Loading the full read
NHL · NYI @ OTT · 7:00 PM ET·
Tony DeAngelo
Assists 0.5 · NYI · @ OTT · Islanders at Senators
WATCH
OVER
0.5
UNDER
0.5
Our win %
76%
Market %
73%
Diff
+54.0%
Market
Assists 0.5
DK
-390
1.26
Projection0.2vs0.5(+54%)
L5
80%
4/5
L10
83%
5/6
L20
83%
5/6
SZN
83%
5/6
H2H
-
ACC
69%
n=25090
ACC is Bet Data Labs' graded record on NHL assists under plays: 69% of 25090 settled cards. It measures us, not this player.
About ACC Our plays in this bucket (NHL assists under) hit 68.8% over 25,090 graded cards, 25,015 of them replayed from historical lines. That is our settled record in this market, not this card's win probability and not a claim about the board as a whole.
Edge
+3.0pp
EV
-4.4%
Alpha Score
50
Recommended
0.00u
PASSWATCHLEANSTRONGPRESS
AB 4-4 on DeAngelo
Best lineBuy up to 1: an exact 1 returns your stake (29% push)
⧉
Same-Game Cap
This is read 8 of 14 the board is showing on this one game. Cards ranked 3rd or deeper inside their own game went 22-43 (33.8%) in July while the 1st and 2nd reads went 127-114, so it is decayed a tier band. The reads share one game state: a blowout or a washout takes all of them together. All 14 lose together 0.0% to 14.2% of the time. The low end assumes the legs are unrelated, the high end assumes one game state moves all of them. Same-game legs sit near the top of that band. Size it down and do not stack it with its siblings.
Why this card
- ›Assists projection 0.2 vs line 0.5
- ›Model read P(under) 62% sits 11pp off the market. Assists has no graded record yet, so it keeps the default 5.0pp lean, so the read was pulled back to that cap before pricing.
- ›Edge test fails: books are not quoting the same number, so there is no single line to price against.
- ›Prop market fair P(under) 73% across 2 books
- ›Calibration: NHL Assists plays predicted ~60% have hit 69% over 25090 graded, of which 12583 still describe the current model. Win prob raised 8.0pp.
- ›Baby line: a 0.5-line prop is an opportunity bet by definition. Capped at LEAN.
- ›Thin sample: the projection rests on 6 games. Capped at LEAN until 8.
- ›Opponent OTT missing 2, context on this card and not a knock against it
- ›Same-game cap: read 8 of 14 the board is showing on this game, and reads past the 2nd went 22-43 (33.8%) in July. Sized down, do not stack with its siblings. All 14 lose together somewhere between 0.0% and 14.2% of the time, and same-game legs sit near the top of that band.
KILL SIGNALS
- ⚠ baby line
PLAYS LIKE THIS
Bet Data Labs has graded 25090 NHL assists cards. The model predicted 60% on average and they hit 69%. Weighted for model changes since, that history is worth 12583 cards about the model running now. This card's win prob was raised 8.0pp for it.
Sharp anchor
SourceDe-vigged book market
Implied P(under 0.5)73.1%
Fair price+272 · 3.72
Books de-vigged2
Pulled10/15, 6:14 PM ET
Fair value is the mean of 2 independent two-sided de-vigs. Reference books (BetMGM, Caesars, BetRivers, reduced-juice) count toward this number but can never set the price above, which comes from a Connecticut book only.
Our projection
Projection vs line0.23 vs 0.5
Sourcegamelog
SportNHL
Games6
Current games6
Prior games0
Prior games fetched0
All model inputs (18)
Variance0.60
Modelgamelog-v0.1-generic
Pricingpx-shared-v2+logistic-normal-v1
sourcegamelog
modelgeneric-recency-blend
sportNHL
propHOCKEY_ASSISTS
games6
current_games6
prior_games0
prior_games_fetched0
current_mean0.170
prior_meannull
current_weight1
role_shift0
variance_floor0.600
opp_adj1
player_espn_id3114769
Market lines
| Book | Line | Over | Under |
|---|---|---|---|
| DK★ Best | 0.5 | +2753.75 | -3901.26 |
DK implied, with vig (this side): 79.6%
Confidence readOn the fence
76% to hitrange 69-83%price needs 80%
The price's breakeven sits inside our range, so this one turns on exactly where the projection lands.
