Help and how to read the reports

Quick start: read a report in five steps

Verdict first. The coloured card states the conclusion and the evidence gates (PASS / WARN / FAIL / PENDING). Do not read charts before knowing whether the strategy passed.
Pick the segment. Back-test = the development window. Validation = a sealed later period. Forward = live paper run with no real orders. Only validation and forward are out-of-sample.
Pick a period. Presets, a year, a month or custom dates. Everything recomputes.
Check net, not gross. The Costs section shows what remains after slippage, brokerage and tax.
Look for fragility. Robustness table, parameter explorer and the top-10-trades share tell you whether one lucky trade carried the result.

How to change the reporting period

Segment tabs: Click Back-test, Validation or Forward paper run. A lock means the period is sealed (not yet opened).
Presets: All (whole segment), YTD (from 1 January of the latest year), 1Y / 3Y / 5Y (counted back from the last date).
Year: Choose a calendar year, for example 2020 to see the COVID crash and recovery on its own.
Month: Choose one month, e.g. 2020-03. Annualised numbers show n/a because the window is under a year.
From / To: Type or pick exact dates. The window is clipped to the data in the segment.

Example. Select Year → 2018 on Meridian. Period return is negative and the heat-map shows which months hurt. Select Year → 2021 and the same strategy shows a large gain. Same rules, very different years, which is why a single headline CAGR is not enough.

Headline numbers

Number	What it tells you	Example (Meridian, whole back-test)
Period return	Total change in equity over the window, after modelled costs and slippage.	₹500,000 grew to ₹2,123,542 (325%).
CAGR	The steady yearly growth rate that would give the same total. Needs one year or more.	16.0% a year. Compare with the index CAGR shown beneath it.
Max drawdown	Biggest fall from a peak to a later low. This is the pain number: plan for a repeat or worse.	-18.6%: on ₹5 lakh that is a fall of roughly ₹93k from a peak.
Sharpe / Sortino	Return per unit of risk (rf = 0). Sortino only counts downside swings.	Sharpe 1.52. Above 1 is good; below 0.5 is weak. Sharpe is flattered when only ~37% of capital is invested.
Calmar	CAGR divided by the size of the worst fall.	Higher means more return for each unit of drawdown.
Volatility	How much the daily equity bounces, annualised.	Compare with the index volatility under it.
Alpha / beta	Beta: how much the strategy moves with the index. Alpha: return left after removing that. Needs 12+ months.	Beta 0.2 to 0.3 means it takes about a quarter of the market's swings.
Avg invested	Share of capital in shares on average; the rest is cash.	About 37%: low exposure explains the low drawdown.
Best / worst month, day	Extremes, to judge tail risk.	A worst day near -6% shows even a 'safe' book can drop sharply.

Charts

Growth of 100 vs index. Both lines start at 100 on the first day of the window. A line above the index is outperformance. The dotted line is the strategy before costs (approximate): the gap between it and the solid line is what trading costs took.
Drawdown. Always at or below zero. Long flat bottoms mean long recoveries. The longest underwater stretch is in the headline cards.
Calendar-year returns. Look for consistency. One huge year with others flat means the result is fragile.
Capital invested. How exposure moved with the market regime.
Monthly heat-map. Green positive, red negative; darker is bigger. The right-hand column is the year total.

Trade statistics and the trade log

Win rate: Share of trades that made money: 38% for Meridian (464 trades).
Average win / loss: 37.6% / -9.3%. A low win rate is fine if wins are much bigger than losses (payoff ratio).
Profit factor: Gross profit ÷ gross loss: 2.11. Above 1 means profitable.
Expectancy: Average return per trade. Multiply by number of trades to see why costs matter.
R multiple: Result divided by the initial risk (entry to first stop). -1R is a normal stop-out; +5R is a big winner.
Top-10 share: 63% of Meridian's profit came from 10 trades. High numbers mean luck can matter. The best trade was GVT&D: 428% (₹153,784), entered 2023-09-01 and exited 2024-09-02.

