Skip to main content
Precision CapitalAcademy LLP

Methodology

How the research is done

The standard applied to every study published here, and the reproducibility commitment that goes with it.

Standards

Eight standards

These are the checks a study has to pass before it is published here. They are stated publicly so that a reader can hold a paper to them.

  1. 01

    Specification before estimation

    The question, the event definition and the test are written down before the data is examined. Analyses added later are disclosed as exploratory rather than presented alongside pre-specified ones.

  2. 02

    Explicit event definitions

    Anything called an event has a rule that another person could implement from the description. Where the rule has a threshold, the threshold is a parameter and its sensitivity is reported.

  3. 03

    Confound control

    The obvious alternative explanation is controlled for before a result is reported, not mentioned afterwards as a caveat.

  4. 04

    Correct inference for the data

    Intraday observations from the same session are not independent. Day-clustered and block-bootstrap standard errors are reported alongside trade-level statistics, and conclusions rest on the more conservative figure.

  5. 05

    Family-wide multiple-comparison correction

    Correction is applied across every test run, and the full family is disclosed — including the tests that did not survive.

  6. 06

    One held-out evaluation

    Out-of-sample periods are held back until the in-sample analysis is frozen, and evaluated once. A specification that fails is discarded rather than re-tuned.

  7. 07

    Negative results published

    A specification that fails out of sample is written up with the same care as one that succeeds.

  8. 08

    Traceable figures

    Every published number traces to code that produced it from raw data, and the pipeline is runnable end to end.

In practice

What this looks like applied

These standards are only worth anything if a published study can be checked against them. Each paper documents its own processing pipeline, stage by stage, on its own page.

Current study

Open Interest Repositioning around Intraday Price-structure Breaks: Evidence from NIFTY 50 Index Options

Its pipeline, hypotheses, robustness sweeps and out-of-sample evaluation are set out in full — including the hypothesis that did not hold.

Read the paper

Replication

Reproducibility commitment

A result nobody can re-derive is an anecdote. This is what is offered, and what is not.

What is provided

  • For each paper, a consolidated notebook that reproduces its headline figures, run end to end against the source data before publication.
  • A full description of any event-detection or selection rule, its parameters and its thresholds.
  • The complete family of tests run, including those that did not survive correction.
  • Robustness sweeps for the parameters a reader would reasonably want to vary.

What is not provided

  • Licensed vendor data. Market-data archives are not redistributed; each paper's disclosure statement records what can be shared and on what terms.
  • Trading rules or strategies. The published research describes what was tested and what was found; it is not a strategy to follow, and no paper here should be read as one.

Requests

Questions about method, data or replication are welcome through the contact page. Corrections are particularly welcome — if something here is wrong, the useful response is to fix it.