Research · published including the nulls

What was tested, and what it found

Every hypothesis this project has retired, with its interval. None of them found an edge that survives costs. That is the result, and it is published as plainly as a win would be.

Retired hypotheses

ID	Brief	Question	Verdict	Estimate [95%]	n	Games	Why	Script
R01	longshot	Do sportsbooks overprice longshot player props? realized rate in the 0.125-0.150 price bucket (priced 0.1388)	retired — no edge	+0.0898 [0.0552, 0.1429] probability	167	—	Every settled observation priced below 0.15 is a sacks over on a single-book 1-2 line ladder, and even there the priced value sits inside the Wilson interval.	research/longshot.py
R02	S00	Does the model's week-1 entry beat the Kalshi close? executable CLV at 1,000 contracts	retired — no edge	-13.52 [-17.06, -10.4] pp	493	14	Mid-to-mid CLV was +0.62pp [+0.13, +1.22], but the book tightened from below; where money changes hands CLV is deeply negative.	research/clv.py
R03	S01	Held to settlement, does one crossing at the touch pay? one-crossing CLV against the closing mid	retired — no edge	-3 [-3.8, -2.26] pp	—	—	Exactly mid-to-mid minus half the 7.22c entry spread; shrinking the ticket buys back slippage and nothing else.	research/clv.py
R04	M01	Would a resting maker order on the model's side have earned the spread? maker CLV conditional on a fill	retired — no edge	+0.28 [-0.34, 1] pp	704	14	Fill rate 18.0%; the fills were worth nothing while the misses were worth everything (selection gap +4.52pp [+3.58, +5.79]).	research/maker.py
R05	018	Does the model's side selection add value over a model-free maker control? conditional maker CLV, model's own side, 100 contracts	retired — no edge	+0.67 [0.05, 1.39] pp	—	14	The flipped side reads +0.07pp [-0.69, +0.93]; the weekly product peaks at ~$113/week at 250 contracts - real but small.	research/maker.py
R06	019 H1	Do Kalshi ladder rungs violate monotonicity on the touch, tradeably? violation episodes surviving taker fees on both legs	retired — no edge	+232 episodes of 892	892	—	95% are in-game rungs updating out of step; where depth could be checked the thinner leg carried a median of 1 contract, max 2.	research/structural.py
R07	019 H2	Do first-TD-scorer partitions sum to more or less than 1? incomplete first-TD events	not testable — data does not exist	+16 events of 18	18	—	13 of 16 settled week-1 events had no No Touchdown leg, so the partition sum is undefined.	research/structural.py
R08	019 H3	Does quoting both sides capture the spread, per series? conditional maker capture, KXNFLSPREAD	retired — no edge	-0.83 [-1.67, -0.29] pp	—	—	Negative on every tight series and zero on the prop series; spread and maker fee are collinear, so the ordering cannot separate them.	research/structural.py
R09	020	Does Kalshi's team-market price lag the sportsbook consensus by more than it costs to cross? mean gap, Kalshi mid minus de-vigged consensus	null — no effect	-0.19 [-0.39, 0.01] pp	1,060	16	Pre-kickoff gaps above 2.5pp were three claims on minority half-point lines and CLV to Kalshi's own close was -0.9pp; in-game gaps were book clock skew.	research/consensus.py
R10	021	Is the model's probability better than the market's on settled outcomes? Brier(model) - Brier(market)	retired — no edge	+0.0275 [0.0117, 0.041] Brier	682	14	Worse than the market, above the 0.021 minimum detectable effect, and indistinguishable from a smoothed prior-season hit rate.	research/score.py
R11	022 H1	After a player reaches a receptions rung mid-game, is the winning side still offered? executable net per contract at 120s, 10 contracts, depth-confirmed	open — awaiting holdout	+1.81 [1.24, 2.47] pp	16	8	Passes four of five bars and awaits week-2 replication; about $17 for the week, and final-stat determination is look-ahead against the live feed.	research/sweep/h1_settlement.py
R12	022 H2	Does order-book imbalance predict a move larger than the spread plus fee? top-decile move minus half-spread and fee	retired — no edge	—	—	—	The in-game slope is real and replicates on college football, but net of cost is negative in all 30 NFL and all 16 estimable CFB cells; the largest gross move is ~1.1pp against 1.5-24pp of cost.	research/sweep/h2_imbalance.py
R13	022 H3	When is the Kalshi book widest and tightest? spread lifecycle (execution map, not an edge)	null — no effect	—	—	—	Cross game lines 1-6h before kickoff; never cross a prop in-game, where REC spreads reach 16c and RSHATT 63c on 1-2 contracts.	research/sweep/h3_lifecycle.py
R14	022 scan	Does an open scan of week 1 find anything that clears five pre-registered bars? findings among candidates	retired — no edge	0 findings of 101 candidates	101	—	396 search tests, 230 at nominal p<0.05 against 16.4 expected; the replicating effects all lose to cost.	research/sweep/summarize.py
R15	023 Part 1	Walk-forward, does the model beat the de-vigged sportsbook close? Brier(model) - Brier(close), 2025	retired — no edge	+0.0218 [0.0162, 0.0277] Brier	5,574	271	Worse in every season (2023 +0.0272, 2024 +0.0265) against an MDE of ~0.008, and in all 13 variants bracketing the model's unverified constants.	research/walkforward.py
R16	023 Part 2	Do sportsbooks disagree enough for arbitrage or middles? historical receptions arbitrage mean size	retired — no edge	+1.09 pp	—	—	Arbs appear on 2.53% of receptions opportunities from slow books, fail replication, and are worth ~$5 per $500 instance before limiting; every middle loses.	research/bookvbook.py
R17	023 Part 3	Do teammates' lines lag an inactive announcement? inactive events with an announcement timestamp	not testable — data does not exist	0 events	0	—	No source on disk timestamps an inactive; Kalshi pulled a Friday-OUT player's markets on Saturday, removing the game-day window for known outs.	research/inactives.py

