Someone chose the rule, and choosing the rule is the active part

Same companies, different rule, different fund

A smart beta fund tracks an index built on something other than company size — a chosen characteristic, equal weights, or a measure taken from the accounts. No manager picks the holdings; a published rule does. The rule itself was picked, which is why the label is only half true.

Take three companies and nothing else. The largest is worth ₹6 lakh crore, the second ₹3 lakh crore, the third ₹1 lakh crore. A conventional index weights each by what it is worth, so the portfolio sits at 60%, 30% and 10%. An equal-weight rule puts a third of the money in each. Same three companies, two entirely different portfolios.

Run a year through both. The largest rises 10%, the middle is flat, the smallest rises 40%. The conventional index returns 0.6 × 10 + 0.3 × 0 + 0.1 × 40, which is 10%. The equal-weight version returns the plain average of 10, 0 and 40 — 16.7%. Not one company was selected or rejected; the gap is the weighting rule alone. The figures are chosen to make the arithmetic visible and describe no actual index.

That gap is the entire product. Funds built this way are sold as smart beta, factor or strategy index funds, and the family name matters less than the shared structure: a rule written down in advance, applied without discretion, wrapped in an index fund or exchange-traded fund. The characteristic the rule ranks on — company size, price against earnings, return over the past year, stability of profits — is what the industry calls a factor, and factor investing is where those characteristics and the claims made for them are set out in full.

The direction of the tilt is mechanical and worth stating plainly, because it is the one part of this that is not debatable. Taking weight away from the largest companies and spreading it evenly must, by arithmetic, leave more of the portfolio in the smaller ones. What that tilt has earned over any particular period is a different kind of question entirely, and §assumptions comes back to it.

The judgement did not disappear — it moved

Passive and active are usually treated as descriptions of how a fund is run day to day. They are more useful read as descriptions of where the judgement sits. In a discretionary fund it sits with a manager, exercised continuously. In a rules-based fund it sits in the rule, exercised once, by whoever wrote it.

Open any rules-based index and there are at least three discretionary choices inside it, none of which a rule can make for itself. Which companies are eligible before the ranking starts — the largest hundred, the largest five hundred, everything above a liquidity floor. Which metric does the ranking, and how it is defined when the accounts do not agree. And how often the portfolio is rebuilt, plus how many names are allowed to change when it is.

Change any one of those and the fund is a different fund holding different companies. A value rule that ranks on price against book value and one that ranks on price against earnings will disagree about banks, about firms with few physical assets, and about anything that has just written down an asset — which is the argument made in full in value versus growth. Neither ranking is wrong. Both are choices.

There is a quieter point underneath. A rule reaches the market because it looked good when it was tested; the version that tested badly does not become a product. So the design choice you are shown has already been filtered by its own back-history, which is a selection the label “passive” does nothing to disclose. Where that leaves the published track record is §backtest.

The cost sits between the two, and so does the evidence

The fee on a rules-based fund usually lands above a plain index fund and below a discretionary one, and that is the easy part of the comparison. The harder part is that the three products differ in what has to be true for each to work, and the fee is not the place that shows up.

Read the last two rows together, because that pairing is the actual purchase decision and the fee is a poor guide to it. The middle column is cheaper than a manager and narrower than a market, and it asks the buyer to be right about a rule instead of right about a person.

One measurement trap deserves naming, because it is invisible on a factsheet. A rules-based scheme is compared against its own custom index — the very rule being sold. If it lands close, that says the fund executed the rule cleanly, which is useful and is what tracking difference is for. It says nothing at all about whether the rule was worth following, because the rule is on both sides of the comparison. To ask that second question you need the plain cap-weighted index, which is not the benchmark the factsheet is built around.

The fee itself still compounds the way every fee compounds, on the whole corpus, every year, including on the growth the previous years' fees would have earned — the arithmetic is worked through in the expense ratio. What the fee does not contain is the subject of the next section.

Only one weighting rule rebalances itself

Here is the mechanism most explanations of smart beta skip, and it is arithmetic rather than opinion. A cap-weighted portfolio needs no trades to stay cap-weighted. Every other rule has to trade to stay itself.

Take the same three companies and the same year. The cap-weighted book started at 60 / 30 / 10 and, after the first rose 10%, the second went nowhere and the third rose 40%, the values stand at 6.6, 3.0 and 1.4 — which is 60.0%, 27.3% and 12.7% of a total of 11.0. Those are the correct new cap weights. Nobody placed an order. The prices did the rebalancing, because in a cap-weighted index the weights and the prices are the same thing.