The band is our projection's standard error run through the same model that sets this card's win probability, so it shows how much the number could move if our read on the player's mean is off. The tick is what -390 has to beat to profit. We grade against that instead of a threshold we invented.
◆ ALPHA SCORE BREAKDOWN
Why 50?
TOTAL
50/100
Win prob+55/ 55
Calibrated win probability 76%
Market+15/ 15
Market consensus 73% on our side
Price+-10/ 15
Negative EV at best price - juice eats the edge
Edge test+0/ 15
Edge test failed
Kill penalty-8/ 0
1 kill signal firing (graduated)
Components sum to 52; published as 50 after a tier rule (baby-line LEAN cap, or no PRESS without the market ≥55% on our side).
Prop weights: calibrated win probability 55 · market consensus 15 · price 15 · edge test 15 · kill signals -8 then -5 each. Clamped 0-100. Tiers: 90+ PRESS, 75-89 STRONG, 60-74 LEAN, 50-59 WATCH, <50 PASS.
◆ ALT-LINE LADDER
Tony DeAngelo · Under the number
The best move
Buy this under up to 1: an exact 1 returns your stake instead of losing (29% of outcomes), same win chance as the 0.5.
| Line | Win | Push | Loss | Price | EV |
|---|---|---|---|---|---|
| 0.5now | 62% | - | 38% | -390 | -21.9% |
| 1push-safe | 62% | 29% | 9% | -709 fair | - |
Win / push / loss come from our projected distribution. A whole-number line can push - an exact landing returns your stake - so buying to the whole number nearest our projection converts a slice of losses into stake-back. EV is shown only where a book actually quotes that line; other rungs show the model's fair price.
Best EV is the highest-value rung inside the trustworthy band. Tighten is the safest rung that stays +EV, for when a leg needs to survive rather than pay big. EV is our model's edge at the price a book actually quotes. Which rungs we trustis now de-vigged: we measure the book's margin on the two-sided main line (its vig-free read is 75% our side there) and strip that same margin off each one-sided rung. A losing-side rung is shown only when the book's own de-vigged price backs the model up, so a longshot can no longer post a phantom edge built on the model's fat tail. EV in (parentheses)is a rung where the model claims far more than the book's de-vigged number: indicative only, never a recommendation.
Team form
ESPN standings · L10 / streak / splitsAWAYNYI
3-3L103-3-0, 0 PTS
StreakW1
Road0-0
Run diff+0.7
HOMEOTT
5-2L105-2-0, 0 PTS
StreakL1
Home0-0
Run diff+0.3
Recent form is context, not signal. A team on a heater can still be the wrong side of the number. Cross-check with the projection above.
Line history
Loading…
Typical pre-close movement
Across 29 replayed days of NHL, this market moves 0.2pp of implied probability on average between six hours out and the close, and the line itself moves 0% of the time (mean 0.00 units of line across every pair measured, movers and non-movers together).
Absolute magnitude over 3913 players quoting player_assists, so signed moves cannot cancel. A volatile market is not an inefficient one: this says how much the close knows that six hours out did not, not which way.
Injuries
From ESPN · refreshed every 30 minNYI
Out · 1
- Alexander RomanovOUTout
Other · 1
- Mathew Barzal?Barzal (knee) will not join the Islanders for their two-game road trip, Andrew Gross of Newsday reports Thursday.
OTT
Out · 2
- Jake SandersonOUTout
- Nikolas MatinpaloOUTMatinpalo (upper body) won't play against St. Louis on Tuesday, and there is no timeline for his return, Bruce Garrioch of Postmedia reports.
Other · 2
- William Eklund?Eklund had surgery to repair a torn ankle tendon and will be out 10-12 weeks, Bruce Garrioch of Postmedia reports Thursday.
- Kurtis MacDermid?MacDermid (upper body) was moved to the injured non-roster list Monday.
Alpha Score breakdown
EDGE TESTFAILS
TierWATCH
Kelly fraction (full)0.00u
Quarter Kelly (recommended)0.00u
EV %-4.4%
Edge %+3.0pp
Card ID: c-401892507-TonyDeAngelo-HOCKEY_ASSISTSPublished 10/15/2026, 10:14:23 PM