2026 SEASON · EVERY GAME GRADED

Called it

TapeDog's track record. Every finished FBS game this season, scored the way TapeDog would have scored it and checked against the final.

48%of our flagged favorites lost

31%expected by the betting lines

27 games, 2026 season. With a sample this small the true rate could be anywhere from 29% to 67%, so treat it as early evidence. Graded after the fact using only pregame information; our live record, frozen at kickoff, is building now.

27Games flagged FADEAll favorites lost 20% of the time

65%Of FADE underdogs coveredAll underdogs: 47%

Live record

Calls saved before kickoff since Nov 4

7Live FADE calls graded

29%Favorite lost outright

43%Underdog covered

This is the honest number: picks frozen before kickoff, never re-scored. The backtest below is for testing ideas.

Model lab

Same games, different rules

RuleGamesFavorite lostLines expectedDog covered

All favorites Every graded game69320%23%47%

Current FADE TAPE score 50+2748%31%65%

Stricter FADE TAPE 50+ and 3 or more flags2748%31%65%

Two kinds of evidence A schedule spot and a market signal26730%30%52%

Close lines only FADE with a spread of 10 or less2152%35%70%

CALIBRATION · DO HIGHER TAPE SCORES LOSE MORE?

16%TAPE 0–27
515 games · lines 20%

30%TAPE 28–49
151 games · lines 29%

52%TAPE 50–69
25 games · lines 31%

0%TAPE 70+
2 games · lines 36%

Every rule here uses only what was known before kickoff, and every rule runs on the same games. The fair test is "favorite lost" against "lines expected" for the same games: beating all favorites isn't enough, because close games get upset more often anyway. We tried several rules on these games, so the best-looking one may simply be lucky. Treat anything under 100 games as early evidence, not proof.

Model comparison

FADE calls · favorite lost vs. lines expected

Model2026 games20262025 games2025

Current TAPE What the site and app use today2748% vs 31%1822% vs 32%

TAPE + travel +10 for 2+ time zones or 1,200+ miles3247% vs 31%2317% vs 30%

Betting lines alone Every favorite, as a baseline69320% vs 23%87723% vs 24%

The first number is how often the favorite lost; the second is what the betting lines expected for those same games. A model only helps if the first beats the second in both seasons. "TAPE + travel" is a test version and doesn't change any live call.

Last season check

2025 · games the model was never tuned on

GroupGamesFavorite lostLines expectedDifference

All favorites87723%24%-1 pts

FADE1822%32%-10 pts

Same TAPE rules, run on a full season they weren't built from. If FADE favorites lose more often than the lines expected here too, the edge is less likely to be luck.

New signals under test

Not part of the TAPE score yet

Signal20262025GamesVerdict

Wind 15+ mph+39 pts-12 pts3 / 14Too few games

Rain at kickoff-5 pts-10 pts8 / 7Too few games

Freezing kickoff–-5 pts0 / 7Too few games

Long trip (1,200+ mi)+8 pts+7 pts41 / 49Passes

2+ time zones+17 pts+9 pts41 / 42Passes

Rivalry game+20 pts-8 pts16 / 37Too few games

Each number is how much more often favorites lost in those games than the betting lines expected. A signal only joins the TAPE score if it beats the lines in both seasons on at least 30 games each. Weather is the forecast at kickoff; travel is measured from the favorite's home stadium.

By tier

TierGamesFavorite lostUnderdog covered

FADE2748%65%

WATCH15130%54%

HOLD51516%44%

How it's graded: past games are re-scored with the signals we can rebuild for any week (the AP poll at the time, last and next opponents, rest, travel, and the opening and closing lines). Weather, Kalshi and ESPN FPI are live-only, so they aren't part of past grades. A pick is "covered" if the underdog beat the closing spread; pushes are left out.

Week 110-0No fades

Week 102-360% covered

Week 90-425% covered

Week 83-0100% covered

Week 74-0100% covered

Week 62-167% covered

Week 52-267% covered

Week 40-250% covered

Week 30-1100% covered

Week 20-10% covered

Week 10-0No fades