Algo Trading
Build systematic trading strategies for NSE and BSE. Backtest, optimize, deploy.
Foundations
Algorithmic trading — commonly called algo trading — is the practice of executing trade orders automatically using pre-programmed instructions based on defined criteria such as time, price, volume, or mathematical models derived from market data. Instead of a human deciding when to buy and when to sell, a computer program evaluates market conditions against a set of rules and fires orders directly into the exchange in milliseconds, without hesitation or emotion.
In the Indian context, SEBI governs API-based trading under its algorithmic trading circular. The key regulatory boundary for retail traders is this: you may use algos to trade your own account freely. However, if you manage funds on behalf of others or use an algo to generate trade signals for fee-paying clients, you must hold a SEBI-registered Investment Adviser (IA) or Research Analyst (RA) licence. Since 2018, the proliferation of broker APIs — particularly Zerodha's Kite Connect — has triggered an explosion in retail algo participation on NSE, turning what was once an institutional exclusive into an accessible discipline.
The distinction between retail and institutional algorithmic trading is important to understand. Institutional algos — at hedge funds, proprietary trading desks, and market makers — are often latency-sensitive High Frequency Trading (HFT) systems that compete on microsecond execution, co-location facilities, and raw technological infrastructure. Retail algos, by contrast, are strategy-based: they follow a defined set of rules to decide when to enter and exit positions over minutes, hours, or days. Retail traders actually hold a structural advantage that institutions cannot replicate: small position sizes. Because a retail trader's position is tiny relative to daily volume, they can enter and exit positions at or near the market price without moving the market against themselves.
The real value proposition of algo trading for a retail trader is discipline. A well-designed algorithm executes your strategy exactly as written, every single time, without second-guessing at 2:45 PM on a volatile Thursday. It removes the cognitive biases — loss aversion, recency bias, FOMO — that consistently erode discretionary trader performance. It allows you to backtest a strategy across years of historical data before risking a single rupee, and once deployed, it can monitor markets continuously while you focus on other work. For many retail participants, this represents a genuine and durable edge over purely discretionary trading.
Architecture
Provides real-time tick data, OHLCV bars, and order-book depth from NSE/BSE. Can be streamed via broker WebSocket APIs or third-party data vendors. Data quality and latency here directly determines signal accuracy.
Computes technical indicators (moving averages, RSI, ATR, VWAP) and evaluates conditions against each incoming tick or bar close. Emits a signal — buy, sell, or hold — to the strategy logic layer.
Applies higher-level rules: entry conditions, exit conditions, filters, and position limits. This is where your edge lives — the strategy logic enforces exactly when and how to act on a signal.
The bridge between your code and the exchange. Zerodha Kite Connect, Dhan API, Fyers API, and Angel Smart API all expose REST endpoints and WebSocket streams for order placement, modification, and cancellation.
Translates strategy instructions into actual exchange orders — specifying order type (market, limit, SL-M), product type (MIS, CNC, NRML), quantity, and price. Good execution management includes retry logic and rejection handling.
Tracks open positions in real time, calculates mark-to-market P&L, applies stop-loss and target logic, and enforces portfolio-level exposure limits to prevent runaway risk.
Records every order, fill, rejection, and signal event with timestamps to a database or log file. Critical for performance analysis, debugging, and maintaining an audit trail for tax and compliance purposes.
Strategy Library
A fast-period moving average crosses above (bullish) or below (bearish) a slow-period moving average to generate buy and sell signals.
An ATR-adaptive trailing stop that flips from bullish to bearish based on price deviating beyond a multiple of the Average True Range.
Price deviates above or below the Volume Weighted Average Price (VWAP) by a defined threshold and is expected to snap back.
The high and low formed in the first 15–30 minutes of the trading session define a range. A breakout of this range with rising volume triggers an entry.
Rank the NSE universe by 3–6 month price return (excluding the most recent month). Go long the top decile and review monthly.
Validation
Backtesting is the process of applying a trading strategy to historical market data to evaluate how it would have performed. It is the single most important tool in the systematic trader's arsenal — because it allows you to stress-test a strategy across years of market conditions, including bear markets, high-volatility regimes, and choppy sideways phases, before risking any real capital.
