A "yantra" is an instrument. So is this.

Why we built this

Retail traders in India are not short on tools. They're short on tools that tell the truth about themselves. Most signal services show you the win streak and quietly bury the losing stretch. Most "AI-powered" trading apps treat the model as a black box and ask you to trust the output on faith.

Trading Yantra started from a simpler premise: build the kind of analysis tool we'd actually want to trade with ourselves — one where every rule, pattern and model signal is backed by a measurement we can point to, and where a negative result is reported as clearly as a positive one.

How we approach the problem

We combine three layers instead of leaning on any single one:

• A technical indicator engine — RSI, EMA, ADX, ATR, Bollinger Bands and Williams %R computed daily and screened into buy/sell signals.
• A candlestick & price-action engine — twelve candlestick detectors plus a point-in-time context model (trend, pullback, support/resistance, volume, relative strength, volatility and market regime) that scores every pattern in context, benchmarked against an unconditional baseline.
• AI/ML models — trained on historical NSE/BSE price action to surface signals that classic rule-based indicators can miss.

None of these three layers is presented as infallible. Each is a hypothesis that gets tested against real, measured outcomes — and the product reflects what the testing actually shows, not what would be easiest to market.

What "radical honesty" means in practice

Concretely: when our research shows that a popular chart-pattern breakout underperforms a random baseline, that finding ships in the product's own documentation instead of being quietly dropped. When a market regime looks unfavorable, the product can shrink or veto position sizing rather than firing a signal blind to the broader trend. We'd rather be useful than impressive.