Strategy Lab

Strategy Lab

Every system here is replayed on the opening prices books actually posted, at the best price among the books in your state, after your state's fees, and we publish every result, including the ones that lose.

These are backtests: hypothetical performance, not bets anyone placed, and never a promise about the next bet. Replayed on 335 game days, 2024-10-22 to 2026-04-17. The holdout, set aside and reported on its own, is the 2025-26 season.

Members

Build your own system

Choose the markets, the side, the line and the filters from the same catalog the presets use, written in plain words, and replay it the same way: opening prices at the books licensed in your state, after your state's fees, with the range, the luck meter and the holdout on every result. A system has at most 3 legs and at most 12 filters per leg; you can keep at most 10 systems.

Sign in to build and track your own systems.

Preset

Rebounder plus scorer in fast games

Rejected by the LabOn every season replayed it loses at opening prices even at the top of its likely range. It stays here to show what the Lab catches.

Our first system, tested as two straight bets. In games we expect to run faster than both teams usually play, it backs the Over on the main line for a big rebounder (projected 9 or more rebounds) and for a high-volume scorer (projected 20 or more points), as long as each plays real minutes and the game doesn't look like a blowout. It bets the situation, not the price: it takes the best opening price at a book in your state whether or not that price beats the market. The original system is a two-leg alt-line parlay. We test the legs as singles because alt-line prices exist for only part of one season in our history, and a book's same-game parlay price can't be rebuilt from its single legs.

Bets816432 games, 52 void (did not play). Hit rate 46.8%.
Likely ROI after fees-16.9% to -3.5%95% range from resampling game days. Middle estimate -10.3%.
Luck meter99.8%Chance a bettor with no skill, at these prices, does at least this well, counting the 1 variation tried. No better than a bettor with no skill
Holdout, 2025-26-17.0% to +1.8%418 bets, middle estimate -7.5%. No 2025-26 result was seen before this preset was written: a clean test.

Rule card

  • Rebounder: Rebounds, Over, main line (the one most books price)
  • Our projection for the stat at least 9
  • Expected minutes at least 28
  • Sample size in games played at least 8
  • Game pace above his team's norm at least 1.5%
  • Expected margin either way at most 9.5 points
  • The book's price was still current when the last book opened at this line
  • Scorer: Points, Over, main line (the one most books price)
  • Our projection for the stat at least 20
  • Expected minutes at least 28
  • Sample size in games played at least 8
  • Game pace above his team's norm at least 1.5%
  • Expected margin either way at most 9.5 points
  • The book's price was still current when the last book opened at this line
  • Price: the best opening price among the books licensed in your state, 1 unit flat, after the state's per-bet fee

Forward record

No forward record yet. Once tracking is on, every match on the board is logged at the price first seen and graded from the box score, win or lose.

What it was tested on

  • Seasons: 2024-25, 2025-26 regular season, 335 game days with opening prices (2024-10-22 to 2026-04-17)
  • Decision time: the opening market. Player numbers and pace come only from games before each day
  • Books: the books licensed in IL (DraftKings, FanDuel, BetMGM, Caesars Sportsbook, Fanatics Sportsbook, BetRivers, theScore Bet, Hard Rock Bet, Bally Bet); each bet takes the best of their opening prices
  • Fair price: the opening market with its margin removed (power method), 4 or more books pricing both sides
  • Stake: 1 unit flat ($50), after IL's per-bet fees; a push refunds the stake but not the fee
  • Grading: final box scores. A player who did not play is a void, not a loss
SeasonBets Hit rateROI after fees95% range
2024-2539845.0%-13.2%-22.9% to -3.6%
2025-26 (holdout)41848.6%-7.5%-17.0% to +1.8%

Hypothetical performance. A replay of past opening prices, not bets anyone placed. Past results do not predict future results. No closing prices for these bets, so CLV is not available. It is never estimated. Thresholds are the strategy scanner's defaults from the first engine (Oct 5 2026), written before any priced backtest existed. Turned into singles on Oct 9 2026 without looking at results. Whether box-score hit rates informed those defaults is not recorded.

Preset

Early line, later market

No better than luck so farA bettor with no skill matches this at least half the time once the variations tried are counted. Only the forward record will tell.

Some books post a player's line before the rest of the market. When the other books open later and their devigged price says the early book's number is a bargain (2% or more of value after your state's fee, with four or more books pricing both sides), this system takes the early book. It is the stale opener bucket our Oct 8 backtest set aside. The catch is timing: the early price was stamped before the market formed, and our history can't show whether it was still up when the later books opened. Every historical bet here is graded at a price we can't confirm you could have had, so read the backtest as an upper bound and the forward record as the real test.

Bets726290 games, 60 void (did not play). Hit rate 51.4%.
Likely ROI after fees-3.4% to +13.2%95% range from resampling game days. Middle estimate +4.5%.
Luck meter78%Chance a bettor with no skill, at these prices, does at least this well, counting the 16 variations tried (9% for this one alone). Within what luck can do
Holdout, 2025-26-3.4% to +13.2%726 bets, middle estimate +4.5%. Picked after seeing 2025-26 results, so this season is not a clean test. Only the forward record is.

Rule card

  • Early book: Points, Rebounds, Assists, Threes, Points + rebounds + assists, Over or Under, any line
  • Books pricing both sides of the line at least 4
  • Value after fees (your state's per-bet fee) at least 2%
  • Price over the devigged market, before fees at most 25%
  • Sample size in games played at least 10
  • The book opened 10+ minutes before the last book at this line, so its price is not confirmed
  • Price: the best opening price among the books licensed in your state, 1 unit flat, after the state's per-bet fee

Forward record

No forward record yet. Once tracking is on, every match on the board is logged at the price first seen and graded from the box score, win or lose.

