1-minute history from the monthly zips in the project's data/xauusd folder, stored as 1m / 5m / 15m / 1h / 1d candles. Gold is not part of the NSE watchlist or the live feed.

Three simple rules that survived every test: our backtest plus five research agents (candle patterns, news, AI, sessions, high win rate), about 750,000 variants, rules chosen on 2015-2021 and tested on 2022-2026, on 1-minute candles, with costs. Run together, because they do not move together.

The Parabolic SAR on 5m to daily charts at 9 settings, traded three ways (always in, intraday flips, follow it from the open), long / short / both, with and without a trend filter: 1,269 variations, chosen on 2015-2021 and tested on 2022-2026, compared with random entry times, and tried as a filter on the plan's rules. Takes about a minute.

Every strategy family and setting replayed on every day: the zone rules at gold's session opens, Fibonacci retracements (tested against non-Fibonacci levels), opening-range breakouts and previous-day high / low. Settings are picked on 2015-2021 and judged on 2022 onwards, which the pick never saw; then an AI learns which strategy to trade each day. P&L per 1.00 lot after costs. Takes a few minutes.

Runs Renko strategies (this tab's rules) on MetaTrader 5 through the YonexisBridge EA, one row per strategy. When the next brick would be a signal, a BUY STOP / SELL STOP waits at that brick's level with its stop attached, so it fills at the brick level as in the backtest; the stop then follows the exit rule. Orders follow the mode in System › MT5 connection (dry run / demo / live). Defaults are the settings that held up in the MT5 backtests (gold $6 bricks, SuperTrend 10/4 + Brick Pulse, both sides, exit at the first opposite brick; crude 0.2, SuperTrend flips, short only).

Entries on a fresh signal in the trade's direction: SuperTrend (ATR period × multiplier on the bricks, source close, wicks taken into account) plus the chosen filters - Street Trend (close vs 10-brick EMA), Brick Pulse (streak ≥ 2, ≥ 50% of the last 50 bricks with the trade), Trend Health (≥ 80% of the last 10). Fills at the brick level (a live stop order at that level), after spread + slippage and swap. MT5 backtests (Aug-Nov 2026 1-minute, Oct 2025-Nov 2026 5-minute): Buy / Sell on every SuperTrend flip was about break-even (PF 1.0-1.1) with deep drawdowns; exiting at the first opposite brick beat fixed brick targets.

How a trade's result is worked out: an entry fills at the close of the signal brick and an exit at the close of the exit brick (CandleForge mode: the 1-minute close that formed them). With the first-opposite-brick exit a run of n bricks in your direction makes n − 3 bricks: the turn itself costs a brick to confirm the entry and two to confirm the exit (1 brick = −2, 2 = −1, 3 = 0, 4 = +1 ...). Move = exit − entry in your direction, costs = spread + slippage (+ swap overnight); per 0.01 lot = (move − costs) × 1 oz (crude × 10 bbl).

Replays the Renko strategies on the MT5 1-minute history with every combination you tick - several brick sizes at once, YONEXIS or CandleForge indicators, entry rules, sides, exits - after spread and slippage. The earlier part of the dates is where a setting would be chosen ("train"); the later part is never used for choosing ("test") and shows what that choice did next. The AI score ranks what tends to survive: the weaker of train and test, discounted for few trades and for settings whose neighbouring brick sizes lose.

Test your zone rules on any day or date range. Nothing is written to the paper wallet: this is a pure replay to see the P&L.

Levels: Same = the MT5 symbol has the same price as on NSE (zone levels used as they are). Offset = a CFD or other contract priced differently: levels are shifted by the MT5 − Upstox gap at the moment of the order. Test buy / close work in Demo mode only (minimum volume).

What this automation's rules call today on today's Upstox 5-minute candles, lot by lot, whether it is switched on or not (no order is sent). Open lots are valued at the live price; closed lots are after costs. Refreshes every 10 s.

Several automations can run on one website, each with its own symbols / watchlist and settings. A symbol can be in only one running automation per website (the website keeps one position per symbol); on another website the same symbol is fine.

Candles and prices always come from Upstox. Look up runs your recorded symbols flow for every row: it types the search text (ANGELONE for ANGELONE-FUT), clicks the result that is the same contract (futures by Upstox's near-month expiry, never options) and saves the website's name as the web symbol - on all workers in parallel. Live orders and prepare then use that name as {symbol}. (A symbols flow that reads a whole list instead: Look up, then Suggest.)

One row per lot: each lot has its own target and shares the entry's stop. P&L = points × qty × exchange lot size (Upstox). Exit price = the live Upstox price when YONEXIS sent the exit.

Leverage: money blocked for a stock trade = value ÷ leverage. Futures margin: % of contract value blocked. Every change is kept in the ledger.

Quick: one pass over the period with the strategy's settings — results in seconds.

Backtests every strategy in the library on all stored history, ranks them, finds the best one per stock and trains the AI picker that chooses which strategy to trust in which situation.

Compare a converted strategy with the admin's own backtest trades, trade by trade.

Where live prices and candles come from. Switching restarts the live engine on the new source.

Live NSE feed, candles and (locked) orders. Logins expire daily at 3:30 AM.

Needed for real backtests and for training the AI score model. 5m data back to 2024 for 20 stocks takes about 2–4 minutes (Upstox serves one month per request).

Make sure backtests run on complete, consistent candles: see what is loaded, choose what to verify, find gaps, then refetch only what is missing.

Rebuilds each futures symbol's history from the contract that was the near month at the time (expired contracts, needs Upstox Plus), instead of the current contract's thin far-month past. Prices are as traded; at each expiry the series rolls to the next contract.

Export candles already in your database as CSV (one file) or Excel (one sheet per symbol and interval). Click a row in the table below to pick it.

Phase 4 gate: live paper results must stay inside the backtest's expected band.

Promotion path: candidate → approved → paper → live small → live. Only strategies in paper status make live calls.

Learns which of a strategy's signals tend to work, to score live calls.

New models start in shadow; promote to champion to use them live.

Grouped by remark tag. Add remarks when you review calls on the Live tab.