MARKET TRACE RESEARCH / EST. 2026
Trace the signal.
Test the edge.
Track market signals before you believe them. Research focuses on real-time market data, point-in-time contracts, prospective OOS and falsifiable hypotheses.
01ObserveWhat was actually available?
02FalsifyWhat could make the signal fake?
03SeparateFeature, outcome, and execution
04ReviewManual before promotion
01 / PRINCIPLE
Verification before prediction.
Instead of hunting for attractive results, we first attack the ways a result could be false: data leakage, selection bias, clock errors, and spurious lead-lag. Effect metrics remain blinded until the sample reaches the preregistered maturity gate.
02 / ACTIVE WORK
All research →Experiments
H2 V3R2EFFECT BLINDED
Korea → Binance Lead-Lag
We prospectively test whether Korean exchange data adds incremental lead-lag information at the moment a Binance candidate is generated. Current status: 207 / 300 fully matched, non-quarantined decisions.
HISTORYFROZEN ARMS
Invalid results stay visible.
V1, the V2 family, and earlier Rich-L2 arms are not pooled into the active confirmatory arm. Failed studies remain preserved with their provenance.
03 / RESEARCH NOTES
All notes →How research fails.
AI / FALSIFICATION · 6 MIN
Why an AI-discovered strategy can still be rejected immediately.
Why point-in-time validity and reproducibility matter more than an attractive backtest.
Read note → POINT-IN-TIME · 6 MINIf future data is mixed, the backtest is invalid.
Even within the same second, information following the decision may enter past features.
Read note → EXECUTION COST · 5 MINHow fees and slippage reduce returns.
How repeated small costs can create a large gap between gross and net performance.
Read note → RISK / PAYOFF · 5 MINEven with an 80% win rate, you can still lose money.
Why average win, average loss, and expectancy matter more than win rate alone.
Read note → STRATEGY AUDIT · 6 MINWhy a popular trading strategy should not be trusted at face value.
Popular strategies should face the same OOS, cost, and execution tests as any other hypothesis.
Read note → PROVENANCE · 5 MINWhy not delete failed experiments?
Failed hypotheses and failure modes become explicit constraints for the next experiment.
Read note →04 / POSITIONResearch,
Research,
not signals.
MKTRC does not sell trading signals. It is a public research record covering hypotheses, data contracts, failure cases, out-of-sample validation, and research methods.
05 / WATCH
YouTube Shorts →Research in 30 seconds.
AI / FALSIFICATION
Why did we reject an AI-discovered strategy?
Good numbers do not matter if point-in-time validity is broken.
Watch →LOOK-AHEAD BIASWhy a backtest can look brilliant for the wrong reason.
Even a small amount of future information can invalidate a backtest.
Watch →WIN RATEHow can an 80% win rate still lose money?
Average win, average loss, and expectancy come before win rate.
Watch →