Find the real matchup edge.
The six-factor trench-heavy approach used for the Memphis–Tulane handicap: rushing mismatches, line play, passing matchups, third downs, recent form and schedule quality, plus verified availability. No sportsbook blending.
Pregame data filters
Applied to eligible prior games used for success rates, third downs, trench statistics, recent form and strength of schedule.
1. Single Game Lab
Pick any scheduled matchup, set independent model weights, and remove specific prior games from either team’s evaluation.
2. Football model comparison
Analyze all FBS games at once, optionally run both engines across the full date, then apply the filters to every matchup.
| Matchup | Spread | Winner / model confidence | ATS pick / cover rate | ML odds / edge | Agreement |
|---|---|---|---|---|---|
| Analyze a date to show full-slate picks. | |||||
Betting review & two-team parlays
Export the complete slate for analysis, or inspect model-selected moneyline combinations across separate games.
Single-game moneyline value shortlist
Uses currently qualifying games and the selected model. The displayed stress test halves the model's estimated percentage-point advantage over the quoted market break-even rate. This is a sensitivity check, not calibrated probability or verified ROI.
Two-leg moneyline parlay candidates
Pairs different football games from the same identified sportsbook, using actual quoted moneylines on the model-selected sides. Requires each leg to be positive value after the 50% edge haircut. Estimated joint win chance is an independence assumption, not a correlation model. No parlays are offered without prices on both legs.
Research tool, not a recommendation to bet. Simulation confidence is historically overconfident, and historical ROI filters were selected after observing results. CFBD prices may be stale or lack a verified pregame timestamp. Confirm both prices and kickoff status before using any selections. Parlay projections assume independence and are less reliable than straight-bet projections.
ATS frequencies are for covering; ML edge = simulated outright win frequency minus sportsbook break-even probability. Both are uncalibrated research estimates, not proven betting edges. Moneyline odds must be supplied by the bookmaker; missing odds are not inferred. Use Run both models for entire slate to share historical drive retrieval across games. Drive processing can take longer on big slates; unavailable game profiles are clearly marked.
3. Slate matchup analysis
Select a matchup from the slate.
Model weights
30/20/20/15/10/5 default · Renormalized when a factor is unavailable.
The listed percentages are starting assumptions, not fitted from historical games. Changing them recalculates the currently loaded slate and matchup.
This controls how strongly offensive performance versus defense-allowed performance influences matchup scores and baseline points. Default 65/35 is an experimental choice; change to 50/50 with the slider if preferred. No sportsbook lines enter the football score forecast.
Definitions and safeguards
The six matchup components and what we do not assume.
Rush matchups — 30%: success and allowed-success rates, with weaker defenses correctly adding to expected rushing efficiency; adjusted for play volume.
Trenches — 20%: offensive/defensive line yards, stuff rates, front-seven havoc, and sacks per dropback where available. These are imperfect proxies for line play.
Pass matchups — 20%: passing success and defensive passing success allowed in the correct direction, adjusted for pass volume.
Third downs — 15%: offense conversion rate and the opponent’s defensive conversion rate allowed; higher allowed now correctly means weaker defense. Red zone remains excluded.
Recent form / SOS — 10%: last two eligible game margins and opponent-strength quality from prior games. Missing measures are dropped rather than fabricated.
Availability / context — 5%: optional manual verified injury/context rating. Home field receives a separate single fixed 1.5-point adjustment; it is never double-counted. Without injury details, the 5% category is unavailable and remaining factor weights are renormalized.
FCS games: excluded from ratings and, by default, efficiency stats. 35+ point FBS blowouts: excluded from efficiency by default, retained for team strength. Game lines are comparison-only; archived odds may not have verified pregame timestamps.
Limits: this is a structured approximation of the Memphis handicap, not a reproduction of its manual 500,000-trial projections. The website runs 10,000 seeded experimental scoring trials; the confidence percentages are not calibrated.
Basketball Matchup Lab
Possession simulations for regular-season games and preseason projections. Compare last year's playing style, roster continuity, incoming transfers and recruits, and manual team adjustments.
Preseason builder · no scheduled games required
Choose any two teams and project the upcoming season from last year’s box scores, current roster, portal and recruiting records. Season 2027 = 2026–27.
1. Select an NCAA matchup
Basketball seasons use the ending year (2026–27 = season 2027). The preseason strength baseline requires verified opponent-adjusted CBBD ratings.
Matchup
Uncheck earlier games independently for each team, including last-season games when running preseason projections.
2. Select historical games
Every change is applied to the actual statistics used in the next simulation.
Preseason roster & play-style settings
Confirmed arrivals get small rating proxies; when rosters are missing, listed commitments get discounted provisional proxies. Check which players to include and edit the projected roster.
Roster matches determine the suggested returning-minutes share, but it is editable. Transfer/recruiting bonuses are deliberately small; listed but unconfirmed commitments receive only a discounted assumption. Returning players, transfers and freshmen are listed with source-coverage notes. Missing records are never counted as zero or confirmed departures. Recent-form weighting is disabled in preseason simulations.
3. Possession simulator
The final prediction comes from the simulated final scores, not a pre-filled projection.
Basketball factor presets
Six presets. Each factor influences the matchup simulation, rather than simply changing a displayed label.