The trade log lists every closed trade. Click a column header to sort, type a symbol to filter (for example the best trade above), and use Prev / Next. Trades are filtered to your chosen period by exit date.

Anatomy of a strategy page: every section explained

Sections appear in this order on each strategy page. Examples use Meridian.

Trade analysis

What it is. A summary of the closed trades inside your chosen period: how often it wins, how big wins and losses are, and where profit came from. Four panels.

Trade statistics. Trades closed, win rate, average win, average loss, payoff ratio (average win ÷ average loss), profit factor (total profit of winners ÷ total loss of losers), expectancy (average return per trade), average R, net P&L, average and median holding days, best and worst trade, and the top-10 share.
Exit reasons. Why trades ended: STOP (price hit the protective stop), REBAL (sold at a month-end rebalance or trimmed), DELIST (stock stopped trading; sold at a penalty). The bar shows net P&L by reason.
Distribution of trade returns. A histogram in 5% steps from -30% to +60% (anything beyond is placed in the end bars). A long right tail with many small left bars is the signature of trend-following.
Top and bottom 10 trades. The ten biggest winners and losers with symbol, exit date, return and rupee P&L.

Example. Meridian wins 38% of 464 trades but its average win (37.6%) is far larger than its average loss (-9.3%), giving profit factor 2.11. Watch out for: a top-10 share above ~50% (63% here): remove those trades and the result changes materially. Also compare the average loss with the stop distance; much larger losses mean gap-downs through the stop.

Trade log

What it is. One row per closed trade with Symbol, Entry, Exit, Days, Return, P&L, R, Exit reason. Return and P&L are net of slippage and charges. A partial sell (trimming a big winner) is its own row.

How to use. Click a header to sort (click again to reverse). Type in Search to filter by symbol or reason, for example STOP or a ticker. Prev / Next pages through 50 rows at a time. The list follows your period selection by exit date.

Example. Sort by P&L descending to see the winners behind the top-10 share; sort by Days to see how long winners are held versus losers (winners are typically held longer, losers cut quickly). Watch out for: clusters of exits on the same date (a market-wide shock) and repeated entries in the same stock (re-entry after a stop).

Robustness and scenarios

What it is. A set of re-runs of the full simulation with one thing changed at a time, to test whether the result survives. Columns: CAGR, Max drawdown, Sharpe, Index CAGR (index over the same dates) and Trades.

Costs ×2. Doubles brokerage and statutory charges.
Start +1, +3, +5 months. Shifts the start date; a robust strategy gives similar answers.
From 2018 / From 2021. Later windows with better data coverage and no early years.
Top 10 / Top 20, Lookback 252d, Skip 21d. Neighbouring parameter choices.
Slippage 0.5%, 1%, 2% per side. Re-runs the strategy assuming each buy fills that much above, and each sell that much below, the reference price (plus the usual market-impact term, capped at 3%). It shows how much execution quality the idea can absorb before the edge vanishes. Realistic liquid large-caps are far below 0.5%; the higher levels stress illiquid names and bad fills.
Random-rank null. Picks stocks at random, keeping the same stops, regime and costs, 30 times. If the real ranking does not beat these, the signal is not adding anything.
Alpha versus the index. A monthly regression on the benchmark: alpha (extra return not explained by market moves), beta, and a t-statistic that allows for correlated months (HAC). As a rule of thumb a t-statistic above 2 is interesting, above 3 convincing.

How to read. Look at the spread. If the CAGR stays within a couple of points across rows the idea is robust; if one row collapses, that is the fragile assumption. Example. Meridian ranges roughly 14% to 18% across the neighbouring-parameter rows and still beats the random-rank null by about 5 points; the slippage rows show how fast the edge erodes with worse fills: 14.1% at 0.5% per side, 14.0% at 1%, but only 5.6% at 2%, below the index (13.3%). In words: the idea tolerates about 1% per side of slippage, not 2%.

