Table of contents

• Two strategies, one shelf
• What each one actually does
• The turnover difference is the whole story
• What Indian tax does to the comparison
• The honest part: Coffee Can's lost four years
• Where Magic Formula hurts instead
• Which one is yours
• Can you run both?
• Key takeaways

Two strategies, one shelf

Of the five strategies we publish free, two get read more than the other three combined — and they happen to be near-opposites.

Coffee Can says: find businesses that have compounded revenue and return on capital for a decade, buy ten to fifteen of them, and then do nothing for ten years. Not "review annually." Nothing.

Magic Formula says: rank the entire market on how cheap it is and how good the business is, buy the top twenty or thirty, sell them all after a year, and repeat. Forever.

Dono serious hain — both are documented methods with real intellectual pedigree. Both have worked. And they ask for such different things from the person running them that choosing between them on returns alone is almost beside the point — you should be choosing on which set of demands you can actually live with.

Figure 1. Coffee Can and Magic Formula compared across origin, what they buy, holding period, turnover, number of holdings, rebalancing, and the maintenance each demands.

This article is the head-to-head. Including the part most write-ups skip: what happened when Coffee Can stopped working for four years, because it did, and the reason it did is the most instructive thing about it.

One thing worth saying up front: both of these translate to India cleanly. That is not true of every famous strategy — plenty depend on American market plumbing that simply doesn't exist here. These two need only company fundamentals, a screener, and patience, all of which India has in abundance.

What each one actually does

Coffee Can

The name comes from Robert Kirby, an American portfolio manager who in 1984 described a client whose late husband had quietly bought every stock Kirby recommended and then simply never sold any of them — putting the certificates in a metaphorical coffee can. The husband's portfolio comfortably beat the wife's actively managed one, mostly because one position had grown into a fortune while the active account had trimmed its winners along the way.

Saurabh Mukherjea adapted it for India at Ambit Capital, where the Coffee Can PMS launched in March 2017, before co-founding Marcellus in 2018. The commonly cited Indian filter is: at least 10% revenue growth and at least 15% return on capital employed, every year, for ten consecutive years — with separate criteria for lenders, where return on equity and loan-book growth of at least 15% replace the ROCE test. Very few Indian companies clear either version. That is the point: the filter is designed to be almost impossibly strict, because the strategy has no other risk control.

The rules after that are brutally simple. Buy roughly ten to fifteen names. Hold for ten years. Don't rebalance. Don't trim winners. Don't add to losers. Bas — kuch mat karo.

Magic Formula

Joel Greenblatt's method, from The Little Book That Beats the Market, is mechanical in a completely different way. Rank every company in your universe twice: once by earnings yield, defined as EBIT divided by enterprise value, and once by return on capital, defined as EBIT divided by net working capital plus net fixed assets. Add the two ranks together and buy from the top.

Note that Greenblatt's return on capital is not the ROCE your screener reports — his version deliberately strips out goodwill, other intangibles and excess cash, on the argument that what matters is the capital a business actually needs to operate. If you build this on a stock screener's default ROCE field, you are running something adjacent to Magic Formula, not Magic Formula.

Two more rules people skip. Greenblatt applies a market-capitalisation floor, and he explicitly excludes financials and utilities because their accounting doesn't fit the metrics. And he prescribes staggered buying — roughly five to seven names every two or three months until you hold twenty to thirty, each position held about twelve months and then replaced — rather than one annual mass sell-and-rebuy. Run properly it is a quarterly rhythm, not a single afternoon a year.

The philosophy is Graham with a quality overlay: cheap alone catches too many broken businesses, so you require the business to also be good. The mechanism is entirely unemotional — you never form a view about any company, you just follow the ranking.

One practical note if you're building this yourself: Screener.in has no ranking function. There is no RANK or PERCENTILE in its query language. You filter on Screener to get a shortlist, then export and rank in a spreadsheet. Our Screener.in manual covers the workaround.

The turnover difference is the whole story

Everything else about these two strategies flows from one number: how often you sell.

Coffee Can's turnover, run properly, is approximately zero for ten years. Magic Formula's is approximately 100% every single year. Both sit at the patient end of a spectrum that runs through positional trading and on down to swing and intraday — and the further down that spectrum you go, the more the tax and cost arithmetic below turns against you.

That difference cascades into everything:

The last two rows are where people actually get hurt, and we'll come back to both.

What Indian tax does to the comparison

Here is the part most Coffee Can versus Magic Formula comparisons leave out entirely — and where the conventional wisdom turns out to be half wrong.

