Studies

The research, in the open.

A Study is a locked experiment: a fixed universe, method and window, and a result anyone can replay. A new method only reaches a Signal if its Study holds up out of sample. 4 of the 5 Studies below did not fully hold up, and they stay published.

Studies

5

Names studied

886

Markets

India · US

• S-001Passed

  Indicator promotion gate

  Which listed names show an indicator edge strong enough to be worth a locked forward test?

  93 of 886 names passed the gate after multiple-testing correction.

• S-002Did not hold

  Locked forward test of promoted names

  Does the edge found in S-001 survive on data the model has never seen?

  0 of 93 names held their edge. The model trailed a naive baseline.

• S-003Mixed

  Shallow neural models vs technical rules, after costs

  Do small neural models beat classic indicator rules, and does either beat simply holding the stock once trading costs are paid?

  Neural models beat indicator rules on 447 of 498 names, but the best learned model beat simply holding on only 50 after costs.

• S-004Did not hold

  Choosing a stock hedge by volatility regime

  Does picking the hedge by how expensive options are (put when cheap, put spread in between, zero-cost collar when rich) keep most of a put's protection for much less of its cost?

  Failed. The rule cut drawdowns by 24% but still cost 71% of what always buying a put costs.

• S-005Did not hold

  Hedging a stock portfolio with index options

  Does one beta-sized put spread on SPY or QQQ, rolled quarterly, cut a 20-stock portfolio's drawdowns meaningfully for a small yearly cost?

  Failed. Drawdowns cut by 24% for 2.7 points of return a year.

Figures generated from the locked research outputs on 2026-10-22.