Independent Research · Indian Equities

Testing Market Ideas Against Reality

An independent research project testing whether published market hypotheses survive statistical scrutiny and real trading costs in Indian equities.

17 completed research tests, two tracks Freeze-before-test protocol No strategy has yet met the project's threshold for deployment — the reasons matter

Featured case study

Statistical significance wasn't enough

volume_shock, built on published research linking price shocks and trading volume, cleared the project's frozen statistical threshold at two horizons — 10-day (t=2.21) and 15-day (t=2.45).

Clearing that bar was the first test, not the last.

Real trading costs — modeled on actual Indian brokerage, exchange, and statutory charges — consumed 82–125% of the gross return in every version tested.
A concentration check found 5 of 77–116 rebalance periods accounted for 395–2,052% of the total cumulative gain. Remove them, and every variant goes net-negative.

Verdict Statistically significant, but not sufficiently robust for deployment. The apparent edge was economically fragile after real costs and heavily dependent on a small number of extreme periods — not broad-based enough to trust.

Read the full case study →

What I built

A freeze-before-test research pipeline: hypotheses and evaluation criteria fixed before testing, autocorrelation-corrected significance testing, correction for testing many hypotheses at once
A ground-up transaction-cost model built from real Indian brokerage, exchange, and statutory rates
Point-in-time historical universe reconstruction — Nifty 50 membership, Dec 2008–Jul 2021, verified against a real corporate action (Satyam's 2009 delisting)
A concentration and robustness layer to catch results carried by a handful of extreme periods rather than genuine breadth
A version-controlled, timestamped research record for every closed test

View methodology →

Research at a glance

Outcome	Count
Deployable strategies	0
Statistically significant, failed robustness scrutiny	1
Inconclusive	2
No signal / wrong-signed / no transfer	14

Across two tracks — 14 factor hypotheses drawn from published research, and 3 externally-sourced strategies tested for transfer to Indian markets.

View the complete research registry →

Ongoing — not yet confirmed

Ongoing exploratory research

Momentum's below-trend pattern

Within closed tests, momentum's underperformance below the market's long-term trend replicated independently across two historical eras — 7 of 7 periods at 21 days, 4 of 4 at 63 days.

Liquidity and factor behavior

Across a 29-cell sweep, momentum-style signals held in the most liquid names, reversal-style signals in the least liquid — consistent with a 2023 academic prediction for Indian markets. Not yet corrected for testing multiple comparisons at once.

What's next

Next research priority: short-horizon reversal across the 5–20 day window, an area not yet covered by this project's completed test set. Active, not archived.

About

I'm Shaurya, an MBBS student interested in how rigorous evidence and decision-making frameworks translate across disciplines. Pre-specifying a hypothesis before testing it, holding out data the method has never seen, correcting for how many things you tested, and reporting negative results honestly are standard practice in evidence-based medicine — I applied the same discipline to markets instead.

More about this project →