This Rule-Based System BEATS the Market Returns | Ft. Rakesh Pujara | Sanjay Kathuria Podcast EP70

ELI5/TLDR

Rakesh Pujara runs a fully mechanical stock-picking system: no opinions, no “I have a good feeling about this company,” just a weekly scan and a few rules. The idea is that stocks already shooting up tend to keep going up, so you buy the fastest movers and let them run, cutting any that stall. He claims this beat the index handily — his 25-stock portfolio roughly doubled in three years (~27% a year). The whole point is to remove the human, including himself, from the decision.

The Full Story

Why momentum at all

Momentum is the observation that things in motion stay in motion. A stock that’s rising tends to keep rising; one that’s falling tends to keep falling. Pujara points to research stretching back 200 years showing momentum has outperformed most other strategies. The Indian numbers he cites are stark: the Nifty 500 returned about 12% a year over the last 20 years, while the Nifty 500 Momentum 50 index returned about 21% over the same stretch.

“बहुत बड़ा गैप होता है, नाइन परसेंट, वो भी 20 साल के लिए।”

The catch is that momentum doesn’t look at value or fundamentals. It only asks which horse is running fastest. To do that without guessing, you need a rigid framework — “हम तुक्का नहीं, तरीका सीखना चाहते हैं” (we want to learn a method, not take a wild guess).

The emotional engine underneath is a Peter Lynch line: people “cut the flowers and water the weeds.” They sell winners early to lock in gains and cling to losers hoping they’ll come back. Momentum forces the opposite — it punishes the laggards by kicking them out and lets winners keep compounding. Over a long enough period, only winners survive in the portfolio. Pujara goes further than most: he refuses to rebalance, because rebalancing is just disguised profit-booking. If a holding swells from 4% of the portfolio to 15%, he leaves it alone. The extra return comes precisely from that swelling.

The system, step by step

He lays it out as if building a portfolio from zero.

Universe. The Nifty Total Market Index — about 750 stocks (100 large-cap, 150 mid, 250 small, 250 micro). Staying inside this index is his one concession to fundamentals: it filters out the thinly-traded, operator-driven names where price can be manufactured. The constituents are public on the NSE site.

Position count. Maximum 25 stocks. On ₹1 lakh, that’s ₹4,000 per name — a 4% starting weight each.

Entry signal. This is the clever bit. He uses Bollinger Bands, but with an unusual setting. The standard band sits at 2 standard deviations from a moving average — that captures about 95% of price action. He pushes the upper band out to 3.7 standard deviations, which isolates only the most extreme ~0.5% outliers. Out of 750 stocks, roughly 5-6% (about 25 names) ever close beyond it. The logic: he doesn’t want normal stocks, he wants the genuinely extraordinary ones tearing away from their own statistical range.

“आपको ऐसे एक्स्ट्रा-ऑर्डिनरी 25 शेयर चाहिए… आपको नॉर्मल स्टॉक चाहिए ही नहीं।”

Cadence. Run the scan Friday after the close. Any stock that closed above the 3.7-SD band gets bought Monday morning at the open. Nothing else happens Monday-to-Friday. He literally does it over morning tea.

If 40 stocks qualify but you only have room for 25, a secondary filter breaks the tie — relative strength, or 12-month rate of change. In practice they rarely all qualify at once; signals trickle in. Once the 25 slots are full, the 26th signal is ignored until a slot opens via a stop-loss exit.

Getting out — three exits

Initial stop loss of 20%. Buy at 100, and if it closes below 80 on a Friday, you’re out Monday. That slot then goes to the next fresh signal, or to cash.
23-week moving average. Once a stock is positive, you trail it with its 23-week average. A weekly close below that line and you exit. This is the normal profit-trailing mechanism.
ATR filter (Average True Range, 14-period, 1.8x). This handles the vertical rockets. When a stock has, say, gone from ₹2,500 to ₹8,000 in a few months, the 23-week average is so far below that you’d give back half the gain waiting for it to break. The ATR filter sits closer and protects against a sudden reversal — a corporate-governance shock, a cancelled order, news flow turning. You can profit using just one or two of these, but he advises all three.

A fourth safety valve is the 10% cap on any single position. He tells a cautionary tale: a client put 90% into one stock hitting upper circuits daily, it reversed, and they bailed at a big loss. With 5-10% sizing they’d have survived.

FIFO and a market thermometer

Holdings are managed first-in-first-out. The earliest entries — the ones that triggered when momentum first started, the “pole position” stocks — get ridden longest. Stocks arriving “late in the cycle” (laggards finally catching up, often already rolling into consolidation) he’s wary of. And when suddenly 200 stocks qualify instead of 25, that’s the thermometer running hot — the market is overheated and he treats it as a caution flag.

