On this page
The short answer
Does Birch do better if it never touches the biggest stocks? Its results stayed strong without them, but random picks from the same smaller stocks rose just as much. The gain came from the stock list, so we left the list unchanged.
Today's large caps compounded for years, so a trading model working on them may not beat simply owning them. We asked what happens if the model works only on stocks that were mid or small caps at the time.
How did we test it?
- The stock list: each month, the biggest 50, 100, 150 or 200 stocks were taken off the tradable list. "Biggest" means most traded. We use that as a stand-in for market cap, because we have no point-in-time market cap.
- Nothing else changed. Birch ran exactly as adopted, with the shared test setup: ₹10 lakh, 1.5% risk per position, at most 12% in one stock, 30 positions, costs, and trades at the next day's open.
- The checks: 6 start dates, three times the slippage, the 5 best stocks removed, and 2023 onwards on its own.
- The random test: for each list, 20 portfolios that picked stocks at random from the same list, with the same sizing.
What did we find?
No change to the stock list.
Birch does not need the large caps. On the usual stock list it made 27.2% a year. With the top 50 to 200 removed it made 26.3–31.2%, and even its worst start date stayed above 23%.
But random portfolios rose just as much. Their median went from 20.2% on the usual list to 26.0% without the top 200. At one start date, Birch's lead over the random median stayed about 6 points (26.2% against 20.2%, then 31.7% against 26.0%). Small and mid caps simply earned more in 2016–2026, and any structured method on them earned more. The extra growth came from the stock list, not from better picks.
Removing the top 100 came closest to an improvement: 30.3% a year, 26.3% with the 5 best stocks removed, and 18.4% from 2023 with them removed, against 14.8%. It fell deeper, 32% against 27%. It is one result out of five, and two of the other lists disagree.
What failed?
- Holding up in the tougher checks. From 2023, with the 5 best stocks removed, the result fell from 14.8% to 10–13% for three of the four lists.
- Worst falls grew to 31–45% once 150 or more of the biggest stocks were removed.
- Concentration. Without the top 150–200, a single stock made 10–13% of all profit on average.
- Removing only the top 50 was the worst case on most measures: the biggest stocks carried part of the 2023 return.
- Other versions in the same test. Ten machine-learning, sizing and market-condition variants were run too, before exchange-traded funds (ETFs) left the stock list. None beat plain Birch on both tough measures for every list.
What are the limits?
- The price data holds only companies still listed today. That flatters every result here.