# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260403-103635
**Date:** Apr 03, 2026
**Category:** analysis
**Question ID:** ANALYSIS-024

---

## Question

A quantitative trading firm backtests a strategy: 15% annual return, Sharpe ratio 2.1, max drawdown 8%. They want to go live. (1) What could go wrong between backtest and live trading? List at least 5 risks. (2) The backtest used 5 years of data and tested 200 parameter combinations. Calculate the probability this outperformance is due to overfitting. (3) Design a live testing protocol that minimizes capital at risk while validating the strategy.

---

## Winner

**GPT-5.4** (openrouter)
- Winner Score: 9.29
- Matrix Average: 7.10
- Total Judgments: 74

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | GPT-5.4 | openrouter | 9.29 | 9 |
| 2 | Grok 4.20 | openrouter | 8.52 | 9 |
| 3 | Claude Opus 4.6 | openrouter | 7.67 | 9 |
| 4 | MiMo-V2-Flash | Xiaomi | 7.55 | 8 |
| 5 | GPT-OSS-120B | OpenAI | 7.51 | 8 |
| 6 | DeepSeek V4 | openrouter | 7.49 | 9 |
| 7 | Claude Sonnet 4.6 | openrouter | 7.09 | 8 |
| 8 | Gemini 3 Flash Preview | Google | 6.14 | 5 |
| 9 | Gemini 3.1 Pro | openrouter | 2.65 | 9 |

---

## 10×10 Judgment Matrix

Rows = Judge, Columns = Respondent. Self-judgments excluded (—).

| Judge ↓ / Resp → | GPT-OSS-120B | Gemini 3.1 Pro | Claude Opus | GPT-5.4 | DeepSeek V4 | MiMo-V2-Flash | Claude Sonnet | Grok 4.20 | Gemini 3 | MiniMax M2.5 |
|---|---|---|---|---|---|---|---|---|---|---|
| GPT-OSS-120B | — | 1.9 | 5.6 | 8.8 | 7.8 | 6.8 | 6.2 | 8.1 | · | · |
| Gemini 3.1 Pro | · | — | 7.3 | 9.8 | 7.0 | 7.7 | 6.0 | 7.3 | · | · |
| Claude Opus | 7.2 | 0.8 | — | 9.3 | 7.0 | 6.8 | 7.0 | 8.8 | 0.5 | · |
| GPT-5.4 | 4.4 | 1.2 | 5.0 | — | 6.0 | 5.1 | 3.8 | 7.0 | · | · |
| DeepSeek V4 | 9.2 | 5.5 | 9.2 | 9.2 | — | 9.0 | 9.2 | 9.2 | 8.0 | · |
| MiMo-V2-Flash | 8.3 | 5.3 | 8.4 | 9.0 | 8.8 | — | · | 9.0 | 8.2 | · |
| Claude Sonnet | 8.0 | 1.4 | 8.0 | 9.3 | 7.3 | 8.2 | — | 8.8 | · | · |
| Grok 4.20 | 8.4 | 3.3 | 7.5 | 9.0 | 6.8 | · | 8.6 | — | 6.0 | · |
| Gemini 3 | 9.0 | 2.0 | 9.2 | 9.8 | 9.0 | 9.0 | 9.0 | 9.6 | — | · |
| MiniMax M2.5 | 5.5 | 2.5 | 8.8 | 9.3 | 7.8 | 7.8 | 6.8 | 9.0 | 8.0 | — |

---

## Methodology

- **10×10 Blind Peer Matrix:** All models answer the same question, then all models judge all responses.
- **5 Criteria:** Correctness, completeness, clarity, depth, usefulness (each scored 1–10).
- **Self-judgments excluded:** Models do not judge their own responses.
- **Weighted Score:** Composite of all 5 criteria.

---

## Citation

The Multivac (2026). Blind Peer Evaluation: ANALYSIS-024. app.themultivac.com

## License

Open data. Free to use, share, and build upon. Please cite The Multivac when using this data.

Download raw JSON: https://app.themultivac.com/api/evaluations/EVAL-20260403-103635/results
Full dataset: https://app.themultivac.com/dashboard/export