How the audit works

Every report answers one question about the trades you upload: are you making money after charges, and is that steady enough that chance is an unlikely explanation. This page says how each part is measured and what the audit cannot see. The numbers on it are read from the same code that writes your report.

Reading your file

• A broker tradebook (one row per fill) is matched buy-to-sell, oldest first, per instrument, into round trips. A quantity still open at the end has no exit, so it is left out and named.
• A trade list or TradingView export is read as given, one row per trade.
• Every row that is left out is counted on the report's What was read line with the reason, and so is the date range read. Exact repeats are counted and named, never dropped silently: two exports that overlap would otherwise count a trade twice.
• Up to 5 files can be read as one book, one per account.

What counts as one result

Trades that share the same market days are not independent: a bad day hits all of them together. So the audit counts results, not trades. A same-day trade belongs to its trading day, and all of that day's trades are one result. Positions held overnight are grouped into blocks of calendar days long enough to cover a typical hold, and each block is one result. The report says which unit it used.

The verdict

• Steadiness is the average result divided by how much results swing, scaled by how many there are. Around 1 could easily be luck; 2 or more is strong evidence. It is measured on your results as they are and with the market's own move taken out (below), and the lower of the two is the one that counts.
• Strong evidence: steadiness of 2 or more per result and trade by trade, and the last 12 months positive. Moderate: steadiness over 1.5 per result, ahead on average, and the last 12 months positive. Losing, not bad luck: behind on average with steadiness of -2 or worse. Weak: anything else, with the first number that failed named.
• Too little data: under 20 results, no evidence grade is given on any part of the report. The rupee figures and the fixes are still shown.
• The market's own move. Each trade's result is also measured minus the NIFTY move over its hold, signed for its side, so a book that only rode the market does not read as skill. MCX trades are never adjusted.

Is it holding up

• The last 12 months are graded on their own. Older results are context; a strategy that worked two years ago and not since says so.
• The held-out test: give the date you stopped changing the strategy, and the results after it are read apart from the ones before. The report says "Held up" or "Not confirmed" only with at least 10 results after that date.
• Two days: the result is recomputed without the best and the worst day, and in two halves, to show whether a few days carry it.

Charges

• The statutory charges are the ones the exchange and the government levy: STT, exchange transaction charges, SEBI fees, stamp duty and GST at 18 %. Rates as of 2026-10 (STT per Budget 2026, eff 1-Apr-2026).
• Brokerage defaults to ₹20 per order; you can change it on the upload page.
• Slippage, when the file has no real fills, defaults to Cash intraday 0.045 %, Cash delivery 0.045 %, Futures 0.06 %, Options 0.15 %, MCX commodity futures 0.0023 % per round trip. A broker tradebook carries real fill prices, so no slippage is added. When the file carries its own charges, they replace the model.
• Every report shows the result at no charges, at the charges above, and at twice those charges, so you can see how much the verdict depends on them. When a file has no quantities, charges are a share of an assumed trade size and the report says so.

Capital and options

• Margin per trade is modelled as a share of turnover: Cash delivery 100 %, Cash intraday (MIS) 20 %, Stock futures 20 %, Index futures 11.5 %, MCX futures 10 %. The peak of the margin in use at the same time is the least capital that would have carried the book, unless you give your own.
• Option legs opened together on the same underlying and expiry are one position; legs added more than 30 minutes later are adjustments. A position's return is on an assumed margin (11.5 % of strike times quantity for an index, 20 % for a stock), on the maximum loss for a hedged pair, and on the premium paid for a bought option.

The fixes and the what-ifs

• When the report flags a slice of your trades (an hour, a weekday, a name, a size) as dragging the book, the slice is checked against skipping the same number of trades at random. A slice that does no worse than a random skip is not called a leak.
• Try a change recomputes your trades under a rule you pick. It is hindsight: any rule picked after seeing the trades looks better than it will be. The report reads the change against the held-out part and the random skip, and the real test is next month's trades with the rule written down first.
• Every "next step" is a measurement or a change to your process. The audit describes past trades; it never says what to buy or sell.

Not checked

• Whether the rules used any information that was not available at the moment of the trade (look-ahead). Only you can confirm this.
• Whether these exact trades were chosen after trying many variants. A strategy picked as the best of 50 needs a far higher bar.
• Fills: limit orders that would not have filled, circuit limits, and impact on thin names are not modelled beyond the slippage line.
• Margin, leverage and position sizing: returns are per trade on its own turnover.
• Options: option positions are measured on an assumed margin, not the margin your broker actually charged, and the extra STT on options exercised at expiry is not charged.

Questions about a number on your report: write to us. If it is our error, the refund policy covers it.