The rates first. Short-term capital gains on listed equity are taxed at 20%, long-term at 12.5%, with a ₹1.25 lakh annual exemption and a 12-month holding threshold. (Dividends, bonuses, splits and buybacks each have their own treatment — our guide to corporate action events covers what each one does to your holding.) These came in with the Finance (No. 2) Act, 2024 in July 2024, and were re-enacted by the Income Tax Act, 2025 from 1 April 2026 — so they are not new, but they are now the settled position.

The intuitive argument goes: Coffee Can never sells, so it defers tax entirely and compounds on the full amount, while Magic Formula pays every year and compounds on less. Over fifteen years, the story goes, that gap becomes enormous.

Run the numbers and it is more interesting than that.

Figure 2. The same 14% gross annual return compounded three ways on a ₹10 lakh corpus, and how the never-selling advantage scales with portfolio size once the ₹1.25 lakh annual exemption is accounted for.

Take ₹10 lakh, a 14% gross annual return, fifteen years, and — importantly — apply the ₹1.25 lakh exemption properly each year, which most versions of this comparison skip.

• Never sell, pay once at the end: about ₹63.9 lakh.
• Sell annually, just past the twelve-month mark, at 12.5%: about ₹62.5 lakh.
• Sell annually, just inside twelve months, at 20%: about ₹49.2 lakh.

The first two are separated by 2.1%. Not nothing, but nowhere near the gulf the deferral argument implies — because on a ₹10 lakh book, a ₹1.25 lakh annual exemption shelters most of each year's gain. At small corpus sizes, turnover is barely taxed at all.

The gap between the second and third is 27.2%.

That is the real finding, and it inverts the usual advice. The expensive thing is not selling. The expensive thing is selling before twelve months.

The advantage grows with the size of your portfolio

The exemption is a fixed rupee amount, so it shields a shrinking share of your gains as the portfolio grows. Same 14%, same fifteen years, different starting corpus:

So the honest version of the tax argument is this: if you are starting small, tax should barely influence your choice between these two strategies. If you are running fifty lakh or more, Coffee Can's deferral is worth roughly ten percent of your terminal wealth — real money, but still not the main reason to pick it.

Two caveats. This assumes constant returns and complete annual turnover, which no real strategy has. And it assumes both strategies earn the same gross return — which is exactly what is in dispute, since the entire argument for higher turnover is that it earns more gross return to pay for the friction.

The one piece of advice that is unambiguous

Whatever your corpus, do not let an annual rebalance fall inside twelve months. The difference between 12.5% and 20% costs between 16% and 27% of terminal wealth depending on portfolio size, and avoiding it requires nothing but a calendar reminder.

Greenblatt's own refinement is sharper still, and it maps cleanly onto Indian rates: sell your losers just before the twelve-month mark, so the loss is short-term and can be set against short-term gains taxed at the higher 20%; sell your winners just after it, so the gain is long-term at 12.5%. That asymmetry is worth more than any threshold you will tune in the screen itself.

The honest part: Coffee Can's lost four years

Most articles about Coffee Can in India stop at "quality compounds, be patient." Here is what actually happened.

Marcellus — the firm that popularised Coffee Can here — publishes the track record of its Consistent Compounders portfolio, which runs the same philosophy. Their own disclosed numbers, as at 31 July 2026, split into three distinct phases against the Nifty 50 total return index:

Figure 3. Marcellus Consistent Compounders versus Nifty 50 TRI across three phases — strong outperformance, four years of significant underperformance, and a sharp recovery.

\* Phases 1 and 2 are annualised. Phase 3 covers six months and is an absolute figure, not an annual rate. All portfolio numbers are net of fixed fees.

Phase 2 ko dobara padhiye. Four years and two months of returning 2% a year while the index returned 11%. Not a bad quarter. Not a rough year. Over four years, during which anyone running this strategy watched a plain index fund beat them by a wide margin, every month, while holding businesses whose fundamentals were mostly fine. The portfolio's disclosed return on capital employed was around 24% with roughly 80% of earnings reinvested — though that is a snapshot at 31 July 2026, not a series proving it held throughout.

This is the thing to sit with before you choose Coffee Can. Not "can I tolerate volatility" — volatility is easy by comparison. Can you hold a portfolio that is losing to a Nifty index fund, for four years, while doing literally nothing about it?

