Market-Neutral Investing: Isolating the Factor From the Market

A long-only momentum portfolio is two bets stapled together. One is the momentum factor: winners keep winning. The other is simply being long equities. In a year when the Nifty falls 20%, the second bet swamps the first no matter how well the ranking worked.

If what you want is the factor, you have to remove the market. That is what a long/short book does, and it is worth understanding both why it works and why the version most people imagine is not available to an Indian retail investor.

The academic factor was always long/short

This is worth stating because it inverts how most people encounter momentum. Jegadeesh and Titman’s 1993 construction was not “buy the winners.” It was a zero-cost portfolio: long the top decile, short the bottom decile, funded by the short proceeds. Every subsequent factor — HML, SMB, quality, low volatility — is defined the same way, as a long-minus-short spread.

The published premia are spread returns, not long-only returns. When a paper reports momentum earning 1% a month, that is the winner-minus-loser spread. A long-only momentum portfolio earns some of that spread plus the entire equity risk premium plus the entire equity drawdown.

So the long/short version is not an exotic derivative of the strategy. It is the strategy; long-only is the approximation.

What “market-neutral” buys and does not buy

Making the two arms carry equal gross exposure removes most of the shared market beta. What is left is the spread between the ranking’s top and bottom.

That is a genuinely different return stream — roughly uncorrelated with the index, which is why institutional allocators pay for it. But three things it is not:

Not risk-free. Daniel and Moskowitz’s work on momentum crashes matters more here than in a long-only book, because in a market-neutral construction the crash arrives entirely through the short leg. At a bear-market bottom, the beaten-down losers you are short rebound hardest. A long-only momentum portfolio merely lags in that scenario; a market-neutral one loses on both arms at once.

Not free of borrow cost. A short position pays a borrow fee for as long as it is held, and that fee is not constant — it spikes precisely on the crowded, hard-to-borrow names a bottom-decile screen selects. Our engine does not yet model borrow cost or check per-day shortability, which means this template’s short leg is modelled more favourably than reality. That is an honest limitation, not a rounding error, and it is worth knowing before reading the equity curve.

Not beta-zero. Equal gross exposure on both arms is not the same as equal beta. If your long book holds low-beta quality names and your short book holds high-beta smallcaps, the portfolio carries residual short beta and will lose in a rally. True beta-neutrality requires sizing by beta rather than by rupees.

How a signed book is actually built

The mechanical question is: where do negative weights come from?

In this engine, direction is expressed as the sign of the weight vector, and the strategy declares its intent up front:

• direction='long' — negative weights are dropped (the long-only clipping convention)
• direction='short' — weights are negated before sizing
• direction='both' — the pipeline’s signs are left exactly as it produced them

For a market-neutral book you want the third. The pipeline then has to produce the signs itself, which it does with two sleeves:

long sleeve:  rank top 25 by momentum  -> FixedWeighting(+0.04)   -> sums to +1.0
short sleeve: rank bottom 25 by momentum -> FixedWeighting(-0.04) -> sums to -1.0
                                    v
                        allocator, 50/50, normalize on GROSS

The detail that makes it work is that the allocator normalises each sleeve on gross exposure — the sum of absolute weights — not net. Under a net rule the short sleeve sums to −1, fails a total > 0 test, and passes through unnormalised while its long sibling gets scaled, silently mis-sizing one arm against the other. Normalising on gross means both sleeves reach 1.0 and the sign survives the division.

Net result: net exposure ≈ 0, gross exposure = 1.0.

The Indian constraint you cannot design around

Here is where theory meets the rulebook.

A retail cash-market account in India cannot carry a short position overnight. Shorts in the cash segment must be squared off within the session. To hold a short across days you need Stock Lending and Borrowing (SLB), where you borrow the shares from a lender through the exchange’s mechanism.

The SLB market in India is thin. Availability concentrates in the largest, most liquid names; borrow is often simply unavailable in the midcaps and smallcaps where a bottom-decile momentum screen finds most of its candidates. When it is available, the fee is variable and can be punitive on crowded names.

So a monthly-rebalanced short book of 25 low-momentum Nifty 500 stocks is, for many retail accounts, not placeable as written. The engine’s capability layer reflects this — it drops holdings it could not have placed, so a short book that shrinks in the results is telling you something true about the market rather than failing.

Two honest responses:

1. Treat this template as a research tool. It shows you what the pure factor spread looks like, which is genuinely useful for understanding how much of your long-only strategy’s return is factor and how much is market.
2. Use the intraday short template for something deployable. Squaring off within the session sidesteps the borrow problem entirely, because an intraday short never needs to be borrowed.

Try it

The Market-Neutral Momentum template runs 12-1 momentum on the Nifty 500, holds the top 25 long and the bottom 25 short at equal gross exposure, and rebalances monthly.

The comparison that teaches most: run it against the long-only Momentum 30 over a window including 2020. The long-only version will have a far larger drawdown in March and a far larger recovery after. The gap between those two shapes is the market leg — the thing the long/short construction removes, for better and for worse.

Further reading

• Short selling in India: what a retail account can actually do is the detail behind the constraint that shapes this entire strategy
• Momentum investing in India: why the best signal skips last month for the long-only version of the same signal
• The low volatility anomaly on a factor whose long and short legs behave quite differently from momentum’s

Glossary: beta, alpha, securities lending and borrowing, factor investing.