TRADER

Autonomous Trading System

A research-backed trading system for Indian equity markets. We tested 6 strategies, found one that works, then spent 4 iterations making it smarter — not by changing the signal, but by improving how capital is deployed.

Try It Live (Paper Trading)

Backtest Performance

Oct 2020 – Jan 2025 (4.3 years)

8.6%

CAGR

+42% total return | 60% win rate

Including 2025–26 bear market (5.4 years)

6.5%

CAGR

+40% total return | Max drawdown 9–16%

₹1,00,000 simulated capital, 96 NIFTY stocks, includes costs and slippage. The system stayed profitable through the 2025–26 bear market while keeping drawdown 3–4x lower than the broader market. Not investment advice.

The Core Idea

Markets overreact. When a stock drops sharply, fear pushes it below its fair value. Value buyers step in, and the stock bounces back within days. We exploit this effect systematically across 96 NIFTY stocks.

1. Read the market

Classify as Bull, Neutral, or Weak. Adjust how much capital to deploy.

2. Find oversold stocks

Rank all stocks by how much they fell. Buy the biggest losers — they bounce hardest.

3. Hold 5 days, repeat

Wait for the bounce. Sell after 5 trading days. Pick new losers. Every week.

System Evolution — Same Signal, Smarter Allocation

v1Foundation

Validated the core edge: stocks that fall hard tend to bounce. Sat in 100% cash during weak markets — safe but left money on the table.

Baseline

v2Capital Efficiency

Stopped sitting idle. Started deploying capital even in weak markets — because the reversal signal is actually strongest when fear is highest.

+44% better

vs v1 baseline

v3Multi-Engine

Added midcap stocks as a second engine. Allocation scales with market momentum — more midcap exposure when trends are strong, less when fading.

+53% better

vs v1 baseline

v4Adaptive IntelligenceCurrent

Replaced hard thresholds with smooth confidence scoring. System continuously adjusts exposure based on signal quality and portfolio health. Protects during drawdowns, leans in during recoveries.

+53% better

vs v1 baseline

Every improvement came from better capital allocation — the underlying signal never changed.

Research Journey — 6 Strategies Tested

ML Prediction (5-min)

No signal in OHLCV features

Breakout Detection

Fakeouts, no follow-through

Mean Reversion (5-min)

Too weak after costs

Trend Following (30-min)

No intraday trend persistence

Cross-Sectional ML

Zero predictive signal

Daily Reversal

Structural mean-reversion effect

Key Discovery

Intraday trading

Doesn't work for retail in Indian markets. Every 5-minute strategy we tested lost money — institutions and algorithms are too fast. The edge doesn't exist at this resolution.

Daily reversal (5-day holding)

Works consistently. Stocks that drop hard attract buyers and bounce within a week. This is a behavioral effect — driven by human psychology, not patterns that algorithms can arbitrage away.

Where the Edge Is Strongest

We tested the same reversal signal across two different market segments. Midcap stocks overreact more — and bounce harder.

Large-Cap (NIFTY 50)

48 stocks
Return (5.4 yrs)+38%
Sharpe 0.70Max DD 9%

Midcap (NIFTY 100 Extra)

48 stocks
Return (5.4 yrs)+108%
Sharpe 1.06Max DD 16%

Same signal, different universe. Midcap stocks have stronger overreaction — producing 2.8x higher returns.

Capital Efficiency & Risk

Returns Per Rupee Deployed

Our system uses only ~52% of capital on average — the rest stays in cash as protection. Despite deploying half the capital, we generate competitive returns.

Our System (52% deployed)+40% return
InvestedCash buffer

1.4x

more efficient per rupee than full deployment

Worst-Case Drawdown

How much can you lose at the worst point? Our system keeps drawdowns 3–4x lower than typical equity investments.

Our System-12%
Large-Cap Fund-35%
Mid-Cap Fund-45%
Active Traders (avg)-65%

Lower drawdown = less panic, fewer bad decisions, better sleep.

How We Protect Capital

Regime-Aware Sizing

Automatically reduces exposure in weak markets and increases during strong trends. Continuous adjustment, not binary on/off.

Drawdown Protection

When the portfolio drops, the system automatically reduces position sizes. Gentle in uptrends, aggressive in downtrends. Recovers faster after bounces.

Kill Switch

If recent trades show a losing streak, the system pauses automatically. Resumes only when signal quality recovers. No manual intervention needed.

Live Market Status

Historical Replay — See the System in Action

Replay: How the System Behaved
2023 — Strong Bull Market

NIFTY rallied throughout the year. The reversal strategy excelled — buying beaten-down stocks that consistently bounced in the uptrend. 75% win rate with minimal drawdowns.

P&L

+₹49,647

Win Rate

75%

Trades

44

Max DD

9.6%

IC

+0.055

Trade Timeline

Jan 9ENTRYHDFCBANK, RELIANCE, SBIN

Reversal picks — 3 stocks fell 3-5% last week, market healthy

NIFTY +0.8%

Jan 16EXITHDFCBANK, RELIANCE, SBIN

5-day hold complete — all 3 bounced

+₹1,840

NIFTY +1.2%

Jan 16ENTRYINFY, TITAN, BAJFINANCE

New reversal picks — IT and consumer names pulled back

NIFTY +1.2%

Jan 23EXITINFY, TITAN, BAJFINANCE

5-day hold — strong bounce on TITAN (+4.2%)

+₹2,100

NIFTY +0.6%

Mar 6ENTRYICICIBANK, AXISBANK, LT

Banks pulled back on RBI policy — buying the dip

NIFTY +0.3%

Mar 13EXITICICIBANK, AXISBANK, LT

Banks recovered sharply — all 3 profitable

+₹3,200

NIFTY +1.5%

Jun 5ENTRYNESTLEIND, HINDUNILVR, ASIANPAINT

FMCG sector correction — reversal signal strong

NIFTY +0.2%

Jun 12EXITNESTLEIND, HINDUNILVR, ASIANPAINT

Clean bounce — win rate at 78% for the year

+₹1,950

NIFTY +0.9%

Sep 18ENTRYTATASTEEL, JSWSTEEL, HINDALCO

Metals corrected on China concerns — buying oversold

NIFTY -0.4%

Sep 25EXITTATASTEEL, JSWSTEEL, HINDALCO

Metals recovered — system continues winning

+₹2,800

NIFTY +1.1%

Dec 18ENTRYTCS, WIPRO, TECHM

Year-end IT selloff — system buying the dip

NIFTY +0.1%

Dec 25EXITTCS, WIPRO, TECHM

Year closes strong — +49.6% total return

+₹1,647

NIFTY +0.7%

Try the Dashboard (Paper Trading)

No account needed. Simulated trading with virtual capital using real market data.

Built with Python, FastAPI, Next.js, Zerodha Kite Connect, Neon Postgres