Quantitative trading research, tested in the open.
We write down our hypotheses before looking at the data, test them against random controls on survivorship-free data, and publish what we find, including the failures.
How we test · Research log · Learn
Latest
Backtest overfitting and why the best result is usually luckWhy trying many strategy variants almost guarantees an impressive but meaningless backtest, and the habits that guard against it.
Launch note: what is here and what is still to comeThe first research log entry: what SDS Quant Research is testing, what has been built, and what will be published when.
Look-ahead bias: when a backtest cheats without meaning toWhat look-ahead bias is, the common ways it sneaks into a backtest, and how to catch it with automated tests.
What it means to pre-register a backtestBorrowing an idea from medical and social science research: write the hypothesis, rules and success criteria down before you look at the results.
Survivorship bias in backtestsWhy testing a strategy only on companies that exist today makes it look better than it was, and how to avoid the mistake.