Experimental sports research. Saved prices may differ from your sportsbook. Estimated probabilities are not guarantees. Your filters and watchlist are stored in this browser; clearing browser data may erase them. Shared data and models are read-only. No ads or payments are active.
Install 4thLook on your phone
Open 4thlooksports.com in your phone’s browser.
Android · Google Chrome
- Open the site in Chrome and tap the three-dot menu (⋮).
- Tap Add to Home screen, then Install or Install app, depending on your Chrome version.
- Confirm the prompt, then open 4thLook from its new icon. If Chrome offers Create shortcut instead, you can use that to add a Home screen shortcut.
If you opened the link inside another app, first choose Open in Chrome. You can also use the Install app button below when Chrome makes it available.
iPhone · Safari
- Open the site in Safari and tap Share.
- Tap Add to Home Screen.
- Leave Open as Web App enabled if shown, then tap Add.
The free beta does not require a ChatGPT login. The Home screen app uses the same saved data and update schedule. An internet connection is required. Your watchlist is stored on this device and may be separate between browsers or installed copies.
Latest NFL headlines
Research tool only · 21+ — Supplements your existing knowledge. Not recommended for beginners. Estimates do not guarantee wins or profit. Read disclaimer.
Choose a game to view its saved props. Near-duplicate lines are combined into one selection per player, category and over/under direction. The highest estimated success is kept for the selected probability source; estimated value breaks ties.
Model validation & limitations
Load the statistical model to view its held-out 2025 test results.
Top 10 Picks
Spreads, moneylines and game totals, with one pick per game. Each view uses its own probability and EV with the same value scoring as Player Props Best Value. Negative EV can appear unless filtered out.
Value formula & rules
Score / 100 = 75 × selected probability + 25 × value factor. Value factor = (EV clipped to −20%…+20%, plus 20%) ÷ 40%. Consensus uses consensus probability and consensus EV; experimental uses model win rate and model EV, including model push estimates. Consensus EV does not model pushes. Requires probability above 50%, quotes within 72 hours and qualifying availability. The score is experimental, not a win probability or a validated betting advantage. Team model estimates remain unadjusted for injuries, roster changes and weather. Different-game picks can still be dependent.
Top 20 by estimated success
Reload the shared snapshot to view saved player props.
16 requested categories: passing yards, passing touchdowns, completions, pass attempts, interceptions thrown, longest completion, rushing yards, rush attempts, rushing touchdowns, longest rush, receiving yards, receptions, receiving touchdowns, longest reception, passing + rushing yards, and rushing + receiving yards. Availability varies by game and sportsbook. Consensus uses supported two-sided over/under prices only. The experimental model covers the original five categories plus selected expanded categories and positions that passed historical screening. Longest-play categories remain consensus-only. See Model validation & limitations for current coverage. Pregame only; finished / in-play games are excluded. Saved quotes remain visible with their age. This is not a historical weekly leaderboard. Coverage depends on the provider and model history. Fewer than 20 bets are shown if fewer qualify. Individual and weekly props share saved results in this browser, including across page reloads when storage is available. Saved quotes can be replaced by newer shared snapshots or manual updates; old prices may no longer be available. Model and availability data load during discovery without betting credits. Scheduled background updates collect shared data. Reload the shared snapshot to view the latest saved results; manual price updates are separate.
My Watchlist
All your saved bets in one place, across categories and weeks. Saved on this browser. Saved quotes may no longer be available.
Watchlist parlay estimate
Uses every pick’s saved odds. Total return includes your stake; profit excludes it. Same-game or related picks may be repriced or rejected by the sportsbook. Different books and older quotes may not be available together. This estimates payout if every leg wins, not the chance of winning or expected value.
Reported player availability
Press to refresh injury reports and confirmed player availability. Uses no betting API credits.
Reported statuses are flags, not medical clearance. Questionable players remain listed with a participation-conditional estimate. Out, doubtful, unavailable or unclear statuses are withheld. Team predictions are not injury-adjusted. Background injury updates continue. Visitors reload the latest saved report; the owner can request an additional free injury check.NFL official reports and ESPN injury/eligible-roster data. Green means Probable · inferred, Cleared, or Confirmed active. Probable · inferred is our assumption for a verified eligible-roster player with no injury designation in the checked feeds; it is not an official NFL designation or confirmation they will play. Missing or incomplete rosters never establish probable status. New injury or inactive reports override the assumption. Questionable players stay in rankings with a warning; estimates assume they play with similar workload and do not estimate their chance of playing. Out, doubtful, reserve/suspended and unclear statuses withhold statistical player prop predictions. Game-day active confirmation requires an eligible roster and complete official inactive list. Newer injury updates override earlier clearance. Active does not guarantee snaps or full workload.
Weekly scoreboard
Scores, records and game status from ESPN. Completed games are excluded from betting rankings.
Historical game lines & model evaluation
Free historical regular-season lines, prices and final results for 2025 and completed 2026 games. Separate from forecasts saved before kickoff.
About this research tool
4thLook provides access to compiled football data, research, and experimental model estimates for independent analysis. Rankings highlight selections to investigate, not instructions to wager. You decide whether to bet. See the research use disclaimer for limitations and risks.
Probability is an estimate, not a promise.
Game bets use Hard Rock Bet Florida. Prop bets show the actual source book: preferably Hard Rock via PropLine (Florida unverified), otherwise a labeled alternative. Sportsbook consensus averages margin-removed probabilities from Hard Rock, DraftKings, FanDuel and BetMGM in the saved snapshot. At least two books must have both opposing prices for the exact same player, market and line; different lines are never combined. Consensus is unavailable when fewer than two books match. Saved quotes can be old and consensus is not a guarantee.
