI · 2015-16 · 25
64✕ Miss
Kyle Okposo
NYI · 2009-10 · 22
63✕ Miss
Nick Foligno
OTT · 2009-10 · 22
61✕ Miss
Eric Fehr
WSH · 2010-11 · 25
60✕ Miss
Jiri Tlusty
CAR · 2013-14 · 26
59✕ Miss
Michael Grabner
NYI · 2011-12 · 24
59✕ Miss
Jordan Staal
CAR · 2013-14 · 25
58✕ Miss

Ones we under-rated

Breakouts Pythia didn't see coming — the model's honest blind spots (real breakouts it scored low). Shown for credibility, not as calls we made.

Matt Beleskey
ANA · 2012-13
we said -69became 2.11 P/60 (73th)
Mikkel Boedker
ARI · 2012-13
we said -36became 2.14 P/60 (76th)
Kaapo Kakko
NYR · 2023-24· recent
we said -35became 2.59 P/60 (81th)
Andrew Cogliano
ANA · 2011-12
we said -13became 2.00 P/60 (70th)
Ryan O'Reilly
COL · 2010-11
we said -12became 2.24 P/60 (75th)
Adam Henrique
N.J · 2012-13
we said -10became 1.94 P/60 (67th)
Jannik Hansen
VAN · 2010-11
we said -9became 1.96 P/60 (65th)
Ryan McLeod
EDM · 2023-24· recent
we said -2became 2.39 P/60 (78th)

Regression calls

Established forwards (any age) with room to fall, flagged before a sustained decline.

1.0×

Top-5 picks decline at 16% vs the 16% base rate.

A harder, noisier call — the very top-5 lands near the base rate; the edge shows across the wider top decile (1.2×).

Precision @516%1.0× base
Precision @1019%1.2× base
Top decile19%1.2× · n=187
Base rate16%301 of 1868

10 seasons (20092023) · 1,868 eligible · 301 actual declines · out-of-sample

Highest-confidence hits

Decline risk
Joe Thornton
S.J · 2015-16 · 37
30✓ Hit
Alexander Wennberg
CBJ · 2016-17 · 22
30✓ Hit
Nik Antropov
ATL · 2009-10 · 30
27✓ Hit
Valtteri Filppula
T.B · 2013-14 · 30
27✓ Hit
Martin Erat
NSH · 2011-12 · 30
27✓ Hit
Valtteri Filppula
DET · 2011-12 · 28
26✓ Hit
Conor Sheary
PIT · 2016-17 · 25
26✓ Hit
Mike Ribeiro
WSH · 2012-13 · 33
25✓ Hit
Antoine Vermette
CBJ · 2009-10 · 28
24✓ Hit
Nick Foligno
CBJ · 2014-15 · 27
24✓ Hit
Jussi Jokinen
FLA · 2015-16 · 33
24✓ Hit
Teddy Purcell
T.B · 2012-13 · 27
23✓ Hit
Markus Granlund
VAN · 2016-17 · 24
23✓ Hit
Brad Boyes
NYI · 2012-13 · 31
22✓ Hit
Yegor Sharangovich
CGY · 2023-24 · 26· recent
28✓ Hit

Honest misses — top-flagged, didn't pan out

Decline risk
Alex Tanguay
CGY · 2010-11 · 31
40✕ Miss
David Desharnais
MTL · 2011-12 · 25
38✕ Miss
Sergei Kostitsyn
NSH · 2010-11 · 24
32✕ Miss
Marcus Johansson
WSH · 2011-12 · 21
31✕ Miss
Henrik Sedin
VAN · 2011-12 · 31
31✕ Miss
Mike Ribeiro
NSH · 2014-15 · 35
31✕ Miss
Ryan Getzlaf
ANA · 2010-11 · 26
30✕ Miss
Robert Thomas
STL · 2022-23 · 24· recent
30✕ Miss
Jonathan Drouin
COL · 2023-24 · 29· recent
30✕ Miss
David Krejci
BOS · 2015-16 · 30
29✕ Miss
Artturi Lehkonen
COL · 2022-23 · 28· recent
29✕ Miss
Henrik Sedin
VAN · 2015-16 · 35
28✕ Miss

How this works — and why we show the misses

Out-of-sample only. Every prediction is out-of-sample: the model was trained on all other seasons and scored the held-out season it had never seen (leave-one-season-out). No full-data model is used. Breakout and regression are separate track records and are never blended.

Every pick is shown — hits and misses. At these odds, misses are the majority, and that is expected: the breakout board's ~1.7× lift on a ~18% base rate still means most flagged players don't hit. A model that never missed would just be repeating the box score, not predicting ahead of it.

Breakout and regression are graded on separate labels, pools, and samples — they are never blended into one accuracy number. Data: MoneyPuck.com.