Almost nobody can. Aur yahi asli baat hai — that is the reason Coffee Can works for the few who run it properly: the discomfort is the mechanism. If holding great businesses through long stretches of underperformance were comfortable, everyone would do it, and the excess return would disappear.

Phase 3 is the other half of the lesson. Anyone who capitulated in late 2025 — after four years of being wrong, which is exactly when capitulation feels most justified — missed a six-month stretch of beating the index by fourteen points. The recovery, as usual, was concentrated and arrived without warning.

One important qualification: Consistent Compounders is not the Coffee Can filter. It screens on double-digit revenue growth and return on capital above the cost of capital each year for a decade, with forensic-accounting and business-longevity overlays on top. It is the same philosophy — buy a handful of durable businesses and hold — run by the same people, not the same rule set. And it is one firm, one concentrated portfolio, with past performance saying nothing about the future. But it is the most transparently disclosed long-run record of this approach in India, and the shape of it — long good stretch, long bad stretch, sharp recovery — is what quality investing genuinely looks like from the inside. If you want to see how India's better-known individual investors have fared holding concentrated quality books, our superstar portfolio numbers and the Ashish Dhawan portfolio breakdown show the same pattern with real names attached.

Where Magic Formula hurts instead

Magic Formula doesn't make you sit through four years of doing nothing. It hurts in four different places.

Jo pasand aa gaya, wahi bechna padta hai. The ranking has no memory and no sentiment. A company you've researched, understood and become quietly fond of falls out of the top thirty and you sell it, on schedule, with no discretion. People find this much harder than they expect, and the moment they start making exceptions, they have stopped running Magic Formula.

The cost drag is real and permanent. 100% turnover means STT on every sale, stamp duty on every purchase, exchange charges, GST, and depository charges of roughly ₹15–20 per sale per stock regardless of size. Thirty names sold and thirty bought, every year. On a small portfolio this can consume a meaningful share of the edge, which is why we suggested a rough ₹1 lakh floor before running any twenty-plus-name mechanical screen.

The screen surfaces uncomfortable companies. Ranking mechanically on cheapness and return on capital will hand you cyclicals at peak earnings, companies with governance issues the market has already priced, and businesses in structural decline that look cheap because they are. This is exactly where a screen stops and equity research starts. Greenblatt's answer is diversification and discipline: hold enough names that the duds don't matter and don't override the ranking. That answer is correct and still uncomfortable to live with.

It has no financials — and that is deliberate. Greenblatt excludes banks, NBFCs and utilities by design, because his metrics don't describe businesses whose borrowings are their raw material. So this is a sector bet you take on purpose, not a bug. It is worth knowing how large a bet it is in India, where financials are a substantial share of the index.

The trap is what happens if you rebuild the screen on your screener's ROCE field without knowing about the exclusion. Screener.in shows HDFC Bank at a return on capital employed of roughly 7%, so a ROCE threshold deletes the financial sector silently, and you end up making the same bet without having decided to. Coffee Can has exactly the same issue — its 15% ROCE test excludes lenders too, which is precisely why the original Indian version specifies separate return-on-equity and loan-growth criteria for them. If your version of either strategy doesn't handle financials explicitly, it is handling them by accident.

Which one is yours

Forget which sounds more sophisticated. Answer these.

Figure 4. Which strategy fits — a decision guide across five questions on temperament, time, capital, tax and the specific failure mode each strategy imposes.

Can you hold through four years of underperformance without acting? If honestly yes, Coffee Can. If not, Magic Formula's annual rebalance gives your need to act a legitimate, rule-bound outlet — which is far safer than having that need break a buy-and-hold strategy at the worst possible moment.

Do you want to think about companies, or avoid thinking about them? Coffee Can requires deep conviction about a small number of businesses, since you'll hold them through everything — which in practice means reading annual reports properly, not just running a screen. Magic Formula requires almost no conviction about any single name; the diversification and the ranking do the work.

How much capital? Below roughly ₹1 lakh, neither runs at full width — but Coffee Can degrades more gracefully, because you can hold five to eight names once and pay the costs once. Magic Formula's twenty to thirty positions with annual turnover need more than that before the recurring friction stops eating the edge.

How large is the portfolio? On a small book the ₹1.25 lakh exemption shelters most of a year's gains, so tax should barely influence the choice. Above roughly ₹50 lakh, Coffee Can's deferral is worth around ten percent of terminal wealth. Either way, never rebalance inside twelve months.

