Replay · NFL

Test a rule against real closing prices.

Each rule here was replayed against every settled line it would have bet, at the closing price, with how sure the result is. Choose one to read it.

Rules replayed

4 more listed that the engine cannot run

Seasons replayed

1999–2024 · 2025 held out

Priced at

all 4 replayed rules · consensus close · a choice in the builder

01Rules

4 rules replayed · 4 more listed that the engine cannot run · Film Room

Loses01Fade every over13,680 bets · −3.3% ROI · open in the builder
Unclear02Rush attempts unders2,256 bets · −1.8% ROI · open in the builder
Loses03Chase the streak1,150 bets · −13.2% ROI · open in the builder
Cannot run04Follow the model's leansWaiting for walk-forward model predictions persisted for the historical seasons.
Cannot run05Buy the middleWaiting for a rule language with two legs: lab.strategy/1 bets one side at one price.
Cannot run06Arbitrage between booksWaiting for a rule language with two legs: lab.strategy/1 bets one side at one price.
Loses07Home underdogs of 3 to 71,400 bets · −5.6% ROI · open in the builder
Cannot run08Unders in the windWaiting for forecast weather joined to the replayed games.

4 rules replayed on this site so far. The more rules are tried, the more likely one clears zero by chance alone: with 4 tried, a 95% interval that allows for it spans about ±2.50 standard errors (Bonferroni), not ±1.96. The intervals on this page are the plain 95%.

Home underdogs of 3 to 7

spread underdog on spread, 1999-2025 REG weeks 1-18, 2 condition(s), consensus_close at close

Loses2025 held out

−5.6% ROI

1,400 bets, −5.6% ROI. Under the 5th percentile of random rules its size.

“Loses” is set by a fixed rule, not a judgement: the ROI interval lies wholly under zero. Re-checked on this page from the file’s own interval and luck check.

Hit rate48.8% [46.1%, 51.4%]

51.8% break-even

ROI−5.6% [−10.7%, −0.4%]

Cleared 48.8% [46.1, 51.4] of 1,343 graded, against a break-even of 51.8% at the average price paid (−107): the interval sits wholly under it. The ROI interval runs from −10.7% to −0.4%, and sits wholly under zero.

Not every price here is a close: 24 of 27 seasons (1999–2022) are priced from nflverse games line, which the file does not label a close.

Bets: 1,400; 434 weeks · 0 void
Cleared: 655; 688 missed · 57 push
Hit rate: 48.8%; 95% [46.1, 51.4] · of 1,343
Break-even: 51.8%; at the average price paid, −107
ROI: −5.6%; 95% [−10.7, −0.4] · by week
Units: −77.7u; 95% [−150.4, −5.8] · 1u flat

Worst drawdown: coming once the rule’s file carries its drawdown key (a-44).
CLV: none to publish. Every bet here is taken at the close, which is the price CLV is measured against.

Returns over time

units · by week block

This rule's ROI sits at percentile 2.4 among 1,000 random rules of 1,400 bets drawn from the same pool of 6,671: under the 5th percentile: fewer than 1 in 20 random rules this size did as badly.

Returns over time: coming once this rule’s file, lab/nfl/presets/home_underdogs_3_to_7.json, carries its week-block series (the returns key, a-44). Nothing is drawn from anything else in the meantime.

How this is computed

The line is cumulative units, staked 1 unit flat, counted in date order. It moves once per block of weeks, not once per bet: each block is whole weeks of one season holding at least the file's floor of cleared bets, because consecutive points of a bet-by-bet curve would give each bet’s book price back, and the file withholds those (“Deliberately restricted: no bet carries its book price, its book, its stake or its profit, and the file has no equity curve. The prices are The Odds API's, whose terms (read 2026-09-24, updated 31 Aug 2026) say nothing either way about displaying historical odds, so they stay on the server and only aggregates leave it. A winning flat bet's profit is its price minus one, and consecutive points of a cumulative curve difference back to it, so those are withheld for the same reason. ROI and units appear only where they average at least 10 cleared bets' prices. Do not add these fields back without a decision that the terms allow it.”). The band is where 90% of random rules land at the same block, the 5th to the 95th percentile of 1,000 rules drawing the same number of bets, at random, from the same pool of eligible lines on the same side. Inside it, the rule did nothing a coin could not.

