ChallengeFlag - Model Accuracy

Model Accuracy

All predictions below were made and published before the draft. The 2026 results are based on draft position as a proxy for NFL evaluation. Career outcome validation will be available in 2029-2030 when the class has matured.

These are ML predictions, not professional scouting opinions. The model does not watch film. It finds statistical patterns in combine measurements and college production that have historically correlated with NFL career outcomes.

2026 Draft -- Round 1 Results

4/7

Top position picks in Round 1

High confidence Stars in Round 1

0.41

Non-skill macro F1 (test set)

0.55

Dud precision (high confidence)

Top Position Pick Results

Position	Player	College	p(Star)	Confidence	Actual Pick	Result
QB	Fernando Mendoza	Indiana	54.5%	Low	#1 Las Vegas Raiders	HIT
WR	Carnell Tate	Ohio State	38.5%	Medium	#4 Tennessee Titans	HIT
LB	Arvell Reese	Ohio State	49.6%	High	#5 New York Giants	HIT
OT	Spencer Fano	Utah	49.6%	High	#9 Cleveland Browns	HIT
DE	Malachi Lawrence	UCF	51.0%	High	#23 Dallas Cowboys	LATE R1
CB	Treydan Stukes	Arizona	48.9%	High	#38 Las Vegas Raiders	R2
S	Lorenzo Styles Jr.	Ohio State	49.8%	High	#172 New Orleans Saints	MISS

Hits

Spencer Fano, OT -- #9 Browns

Model's #1 OT at 49.6% high confidence. Top 10 pick, exactly where a franchise OT belongs.

Arvell Reese, LB -- #5 Giants

Model's #1 LB at 49.6% high confidence. Top 5 pick.

Carnell Tate, WR -- #4 Titans

Model's #1 WR at 38.5%. Most analysts had him lower. Went top 5.

David Bailey, EDGE -- #2 Jets

High confidence Star prediction. Went #2 overall.

Misses

Lorenzo Styles Jr., S -- #172 Saints

Model's #1 safety at 49.8% high confidence. Went Round 6. Biggest miss of the 2026 class. Scouts saw something in film the combine data missed.

Caleb Downs, S -- #11 Cowboys

Model ranked Lorenzo Styles above Downs. Downs went top 12. The safety model needs work.

2022-2024 Backtest

The model was backtested on the 2022-2024 draft classes before the 2026 draft. Key findings:

81-90%

Non-skill Dud accuracy

Star recall (known weakness)

0.55

Dud precision

0.68

Dud recall

Star recall of 0% on the 2022-2024 backtest was an acknowledged limitation before the 2026 draft. The model is significantly better at identifying busts than identifying stars. This is the primary focus of v2 development.

Notable 2022-2024 Dud calls

Player	Position	Draft Round	p(Dud)	Status
Drake Jackson	EDGE	Round 2	High	2 sacks in 3 seasons
Felix Anudike-Uzomah	EDGE	Round 1	High	Limited production
Cameron Thomas	EDGE	Round 3	High	Limited snaps

Model trained on 2000-2021 career outcomes. Test set: 2019-2021. Players drafted post-2021 excluded from training to prevent data leakage. Full career outcome validation for the 2022-2026 classes available 2029-2030.