That claim carries a condition, and it is worth stating rather than letting the arithmetic imply more than it shows. Price movement alone forces no trades on a cap-weighted book. Other things still do — a company joining or leaving the index, shares being issued, a revision to how much of a company counts as freely tradable. Those are events, arriving when they arrive. A cap-weighted book trades when the world changes; a rules-based book trades then and on its calendar as well.

Now the equal-weight version. Start ₹1 lakh in each, ₹3 lakh in total. After that year the three holdings are worth ₹1.1 lakh, ₹1 lakh and ₹1.4 lakh, a total of ₹3.5 lakh, so equal weights now mean ₹1.17 lakh each. Restoring them takes buying about ₹6,700 of the first, ₹16,700 of the second and selling ₹23,300 of the third. That is roughly 6.7% of the portfolio changing hands after a year in which nothing dramatic happened, and the figures are the illustration's own, chosen for clarity.

Scale that up and it stops being small. A rule that holds the 30 strongest performers out of a hundred and rebuilds twice a year, replacing half its list each time, turns over its entire book once a year — again, an illustration with its inputs stated rather than a measurement of any index. Every one of those trades pays a bid-ask spread, brokerage and whatever the market moves against a large order arriving on a known date. Impact cost — the price movement your own order causes — is the part no fee table contains.

These costs are borne by the scheme, which means they come out of the fund's per-unit price — its net asset value, or NAV — and appear nowhere else. A saver comparing two funds on expense ratio alone is comparing the part of the cost that is advertised. The named mistake here is precise and common: treating the fee as the cost of a rule that trades. The fee is the cost of running the fund; the turnover is the cost of the strategy, and a rule that rebuilds often can spend more of that second kind than it saves on the first.

There is a second-order version of the same problem. A rule that rebalances on a published schedule announces in advance roughly what it will buy and sell, and a large enough fund trading a known list on a known day is trading against people who have read the same methodology. Whether that costs anything measurable in Indian conditions is an empirical question this article does not answer.

What the rule is betting on, said out loud

Every rule is a compressed argument about how markets work, and the argument is rarely printed next to the fund name. Six assumptions sit inside a typical factor rule, and a buyer who cannot state them is buying the conclusion without the premise.

• The metric is comparable across every company being ranked. Book value carries very different information for a bank and for a company whose main asset is its people. An earnings ranking depends on which twelve months it uses and how one unusual quarter is treated.
• The pattern survives being published. A rule is a description of the past that, once sold, becomes a standing instruction to buy the same things. Whether a pattern persists after money arrives to exploit it is contested, and it is not settled by the backtest that found it.
• The relationship is stable across conditions. A rule is calibrated on periods in which its ranking held together, and it carries forward the assumption that the ranking keeps meaning the same thing afterwards. The conventional reading is that holdings move together more in sharp falls than in calm stretches — which, if it holds, arrives exactly when a spread of holdings was supposed to help. Nothing in the calibration asks that question.
• One period is enough. The ranking is computed at the rebalance and assumed to hold until the next one. Nothing in the design says what happens if the ranking reverses a fortnight later; the rule simply waits, holding what it bought, until its calendar comes round.
• Returns are well behaved enough for the statistics quoted alongside the rule to mean what they appear to. Standard deviation, and every figure built on it, assumes a distribution without heavy tails — rare, very large moves arriving more often than the bell curve allows for. The rule itself does not depend on that assumption holding; the risk numbers printed beside it do, which is the objection set out in the risk measures.
• The rule survives size. A backtest fills every order at the historical price, because a backtest is not a participant. A live fund large enough to matter moves the prices of the smaller constituents it is required to buy, and the rule's edge and its capacity are not independent of each other.

None of that makes a rule illegitimate. It makes the rule a bet with conditions, and the conditions are knowable in advance in a way that a manager's future judgement is not. That is a genuine advantage of the format, and it is wasted if nobody reads them.

The evidence question has to be kept separate from all of the above, and this is where most writing on the subject slips. That an equal-weight rule tilts towards smaller constituents is mechanical, and nothing about the past is needed to establish it. That the tilt has been rewarded is a claim about a dataset, and it is worth no more than the three things attached to it: the universe it was measured on, the period it covers, and whether the cost of running the rule was subtracted before the answer was taken. Change any one of the three and the same rule can return a different verdict, which is why the question is contested rather than settled. A sentence of the form “this factor earns a premium”, with none of the three attached, is not a finding. It is a slogan with a decimal point.