By season

ROI · 95% interval

24 of 26 seasons' ROI intervals include zero; 2 sit wholly under it; 0 sit wholly over it.

Season	Bets	Cleared	95%	ROI	95%
1999under 100 bets	53	56.0%	[42.3, 68.8]	+6.5%	[−21.7, +30.8]
2000under 100 bets	47	46.8%	[33.3, 60.8]	−10.6%	[−31.5, +13.5]
2001under 100 bets	53	54.2%	[40.3, 67.4]	+3.1%	[−22.6, +23.0]
2002under 100 bets	54	51.9%	[38.7, 64.9]	−0.8%	[−18.4, +16.1]
2003under 100 bets	59	45.6%	[33.4, 58.4]	−12.5%	[−47.4, +18.3]
2004under 100 bets	50	47.9%	[34.5, 61.7]	−8.2%	[−39.0, +19.9]
2005under 100 bets	53	34.0%	[22.4, 47.8]	−33.1%	[−58.3, −4.4]
2006under 100 bets	54	59.6%	[46.1, 71.8]	+15.5%	[−5.0, +39.0]
2007under 100 bets	51	51.0%	[37.7, 64.1]	−0.5%	[−31.1, +26.9]
2008under 100 bets	48	44.4%	[30.9, 58.8]	−12.7%	[−36.0, +7.9]
2009under 100 bets	47	36.2%	[24.0, 50.5]	−30.2%	[−56.3, −3.8]
2010under 100 bets	61	46.6%	[34.3, 59.2]	−9.1%	[−30.0, +11.5]
2011under 100 bets	57	44.4%	[32.0, 57.6]	−13.4%	[−37.9, +10.7]
2012under 100 bets	57	47.3%	[34.7, 60.2]	−6.3%	[−42.4, +27.8]
2013under 100 bets	46	54.5%	[40.1, 68.3]	+6.8%	[−13.7, +27.9]
2014under 100 bets	53	52.8%	[39.7, 65.6]	+1.9%	[−24.6, +29.0]
2015under 100 bets	48	50.0%	[36.1, 63.9]	−2.0%	[−25.5, +24.6]
2016under 100 bets	43	47.6%	[33.4, 62.3]	−7.5%	[−24.7, +16.3]
2017under 100 bets	58	50.0%	[37.1, 62.9]	−2.6%	[−24.7, +20.1]
2018under 100 bets	45	58.1%	[43.3, 71.6]	+12.1%	[−20.0, +41.3]
2019under 100 bets	60	44.8%	[32.7, 57.5]	−12.8%	[−40.5, +15.4]
2020under 100 bets	65	53.8%	[41.9, 65.4]	+4.5%	[−16.0, +23.2]
2021under 100 bets	60	50.0%	[37.7, 62.3]	−4.1%	[−25.1, +19.1]
2022under 100 bets	51	56.3%	[42.3, 69.3]	+9.3%	[−10.6, +31.0]
2023under 100 bets	55	47.9%	[34.5, 61.7]	−7.8%	[−33.1, +17.9]
2024under 100 bets	72	39.7%	[28.9, 51.6]	−23.1%	[−49.1, +3.9]

How this is computed

A real effect shows up every year; one good season inside a run of intervals that include zero is what chance looks like. The ROI interval resamples whole weeks, so bets in the same week move together. Seasons under 100 bets are greyed: their intervals are wide enough that the row should not be read on its own.

2025 is held out

The file: “held out: no figure in this result includes the 2025 season”. The season is kept back so the rule can be tested once, on games it was never shaped by. Looking is done once, when the rule has stopped changing: the job that replays it reveals the season and writes that into the file, and from then on 2025 is marked as seen on this rule.

Not revealed · no 2025 figure anywhere on this page

Where it won and lost

95% intervals · under 100 bets greyed

6 segments in 4 cuts; 0 under 100 bets, greyed.

