The Strategy Pipeline operationalizes internally developed beta/alpha investment strategies into a reproducible backtest-to-live workflow. Every strategy runs over the same shared investment universe and the same data and backtest layers, so results are comparable and reproducible — and the strongest strategy ensemble achieved a paper-trading Sharpe above 1.5. Production deployment runs through a Dockerized workflow.

​
Two kinds of strategies, two kinds of verdicts

Same data layer, same backtest engine — different question: Beta — risk premia, factor tilts, asset allocation, trend. These usually do make money; the question is whether it is money you could not have had by holding something you already own. The verdict comes from a spanning regression: after controlling for known factors, is any alpha left? A trend strategy that is quietly long equities is an expensive SPY. Alpha — cross-sectional selection, stat arb. These live or die on whether the ranking predicts, then on whether it survives the turnover needed to harvest it. The verdict comes from the alpha report: mean IC, a monotone quantile ladder, and gross-vs-net side by side.

​
Return attribution

Live portfolios are decomposed into where the return actually came from:

​
Reproducibility rules

  • One .md (the research note) plus one .py (the replication) per paper, side by side.
  • Strategy code imports from the shared library, never the reverse — a broken replication can never break shared code.
  • The full test suite runs before any strategy graduates toward live deployment.

Call the library

The zrb Python API — data, backtest, beta, and alpha functions with verified signatures.