A back-history is not a track record

Almost every rules-based index arrives with a chart running back a decade or more. Most of that chart was computed after the rule was written, by applying it to old prices. The line is real arithmetic and it is not a record of anything anyone experienced.

Three things separate a back-computed line from a live one, and each pushes the same way. Nobody paid the costs on the computed part — the spreads, the impact, the brokerage on every rebalance the chart implies. The data was often used as it reads now rather than as it read then, so a figure later restated by the company is treated as though it had been known at the time. And the rule was chosen with the answers already visible; the version that produced the flattering line is the one that got published, out of however many were tried.

That last point is worth being concrete about, because it is not an accusation of bad faith. Test twenty variations of a ranking rule — different lookbacks, different eligibility floors, different rebalance dates — and some will look better than the rest on any given history, whether or not any of them describe something real. The one that looks best is the one that gets a name and a fund. This is selection, and it operates quietly even when everyone involved is honest.

The apparent length of the record does not rescue it either. A rule rebalanced twice a year over fifteen years has made thirty decisions, not fifteen years' worth, and a fifteen-year chart drawn from overlapping windows overstates how much independent evidence sits underneath it — the mechanism is set out in rolling returns. Historical figures describe what happened, not what will happen; past performance does not indicate future results.

The practical form of this is one question rather than a formula: ask on what date the index went live, and read the part after that date separately from the part before it. They are different kinds of object printed in the same colour.

The questions that actually separate one of these from another

Given all of the above, a buyer's useful work is not comparing fees. It is establishing what the rule bets on and what would have to be true for the bet to pay. Seven questions, each answerable from documents that exist.

1. What does it rank on, and how is that metric defined when the accounts are ambiguous? A one-line answer in a brochure is not the methodology.
2. What is the eligible universe before the ranking starts? A rule applied to the largest hundred companies and the same rule applied to the largest five hundred are two different products.
3. How often does it rebalance, how much of the portfolio typically changes when it does, and what has that turnover cost the scheme in practice?
4. What happens at the edges — when a constituent's metric is missing, stale, or negative, and when a company is suspended or leaves the universe between rebalances?
5. When the index went live, as opposed to when its computed history starts? Everything before that date is arithmetic on old prices.
6. How does it compare against the plain cap-weighted index, not only against its own custom benchmark? Only the first comparison tests the rule.
7. What would look bad, and has that happened inside the record on offer? A rule whose weak period is absent from the chart has not been shown failing, which is not the same as never failing.

The last one carries the most weight and is asked the least. Every rule has conditions under which it trails the market for years at a stretch, and the arithmetic of holding it is that the rule only has a chance to work if it is held through exactly those stretches. A rule abandoned during its bad period delivered the cost and none of whatever it was supposed to deliver — which is the same trap that catches concentrated sectoral and thematic funds, arriving here in a more respectable outfit.

One framing helps hold all seven together. Weighting by company size is itself a rule, with its own assumptions — that the market's collective valuation is the right anchor, and that whatever it currently concentrates in is worth owning in that proportion, a property explained in market capitalisation. Choosing anything else is not a step from no opinion to an opinion. It is a swap of one opinion for another, with the trading cost of the swap attached, and where a rule's returns come from once you strip the market's own movement out is the subject of alpha and beta.

Testing a rule rather than arguing about it

Everything above is either arithmetic or a question to put to a document. How a particular rule has actually behaved — through which periods, with what worst stretch, and how much of it survives once the rebalancing is paid for — is something to ask of a price history.

FNOTrader's Mutual Funds app runs on the full AMFI NAV history — around 34 million rows of per-unit prices — and reports rolling-return distributions across every start date available, alongside comparison against a chosen benchmark and maximum drawdown. Running a rules-based scheme against the plain broad index, rather than against its own custom benchmark, is the direct way to see which of the two questions in §cost a published figure was answering. The stock backtesting universe covers roughly 2,390 stocks and 17 NSE sector and size indices, so a ranking rule can be run on constituents directly, with the rebalance frequency as an input rather than an assumption.

What a scheme's own documents will not tell you is on its factsheet in a different form — the benchmark used, the turnover reported, and the period the figures cover.

FNOTrader is not a SEBI-registered investment adviser or research analyst, and nothing here is a recommendation to buy, hold or avoid any scheme, index or security. Whether any rule is worth its cost is an empirical question about a specific product over a specific period, and the point of this article is only to make it the question that gets asked.

Common questions

Continue reading

More in Advanced Investing · App: Mutual Funds · Definitions: glossary · Free tools: calculators · All: every article