Price is the only oracle

Most momentum investors rank by 12-month return. Pujara argues that in a world where information discounts almost instantly, price itself is the indicator — it already contains everything.

“प्राइस मूव कौन करता है? बड़े लोग करते हैं… उनके पास फर्स्ट इंफॉर्मेशन होता है, और आपके पास प्राइस है।”

His metaphor: the big players — promoters, institutions with research teams — have first information. The retail investor only has price. So “we are pillion riders on price.” When the driver gets off, you get off. The system makes no sector bets; whatever runs — China+1 manufacturing, defence, data centres, optic fibre, energy — the scan catches it automatically, as long as it’s inside the 750-stock universe.

Where money actually lives (200 years)

To justify heavy equity weighting, he runs through the famous Siegel-style long-run numbers. Over 200 years, $1 becomes: about 4 cents in cash (purchasing power destroyed), ~$3 in gold (barely beats inflation), ~$157 in T-bills, ~$350 in bonds — and roughly $1.6 million in stocks, about 6.85% real return. The takeaway: if you’re not in equity, you’re missing out, and momentum deserves a meaningful slice within equity.

The honest caveats

He admits the system’s weak spot. When momentum breaks market-wide, drawdowns are brutal — 35% in 2020. It eventually recovered, but standing through it is hard. The improvement he’d like: a rule to rotate into liquid/cash or gold when momentum is lost, sacrificing some return for stability. He also doesn’t chase being #1 — the top spot changes every year. He aims to consistently stay in the top 5-10% of strategies. And he concedes passive index funds (Nifty 50) are perfectly fine for very large pools of money that would otherwise move the market; rule-based direct stocks are for manageable sizes.

Key Takeaways

Momentum’s edge in India: Nifty 500 ≈ 12% CAGR over 20 years vs Nifty 500 Momentum 50 ≈ 21% — a ~9% annual gap.
The 3.7-SD trick: Standard Bollinger Bands use 2 SD; widening to 3.7 SD isolates the ~0.5% statistical outliers, which naturally produces ~25 qualifying stocks from a 750-stock universe. Entry = weekly close above that band.
Weekly rhythm: Scan Friday close, buy Monday open, do nothing in between. Reviewable monthly or quarterly too — consistency matters more than frequency.
Three exits, not one: 20% initial stop, 23-week MA trail for normal trends, ATR(14, 1.8x) for vertical movers that have left their MA far behind.
Don’t rebalance winners: rebalancing is profit-booking in disguise. Let a 4% position grow to 15% on merit — the alpha lives in that asymmetry. Only trim in extreme cases (25-30%+) with momentum loss.
10% hard cap per stock as a wipe-out guard; FIFO management favours early “pole-position” entries over late-cycle laggards.
Market overheating signal: when far more than 25 names qualify, the market is hot — be cautious.
Price over rate-of-change: he treats price action itself as the momentum indicator rather than ranking by 12-month return, on the view that information now discounts almost instantly.
No sector forecasting: the system mechanically catches whatever theme is running (EMS, defence, data centres) as long as it’s in the index universe.
The product: CWM MILT 25 (Momentum Investing Long Term, 25 stocks), a smallcase launched 22 March 2023; backtested to 2016. Claimed ~27% CAGR over ~3 years (₹100 → ₹211) vs ~16% for Nifty 500.

Claude’s Take

This is one of the more concrete “here’s my actual system” interviews you’ll find — he gives real parameters (3.7 SD, 23-week MA, ATR 14 at 1.8x, 20% stop, 10% cap), which is rare. Most momentum talkers stay vague. The framework is internally coherent and the no-rebalancing, ride-the-winner discipline is the genuinely useful insight, because it’s the part humans find emotionally impossible.

Now the BS filter. This is a promotional interview for his smallcase, so treat the performance claims as marketed, not audited. A ~27% CAGR over three years that began in March 2023 is largely a function of a roaring small/mid-cap bull market — the same period flattered almost every Indian momentum strategy. The honest tell is his own admission of 35% drawdowns and the missing “go to cash” rule: momentum systems make their money in trends and hand a chunk back in crashes, and three years isn’t a full cycle. The 3.7-SD Bollinger entry is presented as precision but is really just a knob he tuned on backtested data — there’s nothing magic about 3.7 versus 3.5 or 4.0, and over-fitting is the permanent risk with this kind of thing.

The 200-year asset-class numbers are real (Jeremy Siegel’s research) and worth carrying around. The “price is the only indicator, we’re pillion riders” framing is a clean way to think about following smart money. A 7: substantive, specific, refreshingly mechanical, but single-source, product-linked, and resting on a backtest plus one favourable bull run. Useful as a template to understand, not as a track record to trust.