What it was tested on

  • Seasons: 2024-25, 2025-26 regular season, 335 game days with opening prices (2024-10-22 to 2026-04-17)
  • Decision time: the opening market. Player numbers and pace come only from games before each day
  • Books: the books licensed in IL (DraftKings, FanDuel, BetMGM, Caesars Sportsbook, Fanatics Sportsbook, BetRivers, theScore Bet, Hard Rock Bet, Bally Bet); each bet takes the best of their opening prices
  • Fair price: the opening market with its margin removed (power method), 4 or more books pricing both sides
  • Stake: 1 unit flat ($50), after IL's per-bet fees; a push refunds the stake but not the fee
  • Grading: final box scores. A player who did not play is a void, not a loss
SeasonBets Hit rateROI after fees95% range
2024-250n/an/an/a
2025-26 (holdout)72651.4%+4.5%-3.4% to +13.2%

Hypothetical performance. A replay of past opening prices, not bets anyone placed. Past results do not predict future results. 726 of 726 bets took an opening price stamped more than 10 minutes before the market at that line was complete. We can't confirm it was still there, so read these results as an upper bound. No closing prices for these bets, so CLV is not available. It is never estimated. Picked after the Oct 8 2026 backtest gate showed 15 results (three ways of betting, five markets) on both seasons, this bucket among them.

Preset

Rebounds where the model agrees

No better than luck so farA bettor with no skill matches this at least half the time once the variations tried are counted. Only the forward record will tell.

Rebounds is the market where our props model has been closest to honest. This system bets a rebounds line only when two things agree: a book's opening price beats the devigged market by 2% or more after your state's fee, with four or more books pricing both sides, and our model, built only from games before that night, also makes the bet a winner at that price. The Oct 8 backtest found the model overconfident in every market, rebounds least of all, so the market sets the price and the model only gets a veto.

Bets514433 games, 19 void (did not play). Hit rate 50.6%.
Likely ROI after fees-6.8% to +9.7%95% range from resampling game days. Middle estimate +1.3%.
Luck meterover 99.9%Chance a bettor with no skill, at these prices, does at least this well, counting the 16 variations tried (36% for this one alone). Within what luck can do
Holdout, 2025-26-12.1% to +11.2%289 bets, middle estimate -0.4%. Picked after seeing 2025-26 results, so this season is not a clean test. Only the forward record is.

Rule card

  • Rebounds: Rebounds, Over or Under, any line
  • Books pricing both sides of the line at least 4
  • Value after fees (your state's per-bet fee) at least 2%
  • Price over the devigged market, before fees at most 25%
  • Sample size in games played at least 10
  • Value by our model at this price more than 0%
  • The book's price was still current when the last book opened at this line
  • Price: the best opening price among the books licensed in your state, 1 unit flat, after the state's per-bet fee

Forward record

No forward record yet. Once tracking is on, every match on the board is logged at the price first seen and graded from the box score, win or lose.

What it was tested on

  • Seasons: 2024-25, 2025-26 regular season, 335 game days with opening prices (2024-10-22 to 2026-04-17)
  • Decision time: the opening market. Player numbers and pace come only from games before each day
  • Books: the books licensed in IL (DraftKings, FanDuel, BetMGM, Caesars Sportsbook, Fanatics Sportsbook, BetRivers, theScore Bet, Hard Rock Bet, Bally Bet); each bet takes the best of their opening prices
  • Fair price: the opening market with its margin removed (power method), 4 or more books pricing both sides
  • Stake: 1 unit flat ($50), after IL's per-bet fees; a push refunds the stake but not the fee
  • Grading: final box scores. A player who did not play is a void, not a loss
SeasonBets Hit rateROI after fees95% range
2024-2522552.0%+3.5%-8.8% to +15.6%
2025-26 (holdout)28949.5%-0.4%-12.1% to +11.2%

Hypothetical performance. A replay of past opening prices, not bets anyone placed. Past results do not predict future results. No closing prices for these bets, so CLV is not available. It is never estimated. Picked after the Oct 8 2026 backtest gate showed 15 results (three ways of betting, five markets) on both seasons, this one among them.

How to read it

A range, a holdout and the close

Why a range

A few hundred bets can't pin down a return. The same system can look 10 points better or worse from luck alone, so we show the range the true return likely sits in, from resampling whole game days, and keep the single middle number small.

Why a holdout

Try enough variations on the same seasons and some will look good by chance. So the latest season is set aside and reported on its own, and we count every variation tried. A holdout is a clean test only when nobody saw its results before the system was written; when that isn't true, we say so.

Why closing line value

Closing line value (CLV) asks whether your price beat where the market closed. It sorts skill from luck in hundreds of bets, where profit needs thousands. Our price history has opening prices only, so these backtests can't show CLV; the forward record will, once closing prices are captured for it.

Data gaps: 2024-25: 0 of 1236 games have opening lines with more than one time stamp. In the rest every line carries one stamp, so an early book can't be told apart there. 2025-26: 376 of 1233 games have opening lines with more than one time stamp. In the rest every line carries one stamp, so an early book can't be told apart there. No closing prices for these seasons, so CLV is not available for any backtest bet. No injury report as of each open: a player ruled out after the lines opened is graded as a void. No market spread at open: the expected margin comes from each team's net rating before the day. Alt-line (milestone) prices in the history are one-sided (no Under to remove the margin from) and cover a minority of games, so the rebounder plus scorer parlay itself can't be priced; its legs are tested as singles.