All sports · one honest page

Model accuracy

Walk-forward backtests: every prediction was made before the result was known. The dashed marker on each bar is 52.4% — break-even against standard -110 vig.

NFL

1980–2024 · 45 seasons · 11,352 games

64%

Straight-up

13.66vs 13.30

Margin err · us vs line

Against the spread52%

3,738-3,436-201

45%break-even 52.4%60%

Early edge faded — about 50.8% ATS in the 2020s.

Full NFL track record

College football

2014–2025 · 12 seasons · 8,947 games

70%

Straight-up

17.80vs 15.64

Margin err · us vs line

Against the spread49%

4,104-4,220-154 · Totals 51%

45%break-even 52.4%60%

Never beat the closing line on margin error.

Full College football track record

NBA

2008–2023 · 3 eras · 18,552 games

67%

Straight-up

12.50vs 12.18

Margin err · us vs line

Against the spread50%

5,466-5,370-179 · Totals 48%

45%break-even 52.4%60%

Edge decayed by era: 51.0% → 50.3% → 50.1% ATS.

Full NBA track record

NCAAB

2013–2026 · 3 eras · 47,794 games

71%

Straight-up

12.17vs 11.04

Margin err · us vs line

Against the spread49%

16,744-17,090-473 · Totals 51%

45%break-even 52.4%60%

No era beat the vig: 48.9% → 49.8% → 50.2% ATS.

Full NCAAB track record

The honest verdict

A competent forecaster that roughly matches the Vegas line on prediction error — but with no demonstrated betting edge. None of the four models clears the 52.4% needed to beat the spread long-term, and the edges that existed in older eras have decayed as markets got sharper. Good enough to follow for fun; not a way to make money.