Parameter explorer

What it is. Four drop-downs (names held, lookback, skip, cost multiple) that look up a pre-computed full simulation for that combination and show CAGR, drawdown, Sharpe, trades and end equity next to the base case, plus an overlay chart of growth (100 = start).

How to use. Change one drop-down at a time. The base case (15 names, 182-day lookback, 7-day skip, costs ×1) is the dashed line. What to look for. A smooth landscape: neighbouring settings give neighbouring results. A sharp spike at one setting means the base case was luck. Note. Only the listed values exist; the simulator is not re-run in your browser, and the grid covers the development window only.

Rules and parameters

What it is. The complete, frozen specification: the ordered rules (universe, signal, regime, entry, exit, portfolio risk, cash and costs) and a table of every parameter value. Anything not listed here is not part of the strategy.

How to read. Check that each parameter has a plain reason and that there are not too many: more parameters mean more room to overfit. The Instrument box states the market and the capital used. Example. "ATR stop 2.5×, trail 5.0×" means a stop 2.5 average daily ranges below the buy price that then follows the highest close at 5 ranges. If a strategy changes version, the version number and date change; old results are never silently replaced.

Known limitations

What it is. The list of things that make the results less reliable than they look. It is part of the result, not a footnote.

Typical items and what they mean.

Survivorship / missing members. Companies that left the index and have no price history are missing, which makes older years look better than reality.
Adjustment for splits and bonuses. Detected by price-ratio patterns, not a corporate-action table, so a few errors can remain.
Price-only benchmark. Ignoring dividends understates the index by about 1% a year, and the strategy too.
Tax. Most trades are short-term; tax is not deducted.
Estimated inputs. Cash yield by RBI repo schedule and the cost schedule are estimates until checked against contract notes.

How to use. Mentally adjust the headline in the direction each limitation pushes it (most push results down when corrected).

Logs and raw data

What it is. Plain CSV files behind every number, so anyone can recompute independently. Links appear once a period is loaded.

equity_*.csv: date, equity, invested, bench: end-of-day equity in rupees, share of equity in stocks, and the Nifty 500 price index level. Suffix _backtest, _validation or _forward.
fills_*.csv, audit_*.json, signals_ledger.jsonl, signals_log/: Execution audit files (weekly strategies). fills lists every transaction with its signal date, signal rank and score, execution date, price, slippage, fees and reason. audit holds the checks shown in the Execution audit section. For forward runs, signals_log holds one immutable signal file per decision week and signals_ledger.jsonl chains their SHA-256 hashes; the page can re-hash them in your browser. Trade logs of these strategies also carry signal_entry, entry_rank, entry_price, signal_exit, exit_price, exit_rank.
trades_*.csv: symbol, entry, exit, ret, pnl, reason, days, R, gross, slip, fees, buy_value, sell_value. ret is the net return of the trade, pnl net rupees, gross price P&L before costs, slip slippage rupees, fees brokerage, taxes and charges, with pnl = gross - slip - fees.
scenarios.csv: The robustness table. grid.csv and grid_curves.csv: the parameter explorer results and monthly growth curves. decisions.csv: the market regime at each month-end decision.
positions.csv, orders_next.csv, signals_latest.csv, events.log: Forward paper run only: open positions, the orders planned for the next open, the current top-30 ranking and the run log.

Example. Open trades_backtest.csv in a spreadsheet, sum gross, slip and fees and you reproduce the Costs section.

