$19.99 a month or $119.99 a season · cancel anytime
Net profit on Wedge-certified picks so far: —
Loading this week's matchups…
Select a matchup above.
Loading the schedule…
His hit rate this season and his role.
Defense, game script, weather and injuries.
Wedge % vs the price. A read needs a 3-point edge.
Loading a live example…
Research, not financial advice. Figures are before fees. 21+ · Not available in all states · 1-800-GAMBLER
Select any line to open its file
Every Week 10 line at any price (switch to 45–55¢ in the first menu), updated Mon Nov 16 at 9:27 AM CT. Where Robinhood has no line yet, the price comes from the sportsbooks. Hit rate is how often the line came true in his 2026 games; Wedge % is our chance it comes true this week; Bear · Base · Bull is our projection for that player and stat. Select any line to open its file: the bull, the bear and the base, the injury report, and the game script. ★ marks a high-scoring game. Thursday's JAX @ BAL is final, so its lines are off the board. Out: Sam Darnold, David Montgomery, A.J. Brown, Mike Evans, DeVonta Smith, Garrett Wilson (Sun) and Kyler Murray (Mon).
Every Week 10 line Wedge priced, at any price: Robinhood's own contracts and the Robinhood-style rungs priced from the sportsbooks, each with its 2026 hit rate. Inside 45–55¢ the columns are the same as the main screen. Outside it, Wedge % applies the same market shifts (fit on lines priced 40–60¢), so read it as a rough guide; reads are only made inside 45–55¢. Rows without an outlook show Wedge % as plain text.
| Certify | Bear · Base · Bull |
|---|
Lines priced 45–55¢ per game, and how many of them pass the >50% test.
Main props before books post them. Line: our expected opener. Proj.: our projection. Wedge %: our chance he clears it.
Loading the final card…
Loading the thesis…
Select any line to open its file
Robinhood NBA lines, Oct 24–27. Hit rate: last 20 games. Books %: sportsbooks' chance. Wedge %: ours.
Lines priced 45–55¢ per game, and how many pass the >50% test.
Lines. Every Robinhood NBA contract (Kalshi's order book, which Robinhood routes to) for games from Sat Oct 24 to Tue Oct 27: moneylines, spreads, totals, team totals, and player points, rebounds, assists, threes and points + rebounds + assists. Robinhood lists player props about a day ahead, so props cover tonight's games so far.
Hit rate. Player props: his last 20 games (2025-26 and 2026-27, any team). Team lines: that team's last 20 games; game totals: both teams' last 20. A line qualifies when its Yes price is 45–55¢ and the condition came true in more than half of those games (at least 10).
Outlook. Minutes cases: his games split into thirds by minutes, weighted toward this season and his last 15 games; each case's chance is his hit rate in those games nudged toward his overall rate, and tonight's case odds lean on his recent minutes and shift toward the bear case when the spread makes a blowout likelier than usual. Projection: the case average times tonight's adjustments (defense vs his position, pace, game script from the spread, implied team total, back-to-back, teammates out with his with/without record), capped at −25% / +30%. Our read averages the cases' chance with a gamma (points, PRA) or negative-binomial (rebounds, assists, threes) distribution priced at the projection, shrunk toward 50%; Wedge % = 70% the books' price for that rung + 30% our read. Injuries: CBS Sports injury report, Fri Oct 23 morning. Live record: each posted line settled against the final ESPN box score. Data: ESPN box scores via hoopR, Action Network book lines, Kalshi.
It will appear here in the NFL card’s format.
Keys, scripts and the thesis behind every line.
37,519 closing 2025-26 player props at sportsbook prices, settled from box scores.
Filter 48.9%, model 48.4%; both lost money (−2.5%, −1.3%). In the NBA, the price is the signal.
Dot: hit rate. Tick: average price. Green made money.
By stat. Return on the right.
By Wedge % and by price. On the dashed line = right.
Brier gain over a coin flip.
