Why a recipe beats a hunch
Most stock picking is a story: a product you like, a tip, a chart, a company in the news. Stories are not useless, but they are impossible to test and easy to bend. A recipe is different. You write it once, test it on the past, and then apply it the same way to every stock, every time. The why rules beat gut feeling module covers the behavioural case. This guide is the practical one: the five parts every rule needs, with the choices Indian investors face at each step.
Part 1: the universe
The universe is the list you are allowed to choose from. Pick one NSE already maintains, so membership is objective and history is available:
- Nifty 100 or Nifty 200: large caps, deep liquidity, low trading costs, fewer surprises.
- Nifty Midcap 150 or Smallcap 250: more room for a factor to work, more volatility, wider spreads.
- Nifty 500 or Nifty Total Market: the broad market; the Nifty Total Market had 755 constituents on 30 Sep 2026.
- A sector index: narrow, for a deliberate theme, best kept as a small satellite.
The universe sets the character of everything after it. A momentum rule on the Nifty 100 and the same rule on the Smallcap 250 are two different portfolios with different falls.
Part 2: the filters
Filters remove stocks you could not hold properly. Typical ones: a minimum listing history (NSE's factor indices require a year), a minimum daily traded value so your orders do not move the price, trading on every day of the past year, and sometimes a maximum share price so a modest portfolio can still hold every name in whole shares. Filters are not where returns come from. They are where unpleasant surprises are removed.
Part 3: the score
The score ranks what is left. Use one or two factors with long published evidence, measured the same way for every stock:
| Factor | A simple measure | Tends to do well when | Tends to struggle when |
|---|---|---|---|
| Momentum | 6 and 12 month return, divided by volatility | Trends persist | Leadership flips sharply |
| Quality | Return on equity, low debt, steady earnings | Markets are nervous | Speculative rallies |
| Value | Earnings or book value relative to price | Cheap stocks re rate | Growth stocks keep leading |
| Low volatility | Lowest daily price swings | Falling or choppy markets | Strong broad rallies |
Rank on a combined score if you use two factors, for example the average of each stock's percentile on momentum and on quality. Avoid the temptation to add a seventh and an eighth factor because they improved the back test; the blog note on the factor zoo explains why that usually fails out of sample.
Part 4: the portfolio
Now turn a ranked list into holdings.
- How many: 20 to 50 for most rules. Fewer than 15 is concentrated and swings hard; more than 50 starts to look like the index.
- Weights: equal weight is simple, transparent and easy to rebalance. Score weighted or market cap weighted rules need caps.
- Caps: a maximum per stock (often 5% to 10%) and, if you like, per sector.
Position size is also a risk decision. The position size calculator shows how much one bad stock can cost the whole portfolio at a given weight.
Part 5: the calendar
Decide when the rule runs: monthly, quarterly, every two weeks. Faster reacts sooner but trades more, which costs STT at 0.1% on every buy and sell, any brokerage and fee, and short term capital gains tax at 20% on gains held 12 months or less. Add a buffer: buy a stock when it enters the top 20, but sell it only when it drops below 30. NSE uses buffers in its factor indices for the same reason, to stop names near the cut off being bought and sold every cycle.
- Choose the universeone NSE index, such as the Nifty 200 or Nifty Total Market, so membership is objective.
- Apply the filterslisting history, daily traded value, trading frequency and, if needed, a price cap.
- Compute the scoreone or two factors, measured the same way for every stock, ranked as percentiles.
- Build the portfoliothe top 20 to 50, equal weight, with a cap per stock and per sector.
- Fix the calendar and bufferthe review dates and the rank at which a holding is sold.
- Test before moneyrun it on past data with costs and tax, and on a period the rule was not built on.
How to test the recipe honestly
A back test tells you what the rule would have done, not what it will do. Three tests catch most self deception. Deduct realistic costs: charges, STT and impact. Test out of sample: build the rule on one period and judge it on a later one it never saw. And look at the worst stretch, not just the average: the deepest fall and how long recovery took. The how backtesting works module and the out of sample module go through each pitfall, from survivorship bias to look ahead errors.
Monthly model returns for Midcap Smallcap from 2022, from the published nav file.
A rule with a strong CAGR still has red months, and often runs of them. If a recipe's back test shows no bad months at all, suspect the test.
MythA good screener result is a good investment rule.
FactA screen is only part 3. Without a portfolio rule, a calendar and a cost aware test, it is a list of ideas, not a strategy.
MythThe more factors in the score, the better.
FactExtra factors mostly fit the past. Past three, a combined score rarely improves out of sample results.
A worked example: one recipe on the Nifty 200
Here is a complete recipe you could run with a spreadsheet, written the way a rule should be:
- Universe: the Nifty 200, from NSE's published constituents on the review date.
- Filters: listed at least a year; average daily traded value above ₹20 crore over six months; share price below ₹5,000 so a ₹2 lakh portfolio can hold every name.
- Score: each stock's 12 month return, skipping the latest month, divided by the volatility of its daily returns; ranked as a percentile.
- Portfolio: the top 25, equal weight at 4% each, no more than 6 stocks from any one sector.
- Calendar: the first trading day of each quarter; a holding is sold only if it falls below rank 40.
Every line can be checked by someone else and gives the same answer. That is the test of a real rule. Back test it on at least ten years with costs, look at the worst fall, and only then decide whether it deserves money. This recipe is an illustration of the structure, not a rule RupeeCase runs or recommends.
Common mistake: overriding the rule
The rule says sell a stock you have researched and like. The rule says buy one that has already doubled. Each override feels like judgement, and each one quietly replaces a tested recipe with an untested one. If you find yourself overriding often, the rule does not match your beliefs; change the rule in writing, test the new one, and follow that.
How this shows up in RupeeCase
Each RupeeCase index algo is a finished version of this recipe, published on its page: the universe, the number of stocks, the weights, the price cap and the calendar, with the model back test and every rebalance list. Midcap Smallcap runs momentum on the NSE universe, buying only stocks priced under ₹2,000: 40 stocks, equal weight, rebalanced every 2 weeks, from ₹80,000, with a model CAGR of 42.64% and a deepest fall of −26.17% as of 27 Apr 2026. The stocks pages show, for each NSE name, which rules have held it and when, so you can see a recipe's picks from the stock's side. The blog note on reading a methodology page shows how to judge any rule's write up, ours included.
Questions investors ask
How do I select stocks for the long term systematically?
Write a rule with five parts: a universe such as the Nifty 500, filters for liquidity and listing history, a score on one or two factors such as momentum and quality, a portfolio of 20 to 50 names with weight caps, and a review calendar. Test it with costs, then follow it.
What are the best criteria to pick stocks?
No single criterion works in every market. The factors with the longest evidence are momentum, quality, value and low volatility. Combining two that tend to disagree, such as momentum and value or momentum and quality, gives steadier results than any one alone.
How many stocks should a portfolio hold?
For most rules, 20 to 50. Fewer than about 15 makes each stock's fall a big hit to the whole; more than 50 starts to behave like the index, with less benefit from the rule.
How often should I review a stock selection rule?
On fixed dates, from monthly to twice a year depending on the factor. Momentum needs more frequent reviews than value or quality. Use a rank buffer so stocks near the cut off are not bought and sold every time.
Can I use a stock screener to pick stocks systematically?
A screener handles the filters and the score. You still need the portfolio rule, the calendar, the buffer and a cost aware test; without them a screener gives a list, not a strategy.
Do I need coding to pick stocks systematically?
Not to start. A spreadsheet with an index's constituents and a few measures can run a simple rule. Testing it properly over many years with costs is where code, or a platform that has already done it, helps.