Observe
Capture public market events, liquidity, funding, basis, book quality, and runtime context.
Research system
Research designed to fail honestly.
TheDuckTrader does not optimize for the number of strategies that trade. It optimizes for the quality of evidence that survives data, economics, tail risk, and time.
Latest reviewed snapshot
Interesting evidence, no investable strategy yet.
Reviewed 24 July 2026 across approximately 278 hours of server market data.
System governance42 / 21 / 0
42 mapped families, 21 active observers or inputs, and zero paper or live candidates.
Carry economicsBenchmark only
Average net edge remains near -40 bps after conservative round-trip costs.
Correlated relative valueBest research lead
About 59% hit rate and +4.47 bps average response, but costs and roughly -226 bps tail excursion still block promotion.
Funding data qualityMixed
BTC and ETH are research-usable; SOL remains unreliable for promotion evidence; ADA, LINK, and XRP are caution-only.
The research loop
Observe. Reject. Learn. Validate.
A repeatable process keeps attractive historical results from becoming unearned confidence.
Reject
Remove stale, contaminated, uneconomic, clustered, or non-evaluable observations before scoring.
Learn
Attribute false positives and adverse excursions to market state, data quality, and execution conditions.
Validate
Use chronological splits, embargo, independent assets and venues, and frozen promotion criteria.
Portfolio status
Every family has an explicit role.
Research infrastructure, conditioning features, offline hypotheses, and paused ideas are never presented as equivalent evidence.
Funding & basis carryBenchmark
Economically valid concept; observed net edge remains below realistic execution costs.
Premium / funding flipConditioning
Useful regime information, especially in compression states; standalone returns remain too small.
Correlated relative valueOffline review
Pair-specific pockets exist, but adverse excursion, freshness, and cross-window persistence remain weak.
Order-flow imbalanceSecondary feature
Direction and basis-response variants do not currently improve outcomes enough to stand alone.
Crowding and funding timingRegime input
Maintained as context for liquidity and stress, not as independent alpha.
Microstructure risk scoreActive experiment
Tests whether short-horizon volatility and liquidity stress can reduce downstream false positives and tails.
Current experiment
Predict activity and risk, not price direction.
Recent market evidence supports a narrower hypothesis: short-horizon volatility or liquidity-state prediction may be useful as a risk overlay even when directional prediction is weak.
Depth-20 event data
One-second BTC, ETH, and SOL public order-book and aggregate-trade features, with versioned schemas, cumulative depth, and fixed-notional slippage.
Non-directional labels
Future adverse movement, spread widening, depth collapse, and combined liquidity stress at 10s, 60s, and 15m horizons.
Downstream proof
A model is useful only if it reduces false positives and adverse excursion while retaining enough strategy observations.
Evidence gates
What must be true before promotion.
Passing a model metric is not sufficient. The complete decision must remain economically and operationally coherent.
Data integrity
Fresh books, explicit receive-time versus exchange-time semantics, viable depth, stable schema, source reconciliation, and no testnet/mainnet contamination.
Statistical validity
Independent outcomes, chronological splits, embargo, informative prevalence, calibration, and multi-window stability.
Economic validity
Positive forward response after conservative costs, sufficient sample retention, controlled false positives, and acceptable tails.
Operational validity
Explicit permissions, no silent router promotion, stable outputs, reproducible tests, observability, and human approval.
Current decision: promote nothing.
The platform is functioning correctly by rejecting evidence that is interesting but not yet investable or product-ready.
Research partnerships
We want independent challenges, not confirmation.
We are interested in conversations with institutional operators, market-microstructure researchers, fintech product teams, and quantitative risk specialists.
Independent venue validation
Institutional workflow discovery
Risk and market-quality pilots