Average miss: book line 3.1, his average 3.28, our projection 3.26.
NBA picks will be tracked the same way.
Every 45–55¢ NBA line, settled from box scores. Return: one Yes each, before fees.
Dot: hit rate. Tick: average price. Green made money.
By market. Return on the right.
By Wedge % and by price. On the dashed line = right.
| Night | Game | Line | Price ¢ | Hit rate | Wedge % | Picked by | Result | Outcome | P/L ¢ |
|---|
Premier League, Champions League, MLS
Every 45–55¢ Robinhood line in the top leagues and MLS, with the full read.
Wedge analysts post picks every week and get a public record on the Analysts tab: win %, record and net profit, tracked all season. Standout picks can earn the Wedge certification.
Opens your email app with the application filled in.
One contract per pick at its listed price. Settled picks only; ranked by win %, then profit.
| Edit |
|---|
Week 6, 45–55¢ lines. Return: one Yes each, before fees.
Every Week 7–8 line, settled from box scores.
| Certify |
|---|
Roles and spots that cleared the book line. “Held” = also beat its 2023–25 average. Overs hit about 45% overall.
Projection = volume × efficiency. Volume follows recent games; efficiency needs a long window.
Yes win % by week. Hover for record and return.
Overs vs a book-style line, by week. Green beats the week's average.
Hot or cold streaks barely carry forward. Role changes do.
First four games (hollow) vs last four (filled), opponent-adjusted. Team momentum carries little.
45–55¢ lines, Weeks 1–5. Past the tick = hit more than its price implied.
Tick factors to combine them. Small groups are hidden.
Hit rate by pair of factors. Green above 50%, red below.
Week 6, 45–55¢ lines, by group.
Price vs hit rate for each Week 10 line. Above the diagonal = hit more than it cost.
Team lines (moneyline, spreads, winning margin, quarter/half winners and spreads, 1st-half/full-time) are checked against that team's own 2026 games. Game totals, quarter/half totals, ties and overtime are checked against every 2026 game either team has played. Player props are checked against that player's 2026 games, including games he played with no stat line (counted as zero).
Prop categories: positions and teams come from current ESPN rosters. The favorite is the side with the higher Robinhood moneyline (Sat Nov 7 ~2:25 PM CT). Player props Robinhood doesn't list use the Robinhood-style rungs priced from the sportsbooks described on the Week 10 lines tab. Defense softness counts rushing and receiving production by opposing RBs/FBs and by opposing WRs/TEs (PPR: 1 per catch, 0.1 per yard, 6 per TD). Each prop sits in its most specific group, so a prop in tab 3 is not also in tab 1.
Model update (Sun Oct 25). On 385 settled Week 6–7 lines priced 40–60¢, the Yes side came true 45% of the time against a 50¢ average price; buying every Yes lost 13%, buying every No made 9%. The >50% hit-rate filter didn't help (44.5% vs 45.8%), and the old Wedge % ran about 5 points high. Wedge % is now recalibrated on those lines: a logit shift by market (rushing and anytime-TD overs down about 9 points at 50¢, receiving about 2, passing up about 8) and for players with under 12 games for their current team (down about 12). 5-fold cross-validated Brier 0.2417 vs 0.2490 before (price alone 0.2505); trained on Week 6 and tested on Week 7, 0.2345 vs 0.2394. The chip now names the better side (Yes or No) when it clears its price by 3+ points. Two weeks is a small sample; the shift is re-fit as each week settles.