generated Tue, Sep 15, 5:22 PM ET

Model and market calibration

NFL week 1 2026, KXNFLREC + KXNFLRSHATT, common set n=682, 14 games

model (ECE 0.0764)market (ECE 0.0387) identityhollow = n < 30, not a data point

Wilson 95% intervals. Brier: model 0.1957, market 0.1682, naive 0.1924; model − market +0.0275 [0.0117, 0.041]. Source: research/score.py (brief 021).

Series	Bin	n	Forecast	Realized	Wilson 95%
model	0.0–0.1	148	0.049	0.122	[0.078, 0.184]
model	0.1–0.2	107	0.147	0.252	[0.180, 0.342]
model	0.2–0.3	91	0.251	0.264	[0.184, 0.362]
model	0.3–0.4	79	0.353	0.494	[0.386, 0.602]
model	0.4–0.5	67	0.444	0.507	[0.391, 0.624]
model	0.5–0.6	61	0.547	0.525	[0.402, 0.645]
model	0.6–0.7	47	0.651	0.596	[0.453, 0.724]
model	0.7–0.8	31	0.750	0.581	[0.408, 0.736]
model	0.8–0.9	42	0.833	0.905	[0.779, 0.962]
model	0.9–1.0	9	0.923	0.778	[0.453, 0.937]	not a data point
market	0.0–0.1	166	0.055	0.078	[0.046, 0.129]
market	0.1–0.2	90	0.137	0.178	[0.113, 0.269]
market	0.2–0.3	62	0.245	0.290	[0.192, 0.413]
market	0.3–0.4	47	0.345	0.404	[0.276, 0.547]
market	0.4–0.5	75	0.451	0.387	[0.285, 0.500]
market	0.5–0.6	84	0.548	0.595	[0.488, 0.694]
market	0.6–0.7	61	0.645	0.639	[0.514, 0.748]
market	0.7–0.8	52	0.745	0.808	[0.681, 0.892]
market	0.8–0.9	42	0.839	0.857	[0.722, 0.933]
market	0.9–1.0	3	0.940	1.000	[0.439, 1.000]	not a data point

The execution map

Cross game lines 1-6h before kickoff; never cross a prop in-game.

Median quoted spread (cents) · median touch size
Series	>72h	24-72h	6-24h	1-6h	0-1h	in-game
KXNFLREC	9¢ · —	4¢ · —	3¢ · —	3¢ · 50	2¢ · —	16¢ · 2
KXNFLRSHATT	6¢ · —	13¢ · —	10¢ · —	9¢ · 3	9¢ · —	63¢ · 1
KXNFLSPREAD	2¢ · —	1¢ · —	1¢ · —	1¢ · 7,943	1¢ · —	1¢ · 89
KXNFLTOTAL	2¢ · —	1¢ · —	1¢ · —	1¢ · —	1¢ · —	1¢ · —
KXNFLGAME	1¢ · —	1¢ · —	1¢ · —	1¢ · —	1¢ · —	1¢ · —

Spread relative to 60-minute price volatility
Series	Ratio
CHAMP	3.88
WINSWEEK	2.76
KXNFLRSHATT	2.51
WINS	2.13
DIVISION	1.18
KXNFLREC	1.05
KXNFLTOTAL	0.31
KXNFLSPREAD	0.30
KXNFLGAME	0.30

research/sweep/h3_lifecycle.py (brief 022) · generated Tue, Sep 15, 5:22 PM ET