Preseason

Exhibition games, scored by the early-season carryover model on last season's games only and priced from Kalshi. Starters rest, so these rows show how far the offseason view travels, not regular-season skill.

Tracked apart from everything else: never in the Ledger, Calibration, hypotheses or any fit (tag fourfactors_hl25_b2b_carry25_preseason_g0).

How to read these numbers

Judge every ROI against its null: the ROI if the market's no-vig prices were exactly right. That is about −4% (the hold: Kalshi's taker fee and spread, or the bookmaker's margin), not zero. Market is Kalshi's last pre-tip price wherever a game has one (2025-26 on), else the sportsbook close; the hypotheses stay on the sportsbook prices they were registered on. ± is one standard error (1 SE). Colour marks only gaps of at least 2 SE from the null: green above, red below; everything else is left plain. Shaded columns are the ones to read first.

Glossary

Lean: The side the model gives ≥ 50% (as recorded on the row).
Value side: The side where the model's probability beats the market's no-vig probability. Graded only through the pre-registered hypotheses; the blanket value bet is retired (see the note under the hypotheses).
Model %: The picked side's mean model win probability.
Market %: The picked side's mean no-vig market probability (q): the price with the bookmaker's margin removed.
Win %: How often the picked side actually won.
Break-even: The win rate the posted price needs to pay back; above q by about the hold.
ROI: Profit per 1-unit bet.
Null: The ROI expected if the market is right (q × payout − 1). vs null = ROI − null.
z: (ROI − null) ÷ SE. Across many cells, about one in twenty shows |z| ≥ 2 by chance.
EV: Win % − break-even; its null is q − break-even.
Fav ROI: 1u on the market favourite of the same games at the same price: the same-row baseline a pick rule has to beat (a tie at 50% goes to home).
Log loss, Brier: Proper scores of the probabilities; lower is better. Compared with the close on the same games.
Preseason: Written before tip and frozen, like native rows, but exhibitions.
Reconstructed: Scored afterwards with that season left out of the fit and graded at the close: hindsight, never forward evidence.

66 games recorded: 66 with a model P, 66 with a pregame price, 66 graded, 66 graded with a model P and a close.

Preseason · exhibitionKalshi close · 2026-27log loss model − market +0.1399 ± 0.0618 · n=66

Probabilities vs the close (log loss)

Market better

by 2.3 SE · model − market +0.1399 ± 0.0618 (1 SE)
0.8128 vs 0.6729 · n=66

Brier (model / market)

0.2937 / 0.2397

model − market +0.0540 ± 0.0254 (1 SE) · market better by 2.1 SE

Picked the winner (model lean / market favourite)

50.0% / 54.5%

same games · accuracy is not calibration

Model vs market calibration, P(home win)

Model Market (no-vig close)

Each point is one probability bin: across the x-axis what was stated, up the y-axis how often it happened. On the dashed diagonal = calibrated. The bar is ±2 SE, where a correct forecast would usually land; a bar that misses the diagonal is a real miss, not noise. Hover a point for its numbers; the table below has them all. Both forecasts are for the same games; each is binned on its own probability. Negative model − market on the scores means the model did better; an interval spanning 0 has not separated them.

Model bins in numbers — with the market's mean on the same games

Model P(home)	n	Model mean	Market mean (same games)	Actual	Actual − model (± 1 SE)
0.0–0.2	1	16.3%	48.8%	100.0%	+83.7 ± 37.0
0.2–0.3	5	26.2%	48.6%	60.0%	+33.8 ± 19.6
0.3–0.4	8	35.6%	45.8%	37.5%	+1.9 ± 16.9
0.4–0.5	9	45.3%	50.8%	66.7%	+21.4 ± 16.6
0.5–0.6	7	54.4%	50.2%	42.9%	-11.5 ± 18.8
0.6–0.7	10	64.2%	60.9%	40.0%	-24.2 ± 15.1
0.7–0.8	13	74.7%	64.5%	76.9%	+2.2 ± 12.0
0.8–1.0	13	85.3%	63.2%	46.2%	-39.2 ± 9.8

Flat 1u bets at the pregame price — Preseason · Kalshi close · 2026-27

Lean — the model's favourite

Picks	n	W–L	ROI	Null	vs null (± 1 SE)	Win %	Break-even	Model %	Market %	Fav ROI
All picks	66	33–33	-18.3%	-4.0%	-14.3 ± 10.6	50.0%	60.7%	69.1%	58.4%	-17.0%