Live check and re-fit (Mon Oct 26). Sunday's 12 games were the first real test: 487 lines posted at ~11:56 AM, before kickoff. Wedge picks (best side, 3+ pts) went 148–101 (59.4%), +5.4% on what they cost: No picks 144–92 (+6.5%), Yes picks 4–9 (−22%); picks with 8+ pts of edge 68–47 (+11.5%). For comparison, every Yes −5.7%, every No +1.3%, the >50% hit-rate filter's Yes picks 42–54 (−13.4%). As a probability, though, v2 ran too low: on 40–60¢ lines it averaged 43% and they came true 47.8% (price 49.8¢), Brier 0.2558 vs 0.2529 before and 0.2523 for the price; rushing overs were the worst miss (36% vs 44%), and the thin-history penalty flipped (those players hit 29% in Week 6, 49% in Week 7). Re-fit on all 678 settled Week 6–7 lines with the thin term dropped: at 50¢, receiving 46%, rushing 41%, anytime TD 43%, passing 58%. Trained on one week and tested on the other it beats the old blend and the price both ways (Week 7: 0.2497 vs 0.2506 and 0.2511; Week 6: 0.2461 vs 0.2508 and 0.2513). The margins are small and two weeks is still a small sample. DAL @ PHI uses the re-fit; finished games keep their posted numbers.
Model v3 and the momentum study (Tue Oct 27). With all 14 Week 7 games settled (Monday night's DAL @ PHI went 5–9 for the picks; the week finished 153–110, +3.8%), our own matchup read showed no correlation with result minus price on 706 real Week 6–7 lines, so Wedge % now leans harder on the market: 85% books, 15% our read (trained on one week, tested on the other: Brier 0.2474 vs 0.2480 at 70/30; price alone 0.2512). Calibration re-fit on all 706 lines: a Yes priced 50¢ comes true about 45% for receiving and anytime TDs, 40% for rushing and 56% for passing. New "Recent role" factor: when a player's snap share over his last 4 games moved 10+ points from his season share, his volume moves with it (2023–25: 34–55% of the gap carried into the next game after a role change vs 6–21% otherwise). Weighting the last 4 games heavily (the last-4 view) tested worse on real prices (0.2487), so it is shown as a filter, not used in Wedge %. Week by week at real prices (Charts): every Yes 45–55¢ lost 6.6% over Weeks 1–7 and every No made 4.4%; passing props hit 60% (54–36), rushing props 39%, game-total overs 56%, favorites on team lines 40%.
Model v4, roles and the defense study (Tue Oct 27, evening). Projections now blend the usage cases 50/50 with recent volume × long-run efficiency: volume is an exponentially weighted average with a half-life of about 4 games for WR targets, RB carries and QB attempts (longer for TE and RB targets), prior seasons discounted to 30–50%; efficiency uses a 12-game-plus window shrunk toward the position average by 100–200 targets, carries or attempts. The defense factor now measures what a defense allowed against each opponent's own average instead of raw yards allowed: across 2023–26 the raw number barely predicted the next player (WR receiving: correlation 0.00), the adjusted one a little better. For rushing, only the extra carries a defense allows and the yards before contact its front gives up carry over; what the runner did after contact (one big Gibbs-style day) does not (slope 0.04 vs 0.19). A role-and-matchup adjustment, fit only on 2023–25, nudges Wedge % by how players in the same role, game script and scheme fared vs the line (real Week 6–7 prices: Brier 0.2467 vs 0.2472 without it). On real prices our own read still adds little to the books' price, so these are small changes; the biggest gains come from roles and spots (Charts).
Wednesday update (Wed Oct 28). Prices re-pulled from Robinhood, Kalshi and five sportsbooks at ~2:31 PM CT: Robinhood now lists player props for most Sunday games, so 220 of the 408 lines priced 45–55¢ are Robinhood's own. Injury statuses updated from the Packers' and Panthers' final report and Wednesday news; weather re-pulled. Model, roles and the defense history are unchanged from Tuesday evening.
Thursday update (Thu Oct 29). Prices re-pulled from Robinhood, Kalshi and five sportsbooks at ~5:15 PM CT: Kalshi now lists player props for every Week 8 game (LV @ NYJ, CLE @ PIT, ARI @ DAL, PHI @ WSH and CHI @ SEA were added), so 380 of the 459 lines priced 45–55¢ are Robinhood's own. Injury notes now use the official Wednesday practice report for every Sunday team (defenders too, on the Matchup tab), with Thursday news on top; weather re-pulled. Model, roles and the defense history are unchanged from Tuesday evening.