Other parts of the page

Header and badge: Codename, version, previous name and status: PAPER (forward paper run active), CANDIDATE, REJECTED, NO EDGE, PARKED.
Verdict card: One-sentence conclusion plus the evidence bullets. Its colour: green good, amber caution, red not supported.
Promotion gates: PASS / WARN / FAIL / PENDING checklist (see below).
Explain the strategy: Four reading levels of the same description.
Segment note: The grey line under the tabs states what the selected segment is, its dates, and whether it is sealed, opened or live.
Validation tab: The sealed hold-out. Once opened it shows normal results for that window; it is spent and cannot be re-run with different settings.
Forward paper run: Live prices from 2026-10-19, simulated nightly. It also shows the Paper trading book: open positions with stop levels, the orders planned for the next open, the current top-30 ranking, and the run log.
Cost model and slippage selectors: Re-price the same trades under another broker or slippage level; see the Costs section.

Costs: gross vs net and cost per transaction

For Meridian's back-test: gross profit ₹1,355,792; slippage ₹58,691; brokerage, taxes and fees ₹64,278; net profit ₹1,232,823. Costs took 9% of the gross profit.

Gross vs net. Gross is price movement before costs. Net is what an account would have kept.
Cost per transaction. Total slippage and charges divided by the number of orders (a trade has a buy and a sell). Basis points (bp) express the cost as a share of traded value: 100 bp = 1%.
Broker model. Choose Zerodha, Dhan, Kotak Neo or Anand Rathi in the dropdown. The page re-prices every trade in your window using that rate card. Delivery brokerage differs widely (₹0 vs 0.20% of value) while statutory charges (STT, exchange, SEBI, stamp, GST) are the same for everyone.
Caveats. Rates are as published on 2026-10-18 and are applied to all years. Starred (*) models use an assumed DP charge. See rate cards and sources.

How to change the strategy parameters

Parameters cannot be recomputed live in your browser because each run is a full simulation of hundreds of stocks. Instead, a grid of combinations was computed in advance with the full simulator. In Parameter explorer choose the number of names (10 to 30), lookback (182 or 252 days), skip (7 or 21 days) and cost multiple (×1 or ×2). The cards show CAGR, drawdown, Sharpe, trades and end equity for that choice next to the base case, and the chart overlays the two growth curves.

Example. Set Costs × to 2: the end equity falls a little but the strategy still beats the index, which says the result is not fragile to costs. Set Names held to 10 and the drawdown changes: concentration changes risk. If a small change in a setting flips the result, distrust the strategy.

The Robustness and scenarios table adds start-date shifts and sub-periods; the random-rank null asks whether the stock ranking adds anything beyond the stops and regime rules.

Explain levels: novice, amateur, expert, pro

Each strategy has four explanations of the same strategy. Novice: plain language, no jargon. Amateur: rules and numbers. Expert: formulas, parameters and what the evidence says. Pro: risks, biases, capacity, tax, audit path. Click the level tabs under Explain the strategy; the report below does not change.

Verdict and gates

Gate	Question it answers
Beats random-rank null	Does the signal add anything beyond the stops and regime filter? (30 random-ranking runs)
Robust to costs and start dates	Does the result survive double costs and shifted starts?
Stable across sub-periods	Does it work in both halves of history, not just one?
Not driven by a few trades	Is the top-10 share of profit reasonable?
Data completeness	Are the prices of all historical index members available? (survivorship)
Sealed hold-out	Does it hold on data never used in design? Opened once.
Forward paper run	Does it work live, before real money?

PASS WARN FAIL PENDING A strategy only moves towards real money if gates pass; failures stay published.

FAQ

Why is the validation tab locked?: The hold-out is opened once, after the specification is frozen, so it stays an honest test.
Is paper trading profit real?: It uses real prices with modelled fills, so it excludes real execution, taxes and behaviour.
Why do numbers differ from another site?: Price-only vs total-return indices, different costs, different universe (point-in-time here), different start dates.
Can I download the data?: Yes: the CSV links at the bottom of each strategy page (equity, trades, scenarios, grid).
Is this advice?: No. Research and education only. Hypothetical and back-tested results have inherent limitations and do not guarantee future results. Nothing here is investment advice, a recommendation or a solicitation. The publisher is not presented as a SEBI-registered investment adviser or research analyst.

How to read these reports