Autoresearch Trade Bot

Continuously mutate train.py with an LLM against a frozen crypto campaign and keep only score improvements.

Phase: LLM autoresearch worker active Research Champion Paper Ready: NO Research Champion Fast Holdout: NO Research Champion Rollout Certified: NO Rollout Champion: none Latest Cycle Rollout Ready: NO Research Rollout: NOT READY Primary Status: STALE (3d 15h old) Loop State: searching

Research Champion

Karpathy-style single-file strategy mutation on train.py using frozen historical campaigns.

  • strategy_name: configurable-momentum
  • research_score: -32.92768654482868

Promotion Gate

min_total_return0.0
min_sharpe1.0
max_drawdown0.2
min_acceptance_rate0.6

Baseline Metrics

total_return-0.1191
sharpe-31.8288
max_drawdown0.1191
average_turnover0.0272
bars_processed8639
score-32.9277

Research Blockers

  • Latest LLM candidate underperformed the baseline on the screen window.
  • Latest kept candidate still fails gate: total_return_below_gate.
  • Latest kept candidate still fails gate: sharpe_below_gate.
  • Latest kept candidate still fails gate: acceptance_rate_below_gate.
  • Research champion still fails paper gate: total_return_below_gate.
  • Research champion still fails paper gate: sharpe_below_gate.
  • No rollout champion has cleared holdout validation yet.

Next Milestones

  • Keep finding an LLM-generated candidate that survives full campaign evaluation.
  • Accumulate multiple kept improvements in the research branch.
  • Promote only candidates that satisfy the hard rollout gates.

Multi-Window

  • campaign_path: /var/data/autoresearch/llm/campaigns/render-binance-usdm-futures-5m-20260331t090000z.json
  • campaign_refreshed_this_cycle: False
  • latest_decision: {'decision': 'discard_screen', 'stage': 'screen', 'mutation_label': 'llm-c497e127', 'baseline_score': -32.92768654482868, 'candidate_score': -47.613804964445826, 'candidate_acceptance_rate': 0.0, 'candidate_average_metrics': {'total_return': -0.16307450495573128, 'sharpe': -46.1086629537579, 'max_drawdown': 0.16307450495573128, 'average_turnover': 0.03747146608634381, 'bars_processed': 8639, 'nonzero_turnover_steps': 585}, 'kept_commit': None, 'failure_reason': '', 'provider_name': 'llm', 'model_name': 'gpt-5-mini-2025-08-07', 'gate_failures': ['total_return_below_gate', 'sharpe_below_gate', 'acceptance_rate_below_gate']}
  • baseline_strategy_name: configurable-momentum
  • baseline_score: -32.92768654482868
  • model_name: gpt-5-mini-2025-08-07
  • provider_name: llm
  • current_best_ready_for_paper: False
  • current_best_fast_validation_pass_rate: 0.0
  • current_best_fast_holdout_passed: False
  • current_best_validation_pass_rate: 0.0
  • current_best_validated_for_rollout: False
  • latest_cycle_rollout_ready: False
  • research_champion_strategy_name: configurable-momentum
  • rollout_champion_strategy_name:

Worker Status

  • latest_cycle_completed_at: 2026-04-01T04:08:40.000012+00:00
  • last_processed_bar: 2026-04-01T04:05:00+00:00
  • mutation_win_rate: 0.0
  • current_best_fast_validation_pass_rate: 0.0
  • current_best_fast_holdout_passed: False
  • current_best_validation_pass_rate: 0.0
  • generation_validity_rate: 1.0
  • current_best_ready_for_paper: False
  • current_best_validated_for_rollout: False
  • latest_cycle_rollout_ready: False
  • consecutive_failures: 0
  • latest_decision: discard_screen
  • latest_candidate_score: -47.613804964445826
  • baseline_score: -32.92768654482868

Leaderboard

  • llm-d2178e25 (score -21.5169, accepted no)
  • llm-d2178e25 (score -21.5169, accepted no)
  • llm-ae48eda4 (score -34.0659, accepted no)
  • llm-f4ee9bd5 (score -34.1582, accepted no)
  • llm-b03adfe5 (score -34.5156, accepted no)

Latest Candidate

  • strategy_name: configurable-momentum
  • research_score: -47.613804964445826
  • ready_for_paper: False
  • train_config: {'gross_target': 0.5, 'lookback_bars': 24, 'min_signal_strength': 0.025, 'ranking_mode': 'risk_adjusted', 'regime_lookback_bars': 36, 'regime_threshold': 0.015, 'top_k': 2, 'use_regime_filter': True, 'min_cross_sectional_spread': 0.065, 'volatility_floor': 0.007, 'reversal_bias_weight': 0.2, 'funding_penalty_weight': 0.1}
  • gate_failures: ['total_return_below_gate', 'sharpe_below_gate', 'acceptance_rate_below_gate']
  • provider_name: llm
  • model_name: gpt-5-mini-2025-08-07

Research Champion Fast Holdout

  • No fast holdout summary yet.

