A multi-agent ensemble reads the market. One decision comes out, or none.
VYUHA watches Indian markets (NSE and MCX) and writes down what it would do, using pretend money, before anyone risks a rupee. A vyuha is a battle formation: many parts, one purpose.
How a call gets made
Four stages, always in this order. Most of the time the answer is "stay out", and that is recorded too.
Read the prices
Fresh candles from the broker data feed on 5 and 15 minute bars. Stale or missing data turns the decision into no-trade.
The agents vote
Trend, market regime, statistics, options positioning and others each give a lean. Together they produce a BUY, a SELL, or no trade.
Safeguards decide
Cost, daily count, cooldowns and loss limits can block a call or shrink it. The reason is written next to the decision.
Record and resolve
Every decision is logged. Each call is followed until it hits its target, its stop, or its time limit, then costs are subtracted.
Thirteen safeguards, checked every morning
The rules that limit damage come before the cleverness. A morning check confirms each one is still switched on and unchanged.
| Safeguard | What it does | Limit |
|---|---|---|
| Cost gate | Blocks a call when trading costs eat too much of the expected move | over 35% |
| Daily call cap | Limits calls per instrument per day | 6 |
| One call at a time | No second call while one is open on the same instrument | 1 |
| Cooldown after a stop-out | Waits before the next call | 45 min |
| Pause after stop-outs | Stops the instrument for the day | 2 stop-outs |
| Smaller size after a stop-out | Cuts the next call's size | half |
| Daily and weekly loss limits | Halt trading when losses reach the line | 2% / 6% |
Six more cover open risk, combined losses across instruments, a manual pause-everything switch, slow decisions, and poor data.
What this site is, and is not
VYUHA is a project in private testing. Please read this part.
What it is
- A research project that practises on live market data with pretend money.
- A system built to say "no trade" far more often than it trades.
- Open about its limits: every number is shown after costs.
What it is not
- Not investment advice, a tip service, or a research recommendation.
- Not placing any real order, for anyone.
- Not publishing results. Past or paper performance would say nothing about the future.
Where it is heading
The next stage is more weeks of paper results, then a small closed group of testers. Nothing opens to the public until the legal and regulatory work is done. Contact details will appear here when it does.