Game selections display saved Hard Rock Bet Florida quotes. Player props display the selected source book, preferring Hard Rock via PropLine with Florida availability unverified, or a labeled alternative. These are not a market-wide best-price comparison. For a bet without a push, expected value per unit staked = estimated win probability × decimal payout − 1. When pushes are modeled as refunds, EV = win probability × (decimal payout − 1) − loss probability. A high probability does not necessarily mean a good price. Different point lines remain separate; we do not interpolate between them.
The consensus column averages matching sportsbook probabilities after removing each book’s margin. The separate experimental model uses historical game and player statistics, with team scoring coefficients fit on 2021–2023, error distributions from 2024, and a held-out 2025 test. Reported injury statuses flag availability, retain Questionable players with conditional estimates, and withhold unavailable or unclear player props; statistical forecasts do not quantify injury impact or weather effects. Player projections use recent recorded appearances and usage; opponent-defense adjustments are not included. Market probabilities assume a decisive outcome and exclude pushes. Statistical estimates report win, loss and push probabilities separately; model EV includes refunded pushes. For two-way moneylines we treat simulated ties as refunds; check actual sportsbook rules. Related bets are correlated. Team scores are approximated by rounded historical prediction errors; football scoring combinations and key margins such as 3 and 7 are not explicitly modeled.
Saved quotes for upcoming pregame events remain visible regardless of age; each row shows the quote time and age. These are saved prices, not guaranteed current offers. Live/in-play games are visible in scores but not ranked because asynchronous prices can mislead. This site collects shared betting prices on its scheduled update cadence and through manual price-update buttons. Provider update frequency does not determine how current a saved quote is. Displayed quotes can become stale between requests.
Provider coverage is not all NFL data or every sportsbook. Historical weeks retain schedule access. A separate free archive provides available regular-season game lines, prices, and results for 2025 and completed 2026 games; it is not a complete archive of every pregame quote. Historical player-prop lines are not included. Historical model tests are available under Model validation & limitations. Player tests use generated thresholds, not archived sportsbook prop lines. No profitability claim is made.
Provider documentation ↗Prediction history
How are the predictions performing?
Pregame forecasts are preserved separately from price refreshes. New forecasts also archive dated injury reports, roster transactions, news and recent-versus-season production. Snap counts and verified depth charts are not collected; context does not automatically adjust probabilities. Public NFL results settle supported bets without sportsbook credits. Missing player records and book-specific void or tie rules stay unresolved. Model changes require later, unseen-game testing.
Data updates & schedule
Through October 31: Wednesday and Sunday 11 a.m., Saturday 5 p.m., America/New_York (Eastern local time). New betting prices are requested only by the scheduled updater. The model, scores and availability can still be loaded from free sources. With PropLine handling props, a scheduled game-odds update budgets 3 The Odds API credits; insufficient balance skips the run and preserves prior data.
Player props refresh from PropLine hourly, without using The Odds API credits. Hard Rock’s PropLine feed is not yet verified as Florida-specific. Player/team information refreshes daily at 9 a.m. Eastern and Sunday noon. Statistics update model estimates; injuries affect prop eligibility. Headlines provide context only. No numerical injury, replacement or weather adjustment is included.
Checking daily information update…
Data & model details
Data connection & private access
Connect The Odds API to compare Hard Rock Bet Florida only. Coverage varies by book, state, market and subscription. Your session key is sent only to this site's server and the provider. It is kept in this tab's memory, cleared on reload, and never written to browser storage. A server secret named ODDS_API_KEY can provide a shared feed once configured.
This site requires authorized ChatGPT access. Friends must be added through the site's access settings; forwarding the URL alone does not grant access. No sportsbook login is required.
Each refresh can consume API credits. Individual player props load for the game you select; the Top 20 Player Props tab can manually request all available upcoming games for that week. Scheduled background updates also collect shared snapshots; manual updates are additional requests. Each explicit Update game prices request covers moneyline, spreads and totals for four selected sportsbooks (one billing group): normally 3 credits. Player prop price updates use up to 5 credits for the selected game. Viewing saved props and Top 20 Player Props costs no credits. Updating game prices never refreshes player props automatically.Game odds use Hard Rock Bet Florida. Player props prefer Hard Rock via PropLine (Florida unverified); when unavailable, DraftKings, FanDuel or BetMGM may be displayed with their actual book name. Historical model training and simulations use NFL performance data, not sportsbook prices. Availability of each book or prop varies. The saved four-book pool supplies consensus; displayed game offers remain Hard Rock Bet Florida only, while player props can use the labeled alternatives described above. Scheduled updates collect these four books in one billing group. Saved data is reused; this change does not request new prices.
If this bet played out 10 times…
Estimate wins, losses and pushes using 10,000 simulated outcomes based on prior NFL results and player performance. “6.3 / 10” means an average expectation across many sets of ten, not a guarantee of six wins. The model has not demonstrated a profitable betting advantage.
Model not loaded. Load the saved snapshot, or manually update public NFL statistics.
Assumes players participate with similar roles. Availability flags retain Questionable players with warnings and withhold unavailable or unclear player props; no injury impact, weather, depth-chart or opponent-defense adjustments. Predictions abstain on sparse or stale histories. Team model uses scoring, opponent scoring history, rest and venue.4thLook Research Beta
For adults 21 and older. Experimental research, no guaranteed outcomes. Terms of Use · Privacy.