Variance-honest tracking + methods-tree calibration. Per nhl-betting-expert
audit: NHL is high-variance; this page surfaces drawdown windows + CLV
trajectory so users can evaluate edge over multi-bet samples, not single
outcomes.
Methods-tree calibration (N=100 gate)
✓ N=468 ≥ 100. Methods-tree EV calibration-cleared.
NHL methods-tree calibration by outcome.
Outcome
N
Brier
home_reg
468
0.2288
away_reg
468
0.2089
tie
468
0.1776
Aggregate
1404
0.2051
Official record (82 graded calls)
Record 40-40-2
Profit +8.95u
Max drawdown (by day) 5.82u
Max upswing (by day) 4.83u
BET and LEAN calls graded at the closing line, flat 1-unit stakes: the same record the public record page shows.
Profit +10.9 per 100 staked (95% range -13.7 to +35.5).
These prices break even at a 58.0% win rate.
Expected band at 3% edge × N=82 calls:
+2.46u
(95% CI: -15.29 → 20.21)
CLV trajectory (last 20 markets)
NHL CLV at the price taken, cumulative, over the last 20 markets priced.
Market #
Cumulative CLV (pp)
440
+149.86pp
441
+144.31pp
442
+144.31pp
443
+148.83pp
444
+148.60pp
445
+150.01pp
446
+147.44pp
447
+147.78pp
448
+147.26pp
449
+140.36pp
450
+136.38pp
451
+133.73pp
452
+133.52pp
453
+130.69pp
454
+133.61pp
455
+131.68pp
456
+132.79pp
457
+132.65pp
458
+129.05pp
459
+128.09pp
Variance disclosure: 3+ seasons of NHL data needed for meaningful
calibration; hockey single-season variance is too high for definitive
conclusions. Read this dashboard as a multi-bet expectation, not a
single-game tout.
Variance-honest tracking
What to expect. NHL has the highest single-bet variance of any major sport. Expected drawdown for a +5% EV bettor over 100 bets is roughly +/-15 units; over 500 bets is roughly +/-25 units. A 200-bet losing stretch is normal even with positive EV. Read your win rate, ROI, and CLV across hundreds of bets, not dozens.
Expected drawdown windows
Cumulative-PnL quantiles for an N-bet sample at the assumed parameters
(win prob = 0.525,
payout = 0.91u
per win, 1-unit stake). 90% of N-bet outcomes land between the 5% and
95% columns; you should expect to be at or below the 5% column 1 time
in 20.
NHL expected drawdown windows by sample size.
Bets (N)
5% quantile (units)
Median (units)
95% quantile (units)
25
-7.78u
+0.07u
+7.91u
50
-10.96u
+0.14u
+11.23u
75
-13.38u
+0.21u
+13.79u
100
-15.42u
+0.28u
+15.97u
125
-17.20u
+0.34u
+17.89u
150
-18.80u
+0.41u
+19.63u
175
-20.27u
+0.48u
+21.24u
200
-21.64u
+0.55u
+22.74u
225
-22.92u
+0.62u
+24.15u
250
-24.12u
+0.69u
+25.50u
275
-25.26u
+0.76u
+26.78u
300
-26.35u
+0.82u
+28.00u
325
-27.39u
+0.89u
+29.18u
350
-28.39u
+0.96u
+30.32u
375
-29.35u
+1.03u
+31.41u
400
-30.28u
+1.10u
+32.48u
425
-31.18u
+1.17u
+33.51u
450
-32.05u
+1.24u
+34.52u
475
-32.89u
+1.31u
+35.50u
500
-33.71u
+1.38u
+36.46u
CLV at the price taken (459 NHL markets priced)
Markets with positive CLV 202
Markets with negative CLV 212
Cumulative CLV +128.09pp
Mean CLV +0.279pp
CLV here is the price we listed against the closing price, both the best
on the board, for every NHL market we priced: the measure /trust/clv
publishes. Positive CLV across hundreds of markets means the prices we
show are consistently better than where the market closes.