Posted in My Daily Journal · 31 October 2026
We asked our own trade history where the profit actually comes from, whether it repeats at the same time of year or around the same events, and tested seasonality five different ways.
Sahil says option buying is about the days when a trade "explodes". Where did our explosions come from in the last three years?
From very few days. On the weekly ITM-150 pivot buyer, the top twenty days made 157% of the three-year profit — the other 513 exit days netted a loss. Losses were many and small (the bottom twenty days were only 20% of all losses). A burst day looks like this: the position was already held overnight (median two nights), yesterday was a wide-range day (85% of bursts versus 52% of all days), today gapped at least 0.5% (60% vs 18%) in the position's direction (80%), and the day then moved at least 1%. Crash days were the opposite: entered the same day, gap against the position, Friday, and reversed within the session.
Do the bursts come at the same time of year or around the same events?
Partly, and mostly not. Across 2020–2026 the forty-five biggest days made 114% of all profit. March had eleven of them, June eight, April seven; September none. Only fourteen of the forty-five fell within two days of a nameable event (COVID, the 2024 election week, the August 2024 yen-carry crash, the April 2025 tariff crash); 64% of the burst money came from ordinary multi-day trends with no headline. And the position was always there for the same reason — a plain pivot cross taken a median of three days earlier. The bot does not predict events; it is already in the trade when they arrive. Twenty-seven of the forty-five entries were against the previous five-day trend.
If February–March and October–December pay and April, July and August don't, can we trade bigger in the good months?
We ran five tests before answering. Leave-one-year-out passed (6 of 7 years). A permutation test barely passed (p = 0.04). Out of sample on 2015–2019 data the calendar was simply absent — October and November were the worst months then. BankNifty agreed weakly. A mechanism test (does NIFTY trend more cleanly in those months?) failed. Three of five: the pattern is real inside 2020–26 and not before it. It is the footprint of this regime's events, not a law of the calendar, so lot size will not follow the month. A reminder of the historical month-by-month record now sits on our dashboard, labelled exactly that way.
Anything else fall out of the seasonality work?
The most important finding of the week. Replaying the same pivot signals on 2015–2019 index data, the structure earned almost nothing — about ₹78k over five years versus ₹24L for 2020–26 on the identical engine. This is a high-volatility-regime system: 2015–19 had roughly half as many ±1% days. Every pass bar we had used lived inside 2020–26. We now treat realised volatility as a standing input to how much to size, and the quiet years as the scenario to plan for.
What about option writing instead of buying — has anyone tested our signals as a seller?
We had tested intraday writing in July (the 9:20 straddle passed twelve years on one data set and showed no edge on another). New today: selling the opposite option on our pivot signals with the same index trail. Overnight, it produced the smoothest equity curve in the whole programme — ₹8.6L, maximum drawdown ₹76k, every year positive — but at ₹1.5L of margin per lot, with the opposite risk shape to buying, and 0.79 correlated with the buyer book. Intraday-only writing earned less. Our charter is buying only, so this is recorded and not traded.