Two of eight sessions made the entire profit — and the short side lost money
The question that set this study going was blunt: why had a strategy that had been making money suddenly started handing it back? Strota's intraday signal engine was running live, its performance curve had turned down, and the obvious guess was that something had broken. The audit found the premise was wrong. Nothing broke. The engine had only ever earned on a very small number of days, and once those days stopped arriving, what remained underneath was a steady bleed.
Across the sample the idealised book still finished green — +59.5%, or ₹5,945 on a stake of ₹10k per trade. Every rupee of it came from two sessions. The single strongest day returned +88.8%, more than the entire book made. Take that day out and the other 7 sessions add up to −29%. Take out both winning days and the remainder is −55%. Concentration like that is not a detail; it is the result.
More telling was the order the losses arrived in. The three most recent sessions in the sample lost −11.2%, then −7.0%, then −24.8% — consecutive and accelerating. That is the signature of a structure designed to harvest trend days: when the market hands it a strong one-directional session it collects heavily, and when no such session prints it pays out in small repeated stops. Earlier work had flagged that dependency several times. This was the first occasion it was measured on the engine's own live signals rather than inferred from history.
Method matters for a finding this stark. The study took eight end-of-day snapshots of the engine's signals and resolved them into 430 distinct trades once duplicates were stripped out. Every trade was sized equally, and each was charged 0.30% for a round trip — that is everything you pay to get in and back out again, brokerage, statutory charges and the spread. Note what this ledger is not: it ignores the position caps the live account actually applies, so the rupee amounts here run larger than the live ones. The shape of the result transfers; the magnitudes do not.
Two leaks accounted for most of the damage. Shorts — trades that profit when a price drops — were the larger and steadier of them. There were 107, each losing −0.112% on average, together costing −12.0% of the book, about ₹1,200 at the stake tested. Only 42% ended as net winners. For Strota this was the seventh separate occasion the short side had come back negative, counting earlier studies of borrow constraints and of which stocks are realistically shortable at all.
Leak two was narrower and sharper: buying longs in stocks that had already run. On the worst session, six positions were stopped out for a full unit of planned risk each — a −1R loss meaning the trade gave up exactly the amount it had been set up to risk — and every one of those names was already up somewhere between 5% and 9% on the day when it was bought. Buying strength into a tape that kept snapping back was, that day, the whole story.
Both candidate fixes were then re-tested over the identical eight sessions. That is a weaker test than fresh data, and worth saying so, but it holds everything else constant. Removing the short side was the clean one: 323 trades instead of 430, average return per trade rising from +0.138% to +0.221%, and the total from +59.5% to +71.4%. Nothing was surrendered for it. The worst session stayed exactly as bad, because that damage was done on the long side.
Capping extension behaves quite differently, and this is the part worth sitting with. Skipping fresh longs in names already up more than 5% on the day left 280 trades and a total of +34.8%. A tighter version at 3% left 166 trades and +17.5%, and it converted the worst session from −24.8% into −3.7%, with two other red days turning green. It also amputated the upside: the best session's +88.8% shrank to +29. Over a window this trend-heavy the cap lowers total return outright. It is a drawdown lever — drawdown being how far the account falls from its high before recovering — rather than a profit lever, and it was adopted on those terms, at the looser 5% setting.
Both gates are now live, and they apply only when a new position is opened. A fresh short is blocked outright. A fresh long is skipped when the stock has already climbed past the extension threshold at the moment the signal fires. Anything already open is left alone and exits on the rules it was opened under, so no holding gets force-closed by a rule change. Nine gate scenarios were unit-tested before the change went out, and either gate can be switched back off.
Neither gate creates an edge, and the study is explicit about that. Even with both applied, ordinary non-trend days still finish red — three of the eight sessions stay negative after the fix. Eight sessions is a thin sample, and five trading days inside that window are missing from it altogether. The accounting is idealised, so live results will not reproduce these percentages. Most of all, the headline profit rests on one session, which is precisely the fragility the whole exercise set out to name.
Where the work actually points is a change of direction rather than a repair. This engine was built to buy strength, and the tape it has been trading has been rewarding the opposite. A separate Strota study of buying gap-downs of 5% or more inside an uptrend is the one setup that has survived both fees and slippage in testing so far — a measurement over a specific past window, not a standing opportunity. The intention is to repeat this session-by-session audit monthly, because the day a genuine trend session returns, the cost of capping extension will show up plainly in the numbers.
More money stories
- Crypto's momentum formula loses money on Indian stocks. A six-month version doesn't.
- Low-volatility stocks, not momentum, produced a book positive in every completed year
- We tried to make a momentum strategy win every year. The momentum had to go.
- Fading morning gap-ups paid in liquid stocks and lost in illiquid ones
- Widening the stock pool, not the signal, gave this hedged low-volatility test a positive decade
- A short-volatility test on Nifty returned 11.3% a year, inside a 15% drawdown