Home / DPA Insights / Quantitative Investing
Good Signals Pick Stocks. Good Rules Build Funds.
A flexi-cap backtest on Indian equities showed that the factor model picks the stocks, but portfolio construction decides what kind of fund you actually own.
Eligibility filters, a factor model and a momentum screen picked 40 stocks from a universe of 5,500+ BSE and NSE listings. The first backtest showed where the fund’s risk was really coming from. With no segment caps in place, momentum kept tilting the portfolio towards small caps. The client’s mandate was to keep volatility and beta as low as possible. The team shifted from tuning signals to designing constraints: a large-cap minimum, a small-cap cap, a single-stock limit and a mega-cap core. Python automation cut each backtest run from hours to minutes. That made it affordable to try breaking the first idea instead of defending it. The factor model decides which stocks look attractive. Portfolio construction decides what kind of fund you actually own.
The Rules mattered more than the Model
The first backtest told us something no single stock pick could: where the fund’s risk was really coming from. We were building a flexi-cap strategy for an Indian fund house. The universe was 5,500+ stocks listed on the BSE and NSE. Eligibility filters narrowed the field, a factor model scored the rest on fundamentals, and a momentum screen kept the 40 stocks with the strongest recent price trend. It is the same broad recipe behind most factor indices, and we expected strong selection to deliver a strong portfolio.
Momentum has a favourite neighbourhood
Over a 20-year backtest, momentum kept tilting the portfolio towards small caps. With no segment caps in place, volatility ran near 33% a year, and beta, the fund’s sensitivity to market moves, sat around 1.5. For a client whose mandate was to keep both as low as possible, that was a clear signal.
So we shifted from tuning signals to designing constraints. At least 10 large caps had to make the top 40. Small caps were capped at 30% of the portfolio, with a 3% single-stock limit. Later, to anchor the core, we forced in the top five mega caps and tested two weighting schemes: fixed weights, and weights linked to their Nifty 50 share, within floors and caps.
Twenty minutes changes the questions you ask
Next, we stress-tested the factor model itself, varying the weights across quality, growth and value, and the 12 ratios within them, over 50+ iterations. Where data platforms lacked a ratio, we derived it from raw financials. Python automation cut each run from 7–8 hours to 20–30 minutes. At eight hours a test, you defend your first idea. At twenty minutes, you can afford to try breaking it.
The final backtest brought volatility down to about 18% and beta to 1.07, with roughly 4% annualised alpha over the Nifty 500 TRI (total return index, dividends reinvested). Upside capture was 121% against downside capture of 94%: the fund rode the rallies harder and fell less in the corrections.
These are DPA backtest results, not live performance.
The lesson
In a rules-based fund, the factor model decides which stocks look attractive, but portfolio construction decides what kind of fund you actually own.
The signals find the returns. The rules shape the ride.
At a glance: two layers of a rules-based fund
| Factor model and signals | Portfolio construction rules | |
|---|---|---|
| What it decides | Which stocks look attractive | What kind of fund you actually own |
| Its job | Find the returns | Shape the ride |
| What gets tuned | Weights across quality, growth and value, and the 12 ratios within them | Segment caps, a large-cap minimum, a single-stock limit, a mega-cap core and its weighting scheme |
Test the rules before you launch the fund
If your team is building a rules-based or factor strategy, we can backtest your portfolio construction rules alongside your signals, so you see what kind of fund you will actually own.
Book a working sessionFrequently asked questions
What is the difference between a factor model and portfolio construction?
In a rules-based fund, the factor model decides which stocks look attractive. Portfolio construction decides what kind of fund you actually own. The signals find the returns. The rules shape the ride.
What happens when a momentum strategy has no segment caps?
In this 20-year backtest of a flexi-cap strategy on Indian equities, momentum kept tilting the portfolio towards small caps. The client’s mandate was to keep volatility and beta as low as possible, so that was a clear signal. The team shifted from tuning signals to designing constraints.
Which portfolio construction constraints did the flexi-cap backtest use?
At least 10 large caps had to make the top 40. Small caps were capped at 30% of the portfolio, with a 3% single-stock limit. Later, the top five mega caps were forced in to anchor the core. Two weighting schemes were tested for them: fixed weights, and weights linked to their Nifty 50 share, within floors and caps.
Why does faster backtesting change factor model research?
At eight hours a test, you defend your first idea. At twenty minutes, you can afford to try breaking it. Python automation made it practical to vary the weights across quality, growth and value, and the 12 ratios within them, many times over.