Market

Segment	Bets	Cleared	95%	ROI	95%	Against zero
Spread	1,400	48.8%	[46.1, 51.4]	−5.6%	[−10.7, −0.4]

Home or away

Segment	Bets	Cleared	95%	ROI	95%	Against zero
Home	1,400	48.8%	[46.1, 51.4]	−5.6%	[−10.7, −0.4]

Roof

Segment	Bets	Cleared	95%	ROI	95%	Against zero
Open air	1,076	48.4%	[45.4, 51.5]	−6.1%	[−11.9, −0.2]
Dome	324	49.8%	[44.3, 55.3]	−3.7%	[−13.8, +7.2]

De-vigged price

Segment	Bets	Cleared	95%	ROI	95%	Against zero
0.40–0.50	359	49.3%	[44.0, 54.5]	−1.0%	[−12.0, +9.7]
0.50–0.60	1,041	48.6%	[45.5, 51.7]	−7.1%	[−12.9, −1.6]

How this is computed

The same bets, cut by what was known before kickoff. A segment is not a new rule: with this many cuts, some will sit away from the rest by chance, so a single segment’s interval is not evidence on its own. ROI and units are withheld where they would average fewer than 10 cleared bets’ prices, and the file says so on the row.

The rule, as the engine approved it

spread underdog on spread, 1999-2025 REG weeks 1-18, 2 condition(s), consensus_close at close

Bet type: spread
Markets: Spread
Side: underdog
Seasons: 1999–2025 · REG · weeks 1–18
Conditions: game.home is true · price.line is between 3 and 7
Line: main
Price: consensus_close · draftkings, fanduel, betmgm · de-vig multiplicative · at the book's posted juice · -110 where none is posted
Timing: close
Staking: flat · 1u
Limits: per game no limit · per player week 1 · per week no limit
Settlement: push: refund · no snap: void

Where the prices came from

1999–2022 · nflverse games line · not labelled a close · one line per game, provider not verified per season; not labelled a close. Juice from nflverse odds where present, else assumed -110
2023–2025 · The Odds API historical close, draftkings, fanduel, betmgm · a close · last snapshot at or before kickoff

strategy 528680058834c4d316cf3265bfb44b6e0eaaf30bf0d089bfb2132c17fd2bbd18 · AUDIT-2026-09-24, the Lab library launch set

How this is computed

The strategy is stored exactly as the engine approved it, defaults filled, and hashed; the same rule always hashes the same. Every condition is a figure stored as of before kickoff, so nothing known after the game can decide a bet.

Every bet it made

50 shown of 1,400 bets, newest first; page 5 of 28. Filters narrow the list; nothing is removed.