Monday update (Mon Nov 2). CHI @ SEA re-priced at ~3:15 PM CT from Robinhood, Kalshi and five sportsbooks, with the Bears' and Seahawks' final injury report and the current rosters. Every game already played keeps its Thursday Oct 29 numbers, so the board shows what was posted before kickoff. Model, roles and the defense history are unchanged from Tuesday evening.
Wednesday update (Wed Nov 4). Every Week 10 game re-priced at ~3:37 PM CT from Robinhood (73 game-line events), Kalshi's player-prop ladders (now 14 games; TB @ CHI added) and five sportsbooks, with ESPN's current injury statuses and Wednesday's practice news; weather re-pulled. The Lines tab now has an every-line view: all 5,722 Week 10 lines at any price, each with its 2026 hit rate. Inside 45–55¢ nothing changed about how Wedge % or picks work. Outside it, Wedge % applies the same market shifts, which were fit on lines priced 40–60¢, so it is a rough guide there and no picks are made. One fix: the situational splits had been reading last week's injury file, so this week's outs (for example a team's starting QB) were missed in the QB and lineup splits; they now read Week 10's.
Saturday update (Sat Nov 7). Thursday's JAX @ BAL is final (Jaguars 33–10), so its lines are off the board. The 14 Sunday and Monday games were re-priced at ~2:25 PM CT from Robinhood (70 game-line events), Kalshi's player-prop ladders (now all 14 games, BUF @ MIN included) and five sportsbooks, so 448 of the 496 lines priced 45–55¢ are Robinhood's own. Injury notes use every team's final report (Friday; the Bills and Vikings through Saturday: Kyler Murray and Jauan Jennings are out for Monday) and ESPN's latest news; weather re-pulled from game-day forecasts. Model, roles and the defense history are unchanged.
Model v5 and the Weeks 6–8 backtest (Tue Nov 3). We settled every Week 8 line we posted (894 lines across all 15 games, from nflverse box scores and snap counts) and re-tested the model on all 1,333 settled Week 6–8 player lines at real prices, holding out one week at a time. Seven problems turned up. (1) Our matchup read pointed the wrong way: how far it leaned from the price was negatively correlated with how lines settled in every week (−0.03, −0.04, −0.08; −0.29 on rushing). (2) We calibrated the blend of market and read when the price itself was the better thing to calibrate: shifting the Robinhood price by its market's over bias scored a Brier of 0.2460 held out, vs 0.2471 for last week's model and 0.2493 for the raw price (Week 8 alone: 0.2459 vs 0.2510). (3) Receptions were lumped in with receiving yards, and passing TDs and interceptions with passing yards; at about 50¢ receptions came true 48% of the time, receiving yards 45%, rushing 39–42%, passing yards 62%, interceptions 56% and passing TDs 50%. (4) Our projections missed by more than the line itself (average miss on real lines: rushing 25.4 yards vs 21.7 for the line, receiving 24.4 vs 23.2, passing 58.2 vs 56.3). (5) Thursday prices for Sunday games scored worse than Sunday-morning prices (0.2164 vs 0.2098), and late moves kept going rather than reversing; the Sunday re-price was built but never went live. (6) Picks per pull ballooned (59, 167, 275) as Kalshi's ladders grew, and bigger edges didn't win more often than small ones. (7) The certification key still said Week 7, so four Week 8 certified picks never settled; fixed, and they now show under Week 8 on the Results tab.
