BacktestBench - Strategy Selection: leaderboard

The Strategy Selection task family of BacktestBench (2026): choosing which candidate strategy achieves the best performance under the same market conditions, from a strategy written in natural language over a database of real market records. Each model drives the same retrieval, SQL and Python-execution pipeline; the score is accuracy in percent against the executed reference backtest; higher is better.

Metric: Back-test accuracy (%) on the Strategy Selection task family (choosing which candidate strategy achieves the best performance under the same market conditions), synthetic BacktestBench test split (the expert-crafted subset excluded), each model driving the AutoBacktest pipeline of the authors (Summarizer, SQL Retriever and Python Coder agents) over a PostgreSQL market database, temperature 0.6; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 23 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro89.07#145
2Qwen 3 Max85.27#415
3Qwen 3 235B A22B 2507 (Thinking)85.04#466
4Seed 1.884.8#252
5DeepSeek V3.281.47#419
6GLM-4.780.76#380
7Qwen 3 Coder Plus77.91#720
8MiniMax-M2.176.72#575
9GPT-OSS-120B76.48#704
10Qwen 3 30B A3B 2507 (Thinking)75.06#764
11Qwen 3 32B72.21#972
12MiMo-V2-Flash71.26#684
13Qwen 3 Next 80B A3B (Thinking)68.17#594
14Kimi K2 (Thinking)67.46
15Qwen 3 14B64.61#1269

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=backtestbench-strategy-selection · How It Works · Data refreshed daily, snapshot 2026-11-21.