Section 4

Early movers

Did stocks that led in the first weeks of a financial year go on to lead the whole year? This page shows the headline version of that statistic and the properly-tested version side by side, because they say materially different things.

Survivorship bias applies here too

All early-mover statistics are computed on the 2026 Nifty 500 constituent list, so stocks that were dropped or delisted over the decade are absent from both the signal and the outcome — which flatters any "leaders keep leading" result.

The same question, measured two ways

Both panels answer "were the year's winners already strong after 30 trading days?" — but they define "winner" differently. The difference is the whole point.

The headline version

—

of eventual top-30 performers were already in the top quartile of excess return after the first 30 trading days.

Lift vs the 25% baseline: —

Properly tested

—

recall against a non-overlapping forward target (top 30 by return from day 60 to financial-year end).

Lift vs the 25% baseline: —

Random baseline for both: 25% recall (a random quarter of the universe captures a quarter of the winners) and ~7.9% precision (30 winners ÷ ~380 names).

Why the headline figure is inflated

A stock that led for the first 30 days has mechanically already banked part of the year's excess return. So the 55% figure partly measures a tautology. It is not wrong — it is just answering a different question from the one that matters for a decision.

Recall and precision, both versions

Headline version — vs the full-year outcome

Signal and target overlap, so this is partly mechanical.

Properly tested — non-overlapping forward target

This is the number to use. Both versions sit far above the 25% baseline.

Headline version, by signal window

Window	Mean recall	Median recall	Mean precision	Lift

Properly tested, by signal window

Window	Mean recall	Mean precision	Recall lift	Precision lift

How strong is the relationship really?

Correlation of early excess return with subsequent performance. Column A is the overlapping (inflated) version; column B is the honest forward-looking one.

Signal window	A — vs full-year return (overlapping, partly mechanical)	B — vs forward excess (clean, non-overlapping)	B as R²

Correlation with subsequent performance — overlapping vs clean

The gap between the two columns is the size of the artefact.

The honest conclusion: an edge so small it will not survive costs

The clean forward correlation is +0.05 to +0.09 — an R² below 1%. The edge is real in the sense that the sign was positive in 6 of 10 years at day 10, 8 of 10 at day 20, 6 of 10 at day 30 and 9 of 10 at day 60. But a relationship explaining under 1% of variation is not a tradable signal once transaction costs, slippage and taxes are applied. The recall and precision lifts above the baseline (~1.3×, not 2.2×) are the more useful way to see it.

Technical characteristics of early movers

Measured over the first 30 trading days: eventual top-30 performers versus everyone else.

Trait (first 30 trading days)	Winners	Rest	Difference

What the traits do and do not say

Winners showed higher relative volume (1.40× vs 1.11×), roughly two-thirds more gap-ups (2.42 vs 1.47), about one fewer down day, and broke a 52-week high 7.8pp more often (34.0% vs 26.2%). The absence of any drawdown difference is informative: early movers were not simply names that fell less — they were being accumulated and gapping up. At this sample size these are indicative, not proven, and they are characteristics of the winners, not a validated screening rule.

Year by year

Recall / precision for the headline and tested versions, per financial year.

FY	n	Headline d30 (recall/precision)	Headline d60	Tested d30	Tested d60