Research Champion Rollout Certification

  • No rollout certification summary yet.

Rollout Champion

  • strategy_name: n/a
  • research_score: n/a
  • fast_validation_pass_rate: n/a
  • rollout_validation_pass_rate: n/a
  • validated_for_rollout: False

Rollout Shortlist

  • No rollout shortlist candidates yet.

Latest Kept Change

  • strategy_name: n/a
  • baseline_strategy_name: n/a
  • score_delta: n/a
  • No kept config diff published yet.

Endpoints

  • / dashboard
  • /health health probe
  • /api/status JSON snapshot

Strategy Families

These tabs track the parallel wave-1 family searches on the same frozen campaign.

PhaseLLM autoresearch worker active
Loop Statesearching
Research Championibs-reversion-short-horizon
Rollout Championn/a
Baseline Score-167.2813
Research Champion Paper ReadyFalse
Research Champion Fast HoldoutFalse
Research Champion Rollout CertifiedFalse
Latest Decisiondiscard_full
Cycle Completed2026-04-01T04:05:32.979211+00:00

IBS Reversion Leaderboard

  • llm-967e6c3b (score 0.0, accepted no)
  • llm-967e6c3b (score 0.0, accepted no)
  • llm-5c29eb12 (score -63.0898, accepted no)
  • llm-5c29eb12 (score -63.0898, accepted no)
  • llm-9227ddbf (score -92.4751, accepted no)

IBS Reversion Blockers

  • Latest LLM candidate passed screen but underperformed on the full campaign.
  • Latest LLM candidate produced no trades on the evaluation window.
  • Latest kept candidate still fails gate: no_trades_executed.
  • Latest kept candidate still fails gate: sharpe_below_gate.
PhaseLLM autoresearch worker active
Loop Statesearching
Research Championdual-momentum-timeseries
Rollout Championn/a
Baseline Score-45.5099
Research Champion Paper ReadyFalse
Research Champion Fast HoldoutFalse
Research Champion Rollout CertifiedFalse
Latest Decisiondiscard_full
Cycle Completed2026-04-01T04:07:45.290309+00:00

Dual Momentum Leaderboard

  • llm-c04a3bab (score -36.5712, accepted no)
  • llm-c04a3bab (score -36.5712, accepted no)
  • llm-62267d5f (score -37.6854, accepted no)
  • llm-62267d5f (score -37.6854, accepted no)
  • llm-7d25c3dc (score -37.9024, accepted no)

Dual Momentum Blockers

  • Latest LLM candidate passed screen but underperformed on the full campaign.
  • Latest kept candidate still fails gate: total_return_below_gate.
  • Latest kept candidate still fails gate: sharpe_below_gate.
  • Latest kept candidate still fails gate: acceptance_rate_below_gate.
PhaseLLM autoresearch worker active
Loop Statesearching
Research Championdonchian-atr-breakout-trend
Rollout Championn/a
Baseline Score0.0
Research Champion Paper ReadyFalse
Research Champion Fast HoldoutFalse
Research Champion Rollout CertifiedFalse
Latest Decisiondiscard_screen
Cycle Completed2026-04-01T04:04:43.238945+00:00

Donchian ATR Breakout Leaderboard

  • llm-1dbbd54e (score 0.0, accepted no)
  • llm-43f0d2f6 (score 0.0, accepted no)
  • llm-29f2f136 (score 0.0, accepted no)
  • llm-6c6d3e85 (score 0.0, accepted no)
  • llm-29992c55 (score 0.0, accepted no)

Donchian ATR Breakout Blockers

  • Latest LLM candidate underperformed the baseline on the screen window.
  • Latest kept candidate still fails gate: total_return_below_gate.
  • Latest kept candidate still fails gate: sharpe_below_gate.
  • Latest kept candidate still fails gate: acceptance_rate_below_gate.

This dashboard is intentionally read-only in v1. It exposes the current research kernel state, not a live trading control plane.