Scope retained in the historical record.
Sujay Nadkarni / The Nad Coder
@thenadcoder · Active
1 World Cup semifinal set · 2 IPL finish claims
Outcome market stays separate from advance and exact-score claims.
Advance is not substituted for strict 90-minute outcome.
Exact scores remain phase-aware where a phase was stated.
Every tested call in one table.
Source, market, prediction, outcome and verdict stay together. Open any source to inspect the original public post.
| Source | Event | Market | Prediction | Outcome | Verdict |
|---|---|---|---|---|---|
| DaLUaHEujtoFour ML models predict the World Cup knockout path | Hit2026 FIFA World Cup semifinal field | Exact semifinalist set | France, Spain, England and Argentina | France, Spain, England and Argentina | 4/4 slots; a later quarterfinal favorite-identity comparator also had 4/4. |
| DaLUaHEujtoFour ML models predict the World Cup knockout path | PendingEngland vs Argentina | Team to advance | Argentina reaches the final | Pending | Grade A frozen before semifinal; follow-up Reel repeats four-model agreement. |
| DaLUaHEujtoFour ML models predict the World Cup knockout path | Pending2026 FIFA World Cup | Outright champion | France wins the World Cup | Pending | Explicit creator tie-break captured before settlement. |
| DaLUaHEujtoFour ML models predict the World Cup knockout path | Pending2026 FIFA World Cup | Negative outright | Argentina does not win the World Cup | Pending | The reported four-model final path has Argentina losing. |
| DXQhXz6jvVSClaude Code builds an IPL model | HitIPL 2026 | Tournament winner | Royal Challengers Bengaluru | Royal Challengers Bengaluru | Hit; RCB was also the defending-champion baseline. |
| DXQhXz6jvVSClaude Code builds an IPL model | HitIPL 2026 | Tournament runner-up | Gujarat Titans | Gujarat Titans | Hit; no frozen same-time runner-up baseline was available. |
What was tested, source by source.
Public source media stays outside Moshiverse; each card links back to the original platform and preserves only the audit facts needed to reproduce the classification.
Four ML models predict the World Cup knockout path
The Reel shows broad feature inputs and a four-model output, but not model names, code, calibration or a full prediction log.
- Market
- Exact semifinalist set
- Prediction
- France, Spain, England and Argentina
- Outcome
- France, Spain, England and Argentina
Verdict: 4/4 slots; a later quarterfinal favorite-identity comparator also had 4/4.
- Market
- Team to advance
- Prediction
- Argentina reaches the final
- Outcome
- Pending
Verdict: Grade A frozen before semifinal; follow-up Reel repeats four-model agreement.
- Market
- Outright champion
- Prediction
- France wins the World Cup
- Outcome
- Pending
Verdict: Explicit creator tie-break captured before settlement.
- Market
- Negative outright
- Prediction
- Argentina does not win the World Cup
- Outcome
- Pending
Verdict: The reported four-model final path has Argentina losing.
Claude Code builds an IPL model
Public Kaggle data, a creator-shown Claude Code prompt and a local dashboard. The code and model artifacts were not published.
- Market
- Tournament winner
- Prediction
- Royal Challengers Bengaluru
- Outcome
- Royal Challengers Bengaluru
Verdict: Hit; RCB was also the defending-champion baseline.
- Market
- Tournament runner-up
- Prediction
- Gujarat Titans
- Outcome
- Gujarat Titans
Verdict: Hit; no frozen same-time runner-up baseline was available.
What this creator’s format can support
Public sports-model claims are scored, but the underlying models are not recreated because code, model names and frozen predictions were not published.
What is held back on purpose
- World Cup Top 4 claim was posted just 15 minutes 47 seconds before the first covered knockout match.
- The 37/70 IPL validation is creator-displayed visual evidence, not independently reproduced code.
2 source records, no silent deletions.
Detailed video cards appear above when the claim is ready for reader-facing scoring. Corpus rows stay explicit about what has not yet been extracted.
Existing audited corpus
Aggregate scorecard importedThe current per-video ledger grows without retroactively inventing row-level evidence that was not previously preserved.