Date · wk	Bet	De-vig	Result	Actual
2021-11-282021 · wk 12	INDSpread 3 homeIND v TB	51.5%	missed	−7
2021-11-282021 · wk 12	NYGSpread 4 homeNYG v PHI	51.2%	cleared	6
2021-11-252021 · wk 12	NOSpread 7 homeNO v BUF	51.4%	missed	−25
2021-11-212021 · wk 11	NYJSpread 3.5 homeNYJ v MIA	50.5%	missed	−7
2021-11-212021 · wk 11	CHISpread 5 homeCHI v BAL	50.5%	cleared	−3
2021-11-212021 · wk 11	JAXSpread 6.5 homeJAX v SF	49.8%	missed	−20
2021-11-182021 · wk 11	ATLSpread 7 homeATL v NE	48.6%	missed	−25
2021-11-152021 · wk 10	SFSpread 3.5 homeSF v LA	51.2%	cleared	21
2021-11-072021 · wk 9	CARSpread 3.5 homeCAR v NE	50.9%	missed	−18
2021-11-072021 · wk 9	NYGSpread 3 homeNYG v LV	50.7%	cleared	7
2021-10-312021 · wk 8	NOSpread 4 homeNO v TB	50.9%	cleared	9
2021-10-312021 · wk 8	DETSpread 3 homeDET v PHI	47.6%	missed	−38
2021-10-312021 · wk 8	CHISpread 4.5 homeCHI v SF	50.5%	missed	−11
2021-10-252021 · wk 7	SEASpread 6 homeSEA v NO	50.2%	cleared	−3
2021-10-242021 · wk 7	TENSpread 4 homeTEN v KC	48.8%	cleared	24
2021-10-242021 · wk 7	NYGSpread 3 homeNYG v CAR	50.5%	cleared	22
2021-10-182021 · wk 6	TENSpread 6 homeTEN v BUF	50.7%	cleared	3
2021-10-172021 · wk 6	NESpread 3.5 homeNE v DAL	49.3%	missed	−6
2021-10-172021 · wk 6	DETSpread 3.5 homeDET v CIN	48.8%	missed	−23
2021-10-172021 · wk 6	CHISpread 5.5 homeCHI v GB	49.8%	missed	−10
2021-10-172021 · wk 6	WASSpread 6 homeWAS v KC	50.0%	missed	−18
2021-10-142021 · wk 6	PHISpread 7 homePHI v TB	52.0%	cleared	−6
2021-10-102021 · wk 5	JAXSpread 4 homeJAX v TEN	51.2%	missed	−18
2021-10-032021 · wk 4	NESpread 6 homeNE v TB	50.0%	cleared	−2
2021-10-032021 · wk 4	NYJSpread 5.5 homeNYJ v TEN	50.2%	cleared	3
2021-09-192021 · wk 2	BALSpread 3.5 homeBAL v KC	52.8%	cleared	1
2021-09-192021 · wk 2	PHISpread 3 homePHI v SF	52.2%	missed	−6
2021-09-192021 · wk 2	JAXSpread 6 homeJAX v DEN	52.2%	missed	−10
2021-09-192021 · wk 2	MIASpread 3.5 homeMIA v BUF	51.8%	missed	−35
2021-09-192021 · wk 2	CARSpread 3 homeCAR v NO	51.7%	cleared	19
2021-09-192021 · wk 2	INDSpread 4 homeIND v LA	48.8%	cleared	−3
2021-09-192021 · wk 2	NYJSpread 5.5 homeNYJ v NE	50.0%	missed	−19
2021-09-132021 · wk 1	LVSpread 3 homeLV v BAL	47.6%	cleared	6
2021-09-122021 · wk 1	NYGSpread 3 homeNYG v DEN	53.3%	missed	−14
2021-09-122021 · wk 1	NOSpread 4 homeNO v GB	52.0%	cleared	35
2021-09-122021 · wk 1	CINSpread 3 homeCIN v MIN	53.3%	cleared	3
2021-09-122021 · wk 1	INDSpread 3 homeIND v SEA	52.6%	missed	−12
2021-09-122021 · wk 1	HOUSpread 3.5 homeHOU v JAX	51.8%	cleared	16
2021-01-032020 · wk 17	PHISpread 6.5 homePHI v WAS	49.1%	cleared	−6
2021-01-032020 · wk 17	KCSpread 7 homeKC v LAC	52.6%	missed	−17
2021-01-032020 · wk 17	SFSpread 7 homeSF v SEA	51.5%	cleared	−3
2021-01-032020 · wk 17	HOUSpread 7 homeHOU v TEN	51.4%	cleared	−3
2021-01-032020 · wk 17	CARSpread 6 homeCAR v NO	51.2%	missed	−26
2021-01-032020 · wk 17	CHISpread 4.5 homeCHI v GB	50.5%	missed	−19
2021-01-032020 · wk 17	BUFSpread 3.5 homeBUF v MIA	54.2%	cleared	30
2021-01-032020 · wk 17	DETSpread 3 homeDET v MIN	48.8%	cleared	−2
2020-12-282020 · wk 16	NESpread 7 homeNE v BUF	50.7%	missed	−29
2020-12-272020 · wk 16	DALSpread 3 homeDAL v PHI	47.8%	cleared	20
2020-12-272020 · wk 16	NYJSpread 6.5 homeNYJ v CLE	48.0%	cleared	7
2020-12-202020 · wk 15	NYGSpread 6 homeNYG v CLE	49.3%	missed	−14

All 1,400 bets in the rule’s file (JSON) · no price, book, stake or profit per bet, by the file’s restriction

How this is computed

One row per bet as the file lists it: the line, the side, the de-vigged probability of that side at the close, and how it settled. A void is a player with no snap: not a bet, and not a win.

Hypothetical results · nothing here was bet · not advice · lab/nfl/presets/home_underdogs_3_to_7.json, written Sun, Sep 27, 2:14 AM ET · universe built Sun, Sep 27, 2:14 AM ET · strategy 528680058834

Test a rule against real closing prices.

00Build a rule

01Rules

Returns over time

By season

2025 is held out

Where it won and lost

Market

Home or away

Roof

De-vigged price

The rule, as the engine approved it

Where the prices came from

Every bet it made