Bot Tester
External futures bots and the reusable components extracted from them. Each record separates what its upstream claims from what TradeRep has actually reproduced, and no README figure is ever shown as a result. External code enters here first — it does not jump into production TradeRep.
External-bot validation is owned by the external-bot research lane, not by the reboot/organization lane that maintains this index. Stage values, claims and blockers here are seeded from that lane's own committed artifacts (see sources) and must stay consistent with them. Nothing in this file promotes a bot, authorises funded execution, or counts as reproduced evidence.
Full bots
Entire external systems under evaluationMechanical intraday NQ/MNQ strategy + validation methodology
- unverified · Author-reported performance for the published strategy. (upstream author / repository)
- unverified · Production code uses 5m setup/zone detection with 1m execution to address intrabar ambiguity. (upstream code reading (static audit))
- unverified · Code includes explicit commissions/slippage and daily/consecutive-loss stops. (upstream code reading (static audit))
- Nothing reproduced yet. No performance figure may be attributed to this item.
Isolated zero-shot time-series forecasting candidate (pretrained PyTorch model)
- unverified · Emits forecast, uncertainty, direction/confidence and regime from a pretrained LaT-PFN model. (upstream code reading (static audit))
- unverified · The wrapper's lagged context mode corrects an older misaligned legacy mode. (upstream code/comments)
- unverified · Author live-record evidence. (upstream author)
- Nothing reproduced yet. No performance figure may be attributed to this item.
- latpfn-baseline-blocker: An open implementation blocker prevents a reproducible baseline run; owned by the external-bot lane.
Execution/risk architecture + LLM decision layer (paper-trading system)
- unverified · Broker/local position reconciliation, hard safety gates, commission handling, DOM/order-flow features and paper-trading infrastructure. (upstream code reading (static audit))
- unverified · Upstream explicitly describes it as paper trading, not live-money validated. (upstream documentation)
- Nothing reproduced yet. No performance figure may be attributed to this item.
- mnq-ai-trader-no-recorded-sessions: No native recorded sessions exist, so deterministic historical replay feasibility is unresolved (DATA_BLOCKED).
- mnq-ai-trader-llm-cost-control: Historical audit must run with cached/offline LLM mode first to prevent uncontrolled API cost and nondeterministic re-evaluation.
External strategy lane — 5m IFVG retest inside an active 15m bullish FVG
- unverified · A 5m IFVG retest inside an active 15m bullish FVG in a morning window, with declared stop/target, is tradable. (upstream repository / economic hypothesis)
- Nothing reproduced yet. No performance figure may be attributed to this item.
- nq-strategy-b-incomplete-repo: The historical repository is incomplete, so repository reproduction (study A) is limited; the causal clean-room implementation (study B) is the path that can advance through gates.
Components
Reusable pieces extracted from external systems — the part that is usable even when a bot's alpha is notExecution integrity / replay determinism
- unverified · Canonical event ordering, duplicate-event rejection and monotonic replay assertions make a replay deterministic and non-anticipating. (internal design note)
- Nothing reproduced yet. No performance figure may be attributed to this item.
Second-engine replication for finalists
- unverified · Recording code/data/config hashes plus trade-level reconciliation turns engine disagreement into an investigation trigger rather than a silent reconciliation. (internal design note)
- Nothing reproduced yet. No performance figure may be attributed to this item.
- no-finalist-to-replicate: Replication applies to finalists; across A0-A5 and the Strategy Factory campaigns there is currently no surviving finalist.
Order-flow / volume-profile / regime features with causality enforcement
- unverified · CVD/aggressor imbalance, execution pressure, volume-profile levels and a microstructure regime gate are usable research features. (upstream concept source)
- Nothing reproduced yet. No performance figure may be attributed to this item.
- microstructure-data-blocked: TradeRep's canonical data is 1-minute OHLC; aggressor/DOM features require tick or trade data that does not exist in the current dataset. Tests depending on them are marked DATA_BLOCKED rather than approximated.
Null benchmarking / signal-information test
- unverified · Comparing a candidate against randomized entries under matched exposure, holding time, side balance, session window and risk sizing shows whether signal selection adds information beyond drift/exposure. (internal methodology plan)
- Nothing reproduced yet. No performance figure may be attributed to this item.
Evaluation ladder
Every candidate runs this, in order- 01 Static code/methodology audit
- 02 Reproduce author tests/backtests where data permit
- 03 Leakage/lookahead audit
- 04 Cost/slippage stress
- 05 Random/null benchmark
- 06 Walk-forward/OOS evaluation
- 07 Cross-market evaluation where supported
- 08 Paper/shadow audit with orders disabled
- 09 Compare shadow predictions/fills with TradeRep
- 10 Funded/live execution remains a separate explicit decision and is NOT enabled
Read from research-output/registry/bot-testers.json · docs/status/TRADE_REP_MASTER_TRACKER.json