A focused fund of 30 stocks returned 18% a year against 15% for a diversified peer over five years, but three stocks contributed 60% of the excess. Is this repeatable skill or concentrated luck, and how would you tell?

The record is consistent with skill but does not prove it, and the three stocks are the reason it cannot yet. Without them the fund still made about 16.2%, 1.2 points ahead. But 3 points of excess at 6% tracking error is an information ratio of 0.5, which over five years is a t-statistic near 1.1; a stable edge that size takes about 16 years to separate from luck. Judge it on process: were the winners sized by conviction before they ran, and does the pattern repeat across rolling periods.

Less than it sounds, and more than the sales team would like. A cricket side that wins a series on three big innings has still won; the question is whether the batsmen were picked for those innings or happened to be at the crease. In any concentrated portfolio a few names will dominate the excess return, because stock returns are skewed: most stocks do a little, a few do a lot. So 60% from three names is what a 30-stock fund looks like whether the manager is skilled or lucky. The number that separates the two is not the concentration of the result but what the manager did before the result arrived.

Take them out and see what is left. Without the top three the fund made about 16.2% a year, still 1.2 points ahead of the peer, and Rs 100 grew to about Rs 212 against Rs 201 for the peer and Rs 229 for the full fund. The hit rate was 50%: fifteen of thirty stocks beat the peer. That is an ordinary hit rate; the edge came from the winners being bigger than the losers, which is what a good process produces and also what one lucky run produces. The residual 1.2 points from 27 stocks is itself small enough to be noise.

Longer than five years. Divide the excess by the tracking error: 3 points over 6% is an information ratioExcess return over a benchmark divided by the tracking error of that excess. It measures how much outperformance a manager gets per unit of the risk taken against the benchmark. of 0.5, good but not rare. Over five years that gives a t-statistic of about 0.5 x the square root of 5, near 1.1; a statistician would want about 2, which at this information ratio takes about 16 years. No platform waits that long, which is why the decision has to lean on process evidence rather than the return alone.

Five things, each of which can be checked from documents the fund house already has. First, the sizing: were the lender and the hospital chain 6% and 5.5% positions because the manager wrote a thesis and sized it, or because an earlier run had already made them large. The dated investment notes answer that. Second, repetition: split the five years into rolling three-year windows and see whether the fund is ahead in most of them or in one burst. Third, style: if the three winners share a factor, say all mid-cap financials and healthcare in a period those sectors led, the excess may be exposure the platform could buy more cheaply. Fourth, the sells: a skilled concentrated manager also shows up in what he sold before it fell. Fifth, capacity: the fund's size against the liquidity of its 30 names, because a strategy that worked at Rs 2,000 crore may not at Rs 10,000 crore.

Close with a view. List it, labelled as a concentrated fund whose returns will be lumpy, if the investment notes show the big positions were sized on conviction before they ran and the rolling windows show it ahead in most periods. Do not list it on the five-year number alone, and never on the 18%. The limit of the analysis: attribution tells you where the return came from, not why, and the same three stocks could be sold tomorrow; what the platform is buying is the process that chose them.