Current work / Market microstructure
What happens between a signal and a fill?
My current work moves beyond faster bar indicators to event-level market mechanics: order flow, liquidity, queue uncertainty, and price behavior around hypothetical execution.
The replay and evaluation tools exist. A robust profitable execution strategy has not been established. Public-data fill proxies are not actual executions.
The research question
Does the opportunity survive execution?
I study whether observable order-flow and liquidity states contain enough information to improve short-horizon execution decisions after accounting for fill uncertainty, adverse selection, and costs.
The 5-minute and 1-minute studies remain useful context and benchmarks. Event-level data answer a different question rather than automatically upgrading those earlier results.
The engineering contribution
A replayable market-data record.
The C++ core captures and reconstructs public trades and order-book updates with explicit sequence and timestamp discipline. Deterministic replay makes it possible to compare experiments against the same recorded market events.
Python research layers examine fill proxies, post-fill markouts, and cost-aware decision rules. Local compute benchmarking has been performed, but it is not a measurement of exchange or end-to-end execution latency.
System map
Observation first.
Claims second.
These components support research. The diagram is not a live order-routing pipeline, and the website has no trading controls.
Raw trades and order-book updates.
Sequence checks, book reconstruction, and deterministic output.
Fill uncertainty, markouts, adverse selection, and costs.
Negative tests are part of the deliverable.
Recent development screens have not established robust positive execution economics. Improvements relative to an unfavorable baseline do not become profitable strategies merely because the relative comparison is positive.
I retain failed or inconclusive tests instead of presenting engineering throughput, model accuracy, or hypothetical fills as investment returns. Frozen studies remain frozen; this portfolio does not open protected research samples.
What is public.
This page describes the research question, engineering responsibilities, and evidence status. Proprietary coefficients, decision thresholds, raw event records, and execution details are not published here.
Editorial status reviewed on September 26, 2026. This is a dated research summary, not a live system-health feed.