From the alpha workflow.
State variables, sampling decisions, memory features, and what the evidence is showing. Each note names the sample, names the label, and reports the result — pass or fail.
research
01Memory in microstructure: how discrete events become a signal
Raw order-book events are useless to a backtester until you build memory over them. A conceptual taxonomy of the four kinds of memory you can design — intrinsic clocks, asymmetric decay, multi-scale, and state-dependent.
research
02Locking the first three raw atoms
A 1,000,000-event evidence pass on four candidate raw microstructure atoms — what locked, what got dropped, and what survives into Stage 2.
research
03Four trade-aggression atoms — and three of them collapsed
A 1,000,000-event evidence pass on four new trade-aggression atoms. Three of them turned out to be the same axis as raw_trade_ofi. One survived as a WATCH.
blog
04Latency isn't what you think it is
Mean latency is the wrong number. A practical walk through tail latency, jitter, and the engineering that separates paper strategies from live P&L.
blog
05Why your backtest lies to you (and how to make it stop)
The eight ways a backtest quietly overstates P&L — with a checklist for catching each one before real money hits the exchange.
guide
06Market microstructure, explained like you'll build something with it
A first-principles tour of order books, queues, price formation, and the execution reality hiding underneath every candlestick.
Every note names the sample, names the label, and reports the result. Pass or fail. The lab’s harness, study code, and diagnostics live on GitHub.