Independent, reproducible studies of trading ideas. Each paper sets out its hypothesis, data, sample period, variables, costs and statistical tests before reporting results, including results that don't support the idea.
These papers are educational research written while I learn the field. They are not financial advice or recommendations to trade. All results are historical simulations, and past performance does not predict future returns.
Is day trading by chart reading worth the hours? The 10 most popular readings (ranked by a year of Wikipedia page views), from moving averages and RSI to candlestick patterns, tested long and short on 1-, 5- and 15-minute candles against coin-flip traders. All 30 lose money after costs; hit rates 46.5–52%.
A pre-registered test of plain Supertrend (10, 3) and seven one-change variants (execution delay, skipping the open, stops and targets, VWAP and moving-average filters) on 5-minute bars of SPY, QQQ, IWM and DIA. No variant beats the plain rule; stops made results worse on every ETF.
A pre-registered test of buying SPY, QQQ, IWM and DIA after a dip in the first hour of trading, on two years of 1-minute data, against 10,000 random traders. No significant edge after costs; the idea is dropped for all four ETFs.
No papers published yet. The first is in preparation below, and it will appear here, free to read in the browser or as a PDF.
No papers match that search.
Market mechanics and instruments, returns and risk, probability, statistics, hypothesis testing and research method.
In progressA reproducible Python framework: historical data, fixed rules, transaction-cost assumptions and statistical evaluation.
BuiltControlled tests of momentum, mean reversion, volatility, technical indicators and market microstructure.
In progressPapers written up and published on this page, with data, method, results and instructions for reproducing them.
3 papers published