Which failure mode can you survive? Coffee Can fails by being boring and wrong for years at a stretch. Magic Formula fails by being busy, expensive, and occasionally handing you a genuinely bad company. Pick the one whose bad days you can live with, because you will get plenty of both.

Can you run both?

Haan, bilkul — it is a sensible combination, but only under strict conditions, because these two strategies will actively fight each other if you let them.

Separate sleeves, separate capital, fixed shares. Decide the split in advance — 70/30 or 60/40, whatever suits — write it down, and let rebalancing enforce it.

No migration, ever. This is the rule that matters. A Magic Formula holding that falls out of the ranking gets sold. It does not get "moved to the long-term bucket because the business is actually quite good." That single move, which feels reasonable every time, is how two disciplined strategies become one undisciplined portfolio.

Watch the overlap. Both screens like high return on capital. A genuinely excellent business can appear in your Coffee Can core and in a Magic Formula rebalance, and you end up with far more concentration in it than either strategy intended. Check for duplicates at every rebalance and decide, in advance, which sleeve owns it.

Different scorecards. Judge Coffee Can over five to ten years and Magic Formula over three to five. Judging either on a twelve-month view will make you abandon whichever one is currently having its inevitable bad patch, which is exactly the wrong reaction.

Key takeaways

• The turnover difference explains everything else — costs, tax, time, decisions, and how each strategy fails you.
• Tax matters less than you'd think at small size, and more as you grow. With the ₹1.25 lakh exemption applied, never selling beats annual long-term selling by only 2.1% on a ₹10 lakh corpus — but 11.4% on ₹1 crore.
• Never let a rebalance fall inside twelve months. Crossing that line costs 16–27% of terminal wealth depending on corpus size. Better still, follow Greenblatt: losers just before twelve months, winners just after.
• Coffee Can's real test is not volatility. Marcellus's own Consistent Compounders returned 2% a year against the Nifty's 11% for four years and two months. Ask whether you could hold through that before choosing it.
• Magic Formula's real test is obedience — selling companies you like, on schedule, without exceptions.
• Any ROCE-based screen silently excludes banks and NBFCs. In India that is a large unintended sector bet.
• Running both is fine, with fixed capital shares and an absolute ban on moving positions between sleeves.

What this means for your process

Read both playbooks properly before you pick — the entry rules, the exit rules, and specifically the sections on when each one underperforms. Coffee Can Investing and Magic Formula are both free, no login required.

If neither quite fits, Growth at a Reasonable Price sits between them — quality-oriented like Coffee Can, but with valuation discipline and a more forgiving holding period. All three sit in the Style of Picks library alongside 128 others. When you're ready to build the screens, the Screener.in manual has copy-paste queries for both strategies above.

Everyone has a hot tip. Almost no one has a plan.

Related reading

• How to Use Screener.in Like You Actually Mean It — twelve copy-paste queries, including one for each of the strategies above, plus the ranking workaround Magic Formula needs.
• How to Read an Annual Report in India — the work Coffee Can actually requires once the screen has narrowed the field.
• Superstar Portfolio: Individual Investor Holdings, Returns & Numbers — what concentrated quality books have actually returned in India, with real names.
• Equity Research India: How It Works — where a mechanical screen stops and judgement begins.
• What Is Positional Trading in India — the next stop down the holding-period spectrum, if ten years sounds like too long.

Bharatstox publishes NSE and BSE market journalism, corporate filing explanations, live market panels and attributed research calls with visible analyst details and timestamps. Visit Bharatstox for market context, and Style of Picks for the documented strategy playbooks.

Disclaimer: This article is for informational and educational purposes only and does not constitute investment advice or a recommendation. Marcellus Investment Managers is referenced solely because it publicly discloses a long-run track record of this investment philosophy; nothing here is a recommendation of any product, manager or security. Past performance is not indicative of future results. The compounding illustrations are simplified arithmetic on assumed constant returns, not projections. Investing in securities markets is subject to market risks. Read all related documents carefully before investing.

Sources

• Coffee Can Investing playbook and Magic Formula playbook, Bharatstox Style of Picks
• Marcellus Consistent Compounders performance by phase, net of fixed fees, as at 31 July 2026 — Marcellus CCP product deck
• Capital gains rates — introduced by the Finance (No. 2) Act, 2024 with effect from 23 July 2024 and re-enacted by the Income Tax Act, 2025 from 1 April 2026 — TaxGuru summary
• Screener.in query language and the absence of a ranking function — Screener.in knowledge base
• ROCE treatment for banks and NBFCs — Screener.in on differences in reported numbers