A safety rule that was deleting the best returns
"Only trade liquid names" is the most respectable advice in this market. It is also, as measured on this panel, where the returns were being thrown away:
THIN stocks (under Rs 5M traded per day): +15.86%. LIQUID stocks (Rs 50M+ per day): -1.00%. [LEDGER: project_quiet_loaded_edge]
The liquidity gates in our own system "were systematically deleting the highest-alpha bucket in the system". The safety rule was the leak.
And it compounds with a second finding from the same study: a stock that is both quietly accumulated and thin measured +24.85%, while loaded-and-already-ignited measured -0.27%. The uncomfortable combination, small, quiet, illiquid, is where the number is.
The correction, stated exactly
Size to liquidity. Do not reject on liquidity.
A stock too thin for Rs 10 lac is not a bad stock. It is a small position. Those are different decisions, and collapsing them into one destroys the return.
Now the part that makes this honest
This is the lesson most likely to hurt someone, so it carries the loudest warning in the course:
- Capacity is real. +15.86% on a name trading Rs 5M/day cannot absorb serious capital. Your own buying moves the price you are trying to buy.
- Costs are brutal and they scale with thinness. A round trip costs about 3.04 percentage points on average [LEDGER:
reference_backtest_methodology], wider on thin names. A separate optimised model that beat its benchmark by 9-16 points out of sample was blocked from shipping precisely because sub-Rs 5 fill and slippage costs "erase it several times over" [LEDGER:project_model_optimised]. - Survivorship bias hits thin names hardest. Delisted companies are missing from the panel, and companies that delist are disproportionately small and illiquid. The +15.86% is an upper bound, and of all the numbers in this Academy it is the one most inflated by that bias.
- There is still a hard floor: nothing below 10,000 average daily volume is scanned or published at all. Below that there is no price, only prints.
What to actually take away
Not "buy illiquid stocks". Take this: a rejection rule and a sizing rule are different tools, and using the first where the second belongs deletes returns you never see.
Evidence · History · Confidence · Counterargument
Evidence, THIN +15.86% vs LIQUID -1.00%; THIN+QUIET +24.85% vs IGNITED -0.27%. [LEDGER: project_quiet_loaded_edge]
History. One large study on the PSX panel.
Confidence, medium. The direction is strong and was found independently. The magnitude is not investable at face value once capacity, slippage and survivorship are applied.
Counterargument. The strongest case against acting on this: after the 3.04pp round-trip cost, the capacity ceiling and the survivorship correction, the realisable edge may be small or zero. A related model with a bigger measured edge was blocked for exactly this reason. Treat this as a sizing insight, never as a licence to buy illiquid stocks.