However, backtesting done poorly produces dangerously misleading results. The most common pitfall is overfitting — tuning too many parameters to fit historical data so precisely that the strategy captures noise rather than signal. An overfitted strategy may show exceptional historical returns and then fail immediately in live markets. As a rule of thumb, use fewer than five optimizable parameters for any strategy, and always validate that the logic makes economic sense.
Survivorship bias is another serious error: if you test a strategy only on stocks that are currently listed on NSE, you are implicitly excluding hundreds of companies that were delisted, merged, or went bankrupt over your test period. Your results will be optimistic by design. Use a full historical constituent list from a point-in-time database.
Lookahead bias occurs when your code inadvertently uses future information that would not have been available at the time of the trade — for example, using the day's closing price to decide whether to enter at the open of the same day. This is especially common with OHLCV bar calculations. Use a strict timestamp discipline in your backtesting framework.
Cover at least one full market cycle — a bull phase and a bear phase. Indian markets saw major drawdowns in 2020, 2022, and 2015–2016. Your strategy must have a defined response to each.
Include brokerage (₹20 flat or 0.03%), STT (0.025% on equity delivery sell, 0.0125% on intraday), exchange fees, GST on brokerage, and SEBI turnover charges. These compound significantly over hundreds of trades.
Assume you cannot always fill at the exact signal price. Model a 1–3 tick adverse slippage on entry and exit, especially for liquid futures. For small-cap equities, use a percentage-of-spread slippage model.
Split your data into rolling in-sample windows for optimization and out-of-sample windows for validation. Consistent performance across multiple out-of-sample periods is far stronger evidence than a single backtest.
Reserve the most recent 12–18 months of data and do not touch it until your strategy is fully designed. Test on this holdout period once only — treat it as a live trading simulation.
Tools
Your Path
Variables, loops, conditionals, functions, and file I/O. Then move to pandas for data manipulation and matplotlib for chart plotting. Target: 4–6 weeks of self-study with free resources like Python.org docs and Real Python.
Study how exchanges work, order types (market, limit, SL, SL-M), margin requirements, lot sizes, and basic price action. You cannot code a strategy you do not understand conceptually.
Open a demat + trading account with a broker that offers an API. Zerodha and Dhan are the best starting points. Subscribe to the API plan and read the official documentation thoroughly.
Start with a single, well-understood strategy such as EMA crossover on NIFTY 50. Use backtesting.py or vectorbt on 3 years of daily OHLCV data. Review trade-by-trade results, not just the equity curve.
Deploy the strategy in paper trading mode (most brokers support this) for at least one full calendar month. Track every signal, entry, and exit. Compare live signal results against your backtest expectations.
Go live with the minimum viable position size — 1 lot of NIFTY futures or a single lot of liquid options. The goal is validating live execution, not making money. Run for 2–3 months before sizing up.
Increase position size only after consistent live performance matches backtest expectations across at least 100+ trades. Maintain a trading journal. Review monthly. Never size beyond your capital's ability to absorb a 10-trade losing streak.
Common Questions
Yes. SEBI permits API-based algorithmic trading for your own account. You may not use algos to manage third-party funds without holding a SEBI-registered IA or RA licence. Most retail traders run algos entirely within their own demat + trading account, which is fully legal.
You can start backtesting and paper trading with zero capital. For live deployment, ₹50,000–₹1 lakh lets you trade small position sizes to validate a live strategy. Meaningful monthly income typically requires ₹5–10 lakh or more depending on your strategy and risk-adjusted returns.
Python basics are sufficient for most retail algo strategies. You need to understand variables, loops, conditionals, and data manipulation using pandas and numpy. You do not need a formal computer science background — many successful retail algo traders are self-taught.
Yes. Once a strategy is deployed on a cloud server (AWS, DigitalOcean, etc.), it runs autonomously during market hours, places orders via broker API, and logs trades — with no manual intervention required. You only need to review performance periodically and handle exceptions.
Zerodha Kite Connect is the most widely used broker API with extensive documentation and a large developer community (₹2,000/month subscription). Dhan API is the most accessible for beginners with a free tier. Fyers API and Angel One Smart API are strong alternatives with good reliability.
Course in development. Join the wait list — message or call us on WhatsApp at +91 82484 88542 and we'll notify you when enrollment opens.
