I’m an ML Engineer. I got tired of "AI Trading Bot" scams, so I coded my own Cash Swing Trading Engine in public. (No advice, just math)

Hey everyone,

Disclaimer: I am not a financial advisor or a SEBI registered entity. I am an Machine Learning engineer working in industry for more than 8+ years.

Like most developers, I wanted my savings to work for me, but the retail trading space is full of traps:

Overfitted AI Bots: Feeding raw stock prices into LSTM/deep learning models just memorizes noise. They look perfect in backtests, but blow up live.
Lagging Indicators: 1970s charts (RSI/MACD) are lagging averages that get front-run by HFT servers at microsecond speeds.
Shitty performance of current Mutual Funds : 90% of the stocks has given negative return in last 2 years. I was tired of handing over commission to them. Hence decided to try something to build
So, After more than 8 months of struggle, I built StockMind—an automated quant engine focused on Cash Equity Swing Trading.

🚫 Why F&O (Futures & Options) is Excluded

SEBI states that 90% of retail F&O traders lose money. The math is structurally rigged against us:

Theta (Time Decay): Options decay to zero on expiry. In cash equity, you have holding power to wait out drawdown cycles.
Leverage trap: 5x margin leverage means a small 2% market dip triggers a force liquidation at the absolute bottom.
Friction drag: Frequent options trades bleed up to 10-15% of your capital annually in STT, GST, and brokerage fees.

📊 The Math & Modeling Implemented

Instead of predicting price ticks, the engine uses structural probability:

Sector Rotation: Ranks sector indices daily using Relative Strength (RS) against the Nifty 50. All buy signals in weak sectors (bottom 30%) are automatically blocked.
Quality + Momentum Screen: Excludes high-debt companies (Debt-to-Equity > 1.5) and prioritizes high return (ROE > 15%), filtering only macro uptrends (Price > 150-day SMA).
**Friction-Adjusted Backtests: **Adds a flat 0.25% cost per trade to simulate real-world STT and bid-ask slippage.

📈 Metrics & Performance

Backtested metrics over the last 5 years (adjusted for 0.25% cost per trade):

CAGR: [24.5%] | Max DD: [-12.3%] | Sharpe: [1.65]
Here is my current live paper-trading performance. I have started trading with real money as well. Will share the result soon in next post.

Paper Trade performance.

⚠️ Current Weaknesses

Manual Login: Have to log in manually to Zerodha Kite every morning to generate the API session token (no free automated headless login).
Daily timeframes only: Runs end-of-day data for swing trading. Not designed for day traders. Its more for medium to long term perspective.
💻 Open Source & Dashboard

Live App: (https://www.thestockmind.com)
Telegram Channel: https://t.me/stockmindAI
Would love to get feedback from other developers and quants on the slippage modeling and sector rotation index.