# Spain 0-0 Argentina - Final Project Turtle Audit

Audit timestamp: `2026-07-19T21:32:53Z`

## Settlement

- Scope: regulation time
- Score orientation: Spain first, Argentina second
- Locked C4: `1-1`, `0-1`, `1-0`, `1-2`
- Actual regulation score: `0-0`
- C4 result: **MISS**
- Result source: Associated Press, https://apnews.com/article/fccc26aa12d9226e63d06b601b770617

## Immutable Lock

The public card was frozen at `2026-07-19T18:24:41Z`. This audit does not add,
remove, reorder, or relabel any locked score. The original C4 remains the public
prediction that is being judged.

## What The Model Knew

The model did not overlook `0-0`. It ranked the score sixth with probability
`0.071671582290` (7.17%). The fourth and final selected node, `1-2`, had
probability `0.078378670653` (7.84%). The gap was only `0.006707088363`, or
0.67 percentage points.

The contradiction ledger also recorded that the scoreless-draw branch passed
its source floor: six qualifying scoreless draws in 100 admitted rows produced
a posterior rate of 6.86%, above the declared 0.8% threshold. Despite that,
the selector assigned the branch no C4 node.

## Exact Failure

This was a **scoreless-draw branch omission**. The selector allocated two BTTS
nodes and two one-sided clean-sheet nodes while leaving the surviving `0-0`
branch uncovered. A mild BTTS preference placed `1-1` above `0-0`, and the
four-node boundary then excluded the actual score.

This is not an absent-data defense and not ordinary postgame surprise. Turtle
identified `0-0` as a material failure route before kickoff and still chose not
to publish it. That selection decision failed.

## Why The Checks Did Not Save The Card

The 42 stability forks measured whether the selected nodes persisted under the
model's own declared perturbations. They did not prove that the model was
calibrated to real football outcomes. The card covered 37.73% of its modeled
exact-score mass, leaving 62.27% outside C4. Internal reproducibility was
mistaken for practical predictive reliability.

## Final Laboratory Verdict

The current Project Turtle exact-score protocol failed its final test. The
laboratory experiment is closed pending a genuinely rebuilt and independently
backtested method. The evidence, locked card, and miss remain public so the
record cannot be rewritten after the result.