Holdout SetEvery hypothesis registered before it was scored. Every verdict published.

A research programme that mostly says no.

I build and test trading strategies on Indian equities, alone. Every hypothesis is registered in a dated commit before it is scored, every verdict is published whether it passed or failed, and no real capital has ever been deployed.

Registered
23
Dead
22
Qualified pass
1
Built on an unproven signal
0

The complete register is the Ledger — the most useful page here, and the one that would be hardest to fake. The single pass is under a pre-registered forward test whose verdict dates were published before its first observation.

What this is not

A track record
Every figure here is simulated. No real capital has ever been deployed.
Advice
No tips, no signals, no recommendations. I am not a registered adviser.
A data shop
Nothing is for sale and nothing is downloadable.
A claim that this works
The first scheduled verdict is due around February 2027.

Research log

All entries →
  1. 18 August 2026 · DR-13, DR-25, DR-12, DR-27

    How raw NSE Bhavcopy beat premium vendors at survivorship-free data

    Why most Indian equity datasets quietly inflate every backtest run on them, how to rebuild a decade of history that doesn't, and the eight-month hole the acceptance tests found in my own archive.

  2. 11 August 2026 · DR-26, DR-24, DR-27

    We built a gate, then failed it twice in one day — on purpose

    How pre-registration works in practice, why a threshold chosen after the result isn't a threshold, and what it costs to run a research programme that can tell you no.

  3. 28 July 2026 · DR-31, DR-30, DR-36

    Eleven hypotheses, two rounds, zero trades

    A year's worth of quantitative research discipline compressed into six weeks, and what it cost to find out that nothing worked.