What changed. Wedge % = the price shown (Robinhood's own, or Robinhood-style from the books) shifted in log-odds by its market's over bias (passing yards +0.22, interceptions +0.13, passing TDs +0.01, receptions −0.08, receiving yards −0.19, rushing −0.31, anytime TD −0.02), fit on Weeks 6–8 with shrinkage; our read is still shown in the outlook. Held out a week at a time, its picks (3+ pts vs the price) went 283–208 (57.6%), +12.9% before fees; the Wedge picks actually posted in Weeks 6–8 went 248–190 (56.6%), +4.0%. Projections: we took the random-features ridge regression from Kelly, Malamud and Zhou's “The Virtue of Complexity in Return Prediction” (AQR; Journal of Finance 2024) and AQR's “Understanding the Virtue of Complexity” (2025): 3,000 random sine/cosine transforms of the inputs, fit with heavy ridge shrinkage and averaged over five draws, which picks up non-linear effects without hand-picking them. It is trained on 2023–26 player games to predict how far he lands from a book-style line, using recent volume and efficiency, his usage profile (target and carry share, depth of target), role, the spread and team total, the opposing defense's coverage, blitz and run-defense numbers, his offense's pass rate, and weather. Walked forward, it beat our old projection on passing, receiving and rushing yards and added 2–6% of explained variance over the book-style line. On real lines it improved only passing (average miss 54.7 vs 56.3 for the line), so the best estimate is the line moved toward it by a weight fit on Weeks 6–8: passing all the way, receptions half, receiving yards a quarter, rushing not at all. We also checked Moskowitz's “Asset Pricing and Sports Betting” (AQR), which finds sportsbook line moves overreact and partly reverse; on Kalshi's Week 8 props the moves continued, so v5 doesn't fade them. Three weeks is a small sample: the shifts are estimates and will move as more weeks settle.
Wedge % and the outlook: every regular-season game a player has played for his current team since 2023 (2026 through Week 8). Usage cases: his clean games split into thirds by volume (targets for receiving lines, carries for rushing, touches for TDs, dropbacks for passing); each case's chance is his hit rate in those games nudged toward his overall rate, and this week's case odds come from the splits that look like this week (QB, lineup, No. 1 competitor, defense), weighted toward 2026. Projection: the case average times this week's adjustments: defense vs his position (2026 allowed per game vs league, shrunk toward average for small samples), coverage (yards per target allowed by the defenders likely to cover him, using the injury report), game script (his results when his team won by 8+, stayed within 7 or lost by 8+, weighted by the odds from the spread), team total (implied points vs points per game) and weather (wind, cold and rain for outdoor games); combined adjustments are capped at −30% / +35%. Our read averages the cases' chance with a distribution priced at the projection, shrunk toward 50% (calibrated on 2025); Since Week 10 (v6), Wedge % is the market midpoint moved by the backtested matchup lens; last week (v5) it was the price shifted by its market's over bias, with our read shown for reference (Week 8: 85% market and 15% our read; before that 70/30). Recent role: when his snap share over the last 4 games is 10+ points off his season share, the projection scales with the square root of his change in volume (targets, attempts) or moves 75% of the way to his last-4 rushing average, capped at −20% / +25%. Backtest: on 3,436 player lines from 2025, walked forward week by week, our read scored a Brier of 0.238 vs 0.250 for a coin flip and 0.246 for his last-8 hit rate. On 334 real Week 6 prices, Wedge % scored 0.2483 vs 0.2486 for the price alone and 0.304 for the season hit-rate screen: the market already prices most of this, so small edges are noise. Games he left early hurt and final-week games he was rested are left out. ★ marks games whose game total implies more than 49.5 points. Data: nflverse weekly stats, play-by-play, snap counts, injury reports, rosters and depth charts; NFLWeather and RotoWire forecasts (Week 10); ESPN injury statuses; Action Network, VegasInsider and Covers book lines; Kalshi prices.
Flags: "No bid" means only an ask is posted; "Wide spread" means more than 5¢ between bid and ask; "Small sample" means fewer than 3 games; "Inactive" means Robinhood has closed that contract. Sources: Robinhood prediction markets (prices, contract terms), ESPN box scores, game logs and rosters, and nflverse play-by-play (red-zone and target data).