Game by game — newest first

Date	Away @ Home	Model P(home)	Pregame q(home)	Close q(home)	Final (away–home)
2026-10-16	DEN @ LAL	60.8%	66.5%	57.9%	121–106
2026-10-16	POR @ GSW	49.8%	41.5%	48.8%	118–128
2026-10-16	UTA @ PHO	79.3%	61.9%	61.9%	119–116
2026-10-16	ATL @ DAL	32.8%	52.0%	50.7%	135–111
2026-10-16	SAC @ SAS	89.4%	81.5%	80.7%	97–126
2026-10-16	CHO @ MEM	24.5%	56.5%	56.5%	98–117
2026-10-16	WAS @ NOP	77.1%	52.0%	52.5%	106–107
2026-10-16	MIN @ CHI	37.1%	66.5%	66.5%	116–133
2026-10-16	MIA @ ORL	50.1%	66.5%	70.5%	115–136
2026-10-16	BOS @ PHI	42.8%	60.5%	77.5%	125–120
2026-10-16	TOR @ DET	74.4%	76.7%	78.7%	125–129
2026-10-16	MIL @ IND	61.6%	72.5%	71.5%	128–127
2026-10-15	OKC @ HOU	49.1%	43.1%	43.5%	89–116
2026-10-15	TOR @ NYK	63.2%	64.5%	58.3%	120–122
2026-10-14	DEN @ LAC	56.3%	37.5%	34.5%	110–122
2026-10-14	PHO @ SAS	77.1%	76.7%	80.5%	103–123
2026-10-14	NOP @ DAL	52.7%	41.1%	43.5%	101–97
2026-10-14	CHI @ MIL	57.1%	45.5%	32.8%	125–104
2026-10-14	DET @ MIN	46.2%	62.5%	63.5%	128–106
2026-10-14	BRK @ MIA	87.9%	61.9%	60.5%	128–144
2026-10-14	CHO @ BOS	63.3%	71.8%	74.8%	93–114
2026-10-13	GSW @ LAL	76.4%	53.0%	43.5%	101–89
2026-10-13	POR @ SAC	44.6%	41.1%	31.5%	110–98
2026-10-13	IND @ OKC	83.1%	71.5%	67.2%	112–114
2026-10-13	CLE @ ORL	42.8%	46.5%	46.0%	116–132
2026-10-13	NYK @ TOR	63.4%	76.5%	75.5%	98–148
2026-10-12	SAS @ UTA	20.1%	64.5%	66.5%	109–119
2026-10-12	MIN @ NYK	67.2%	50.5%	48.0%	122–104
2026-10-12	OKC @ ATL	50.0%	60.5%	63.5%	110–113
2026-10-12	BRK @ WAS	64.1%	57.5%	55.5%	131–122
2026-10-11	CHI @ DEN	81.1%	68.5%	68.5%	128–133
2026-10-11	MIL @ CHO	83.3%	56.5%	56.5%	90–105
2026-10-11	ORL @ CLE	74.6%	63.9%	63.9%	105–119
2026-10-11	DAL @ HOU	81.2%	52.0%	50.5%	132–115
2026-10-10	SAS @ PHO	39.4%	28.5%	29.2%	98–102
2026-10-10	SAC @ GSW	71.2%	67.5%	68.5%	106–111
2026-10-10	MIN @ MIA	57.7%	32.2%	43.3%	125–124
2026-10-10	PHI @ BOS	74.5%	67.2%	68.8%	106–144
2026-10-10	ATL @ IND	32.7%	53.0%	53.5%	119–117
2026-10-10	DET @ WAS	16.3%	44.9%	48.8%	108–114
2026-10-10	LAC @ TOR	63.8%	60.7%	60.9%	106–114
2026-10-09	MEM @ CHI	69.0%	54.5%	54.5%	104–100
2026-10-09	HOU @ DAL	33.6%	46.5%	42.5%	135–117
2026-10-08	SAC @ LAL	84.4%	77.5%	74.8%	110–114
2026-10-08	ATL @ SAS	70.3%	71.5%	71.5%	123–116
2026-10-08	NOP @ MIA	72.6%	72.5%	71.5%	118–128
2026-10-08	PHI @ BRK	28.3%	27.5%	27.5%	108–114
2026-10-08	WAS @ NYK	90.0%	68.5%	67.5%	111–109
2026-10-08	BOS @ CLE	56.7%	41.5%	63.5%	124–113
2026-10-07	GSW @ POR	75.0%	80.5%	81.5%	118–123
2026-10-07	ORL @ MEM	38.4%	45.8%	47.5%	122–118
2026-10-07	MIL @ OKC	84.2%	82.7%	81.5%	128–126
2026-10-07	PHO @ CHI	41.5%	45.3%	42.5%	117–124
2026-10-07	MIN @ IND	36.4%	43.1%	44.1%	112–123
2026-10-06	LAL @ GSW	47.9%	61.2%	67.8%	98–124
2026-10-06	DEN @ UTA	29.3%	49.5%	50.0%	117–106
2026-10-06	NOP @ OKC	83.4%	46.5%	44.5%	116–110
2026-10-06	BRK @ CHO	90.4%	52.0%	50.5%	124–90
2026-10-05	LAL @ SAC	28.8%	35.5%	42.5%	127–103
2026-10-05	MIN @ MIL	34.6%	34.8%	32.5%	116–97
2026-10-05	MEM @ ATL	86.9%	59.5%	59.5%	132–123
2026-10-05	PHO @ DET	75.5%	56.5%	53.5%	107–109
2026-10-05	NYK @ PHI	43.1%	39.5%	36.5%	97–120
2026-10-04	UTA @ DEN	84.1%	60.5%	59.5%	109–97
2026-10-04	GSW @ LAC	72.9%	42.6%	42.6%	101–104
2026-10-03	MIA @ TOR	65.7%	64.2%	51.8%	129–105

q is the no-vig market probability. Prices are refreshed each build until tip and frozen after; prices are Kalshi's YES asks with the taker fee included (as American odds: what a bet costs); q is the bid/ask midpoints normalised (the market's view), and the close is the last 1-minute Kalshi candle before tip. b2b_net is read from the previous day's scoreboard.

Built 2026-10-19 07:42 UTC. Win probabilities from prior-game four factors, roster talent and the NBA injury report; the market is a benchmark only and never enters the model. Full statistics: ledger_report.txt · model tags and coefficients: Model. Not betting advice.