Factor shares normalize to 100%. Preseason opponent-adjusted efficiency is a REQUIRED scoring anchor and does not disappear when a slider is set to zero; shooting, rebounding and turnovers tune play styles and small matchup residuals. Recent form is disabled in preseason; projections remain uncalibrated.
NFL Matchup & Prop Lab
Matchup-aware game outcomes, plus player-specific rushing, receiving and passing prop simulations. Uses earlier completed games only and lets you adjust usage and player game samples.
1. Choose NFL game
Select any regular-season week and match. Use the previous games to model this exact opponent.
Matchup
All stats are as of before selected game.
2. Game simulation
Offense against its opponent’s defense, including run blocking versus run defense, protection versus pass rush, coverage, and home advantage.
Trenches & coverage matchup lab · adjustable
Measured baseline: rush yards/carry, pass yards/attempt, completions and interceptions. Sacks are shown only when published. These controls are manual percent-strength adjustments, not NFL player grades. Positive = stronger unit. ±20 is a sizeable assumption.
4. Single player prop simulator
Uses the player's previous game logs, opposing defense and expected team drives. Choose any available player.
NBA Matchup & Prop Lab
Pregame opponent-aware NBA simulation, plus points, rebounds, assists, threes and combination props. Earlier completed games only; prior-season data available for opening week.
1. Choose NBA game
Pick a date, load the teams and their last completed games.
NBA historical checks & export
Review earlier dates, run the game model, and grade saved picks. Lines are manually entered home-team spreads (home favorite −5, home underdog +5). No line means no ATS grade.
Pregame labels refer to prediction time only; manual spreads are not verified closing lines and may be entered after tipoff. Saved on this browser/device; export a backup before clearing browser data. Historical reruns use earlier team games but today’s model/settings, and are not saved pregame predictions. Confidence is an uncalibrated simulation frequency. Model scores are regulation only; actual results include overtime. Pushes and pending games are excluded from hit rate. One first-saved prediction per game; later reruns do not overwrite it.
| Date / matchup | Source | Home spread | Projection | Pick | Model % | Final (away–home) | Result |
|---|
2. Full-slate NBA player props
Simulate every listed player across the selected NBA slate, then filter by position or stat. Game and player histories stop strictly before each matchup's tipoff.
| Player · position | Proj. | 80% range | Opponent | MIN | USG* | DEF | Model |
|---|---|---|---|---|---|---|---|
| Run the slate to load players. | |||||||
USG* represents an estimated shot-involvement proxy, not official NBA usage. Historical player stats (including prior teams for trades) and team-level matchup defense drive the model.
NBA matchup
Data cutoff is before tipoff.
3. Game simulator
Match each offense against its opponent's scoring defense. Exclude any historical team games you don't want.
3. Player prop simulator
Simulates player production using historical minutes, game logs, projected team scoring and the opposing defense.
⛳ PGA Golf Lab
Simulate every player together across four rounds. Change strokes-gained inputs, course fit, field selection and cut rules. Results come from simulated rounds, not preset percentages.
1. PGA tournament and field
2. Player strokes gained / round and manual adjustments
Editable values: approach, off tee, around green, putting, recent form and course fit. Unavailable values start at zero; verify before relying on results.
| Golfer | Approach | Off tee | Around | Putting | Form | Course fit |
|---|
3. Tournament settings
4. Simulated tournament outcomes
Top finishes are rank-based; no-cut players count as missed cuts. Scores for missed-cut players are hypothetical 4-round totals, not official tournament scores.
| Player | Win | Top 5 | Top 10 | Top 20 | Make cut | Projected strokes* | R1 avg | R1 under |
|---|
NHL Matchup Lab
Simulate offensive-versus-defensive scoring matchups using verified regular-season scores, recent results, rest, manual xG and goalie adjustments. Detailed shots and power-play game logs are temporarily disabled to reduce data-service rate limits. Regulation, overtime and shootout are handled separately.
1. Choose NHL game
Pick a date, choose an upcoming or completed regular-season matchup and load the teams' previous games. Previous-season history supports October matchups.
NHL matchup
Historical data cutoff is before game day.
2. Historical sample controls
Remove specific outliers, mismatched opponents, or games with unusual goaltending. All values update when you rerun.
3. Game assumptions
Blank or zero xG fields are ignored. To use xG, enter all four positive figures or import team xG below. Positive GSAx means better-than-expected goaltending; selecting a goalie does not confirm a starter. Offensive changes and penalty/shot sensitivities remain optional manual scenarios.
Automatic NHL advanced statistics
Team xGF/xGA and shots can be loaded from published MoneyPuck season files. Goalie names now load independently from NHL/ESPN rosters when a matchup opens. GSAx still requires published MoneyPuck data or a manual value; no starting goalie is confirmed. For completed games, the importer uses only the prior completed season to prevent lookahead bias. For upcoming games, current-season stats may be blended with the prior season. The goalie list is not a verified starting lineup. Offensive lineup changes, shot sensitivity and penalties remain manual assumptions.
Source: MoneyPuck.com data. Free for noncommercial use with attribution; other uses require permission. No key required. A provider response, license, or schema change may prevent syncing.
Analysis lines (not used to anchor the model)