ROBOT
POLL
WEEK 8 · LOCKED TUE OCT 20, 9:04 AM
286–102
STRAIGHT-UP · 73.7%
vs closing line 182–194–12 · 48.4%
BRIER .190
CALIBRATED ✓
Model-generated rankings, picks, and analysis. Every prediction includes its signal breakdown, is locked before kickoff, and is graded against results.
MODEL VS MARKET — 3 LARGEST WIN-PROBABILITY DIVERGENCES
Ranked by disagreement over who WINS the game (model win probability vs. the spread-implied probability). The Betting Analysis below ranks by a different question — edge against the posted number — so its order can differ.
ARMY +2.0
+22 PTS APART ON WIN %
Largest signal contribution toward this side: rating_gap at +12.7 points of lean. Model and market win probabilities differ by 22 percentage points.
UTAH -6.5+19 PTS APART
Market 66% · Model 85%
Largest signal contribution toward this side: rating_gap at +12.5 points of lean. Model and market win probabilities differ by 19 percentage points.
COLORADO +13.5+18 PTS APART
Market 19% · Model 37%
Largest signal contribution toward this side: home_field at -1.9 points of lean. Model and market win probabilities differ by 18 percentage points.
Market-anchored: the consensus line plus the model's residual signals (rest, travel, lookahead, defensive trends). A bet appears only where those residuals move the number enough to clear the vig-implied break-even (52.4%). Each breakdown shows the market's share and every nudge. Stakes sized by quarter-Kelly, |spread| ≤ 21.
⚠ HONEST DISCLOSURE: the spread model is anchored to the market (consensus line + residual signals). Walk-forward 2016-2025, at the card's threshold it went 475–427 (52.7% vs. the line, 69% favorites); its best threshold reached 52.7% on 902 bets. Break-even is 52.4% — so it BEATS the vig. Paper only. The proven product is the calibrated win probabilities and rankings above; real money is your call, not the robot's suggestion.
BET $1 · South Alabama -3.0 @ Marshall
edge +2.2LOG
Tue Oct 20 · 7:30 PM ET · model covers 54.6% · market needs 52.4% · breakdown ▾
LEAN +4.8 (this side)50% to win outright
remaining games: below edge threshold
$99 UNALLOCATED
Paper trading only. These are tracked model outputs, not wagering recommendations.
HISTORICAL PERFORMANCE — BACKTEST VS RESULTS
Every season below was predicted walk-forward: the model only used data from seasons prior to the one being predicted.
| SEASON | GAMES | STRAIGHT-UP | VS THE LINE | ATS % | BRIER | ML ROI (FLAT $) |
| 2016 |
760 | 70.3% | 358–382–20 | 48.4% | .190 |
no ML data |
| 2017 |
776 | 70.4% | 390–365–21 | 51.7% | .185 |
no ML data |
| 2018 |
772 | 72.4% | 391–361–20 | 52.0% | .175 |
no ML data |
| 2019 |
774 | 73.1% | 377–384–13 | 49.5% | .176 |
no ML data |
| 2020 (COVID — excluded from headline metrics) |
534 | 71.2% | 274–251–9 | 52.2% | .187 |
no ML data |
| 2021 |
770 | 69.5% | 383–378–9 | 50.3% | .187 |
-4.0% |
| 2022 |
776 | 68.7% | 389–374–13 | 51.0% | .199 |
-4.6% |
| 2023 |
792 | 72.3% | 395–381–16 | 50.9% | .185 |
-1.2% |
| 2024 |
798 | 69.2% | 391–388–19 | 50.2% | .193 |
-5.4% |
| 2025 |
808 | 72.5% | 396–394–18 | 50.1% | .184 |
+1.0% |
Moneyline analysis, 3,771 games (2021-2025): flat bets on every straight-up pick returned -2.8%. Two subsets were positive: model-backed underdogs (470 bets, +11.4%) and picks with a ≥5-point modeled edge (968 bets, +2.9%). Both samples are small and were identified alongside other tested strategies, so they are tracked as paper-trading experiments pending live confirmation.
SIGNAL LAB — SUBSET BACKTESTS
TOGGLE SIGNALS · 10 SEASONS OF PRECOMPUTED RESULTS
HOME FIELD ✓
RATING GAP ✓
REST DIFF ✓
OPP PASS DEF TREND ✓
RUSH DEF TREND ✓
PASS OFF TREND ✓
LOOKAHEAD SPOT ✓
TRAVEL KM ✓
Every automated run reports here — including whether lines moved. If this goes quiet, the automation is down, not the market.
loading…
ROBOT POLL · regenerated weekly · predictions lock at publication and are not edited
2020 excluded from headline metrics · paper trading only · not wagering advice