Most people who want to start algo trading in India begin in the wrong place. They open YouTube, watch somebody's Python screen recording, copy a moving-average crossover, run it, and lose money for reasons they cannot diagnose — because they skipped the market, the costs and the testing to get to the code.
This is the order that actually works, with an honest note at each step about what it costs and how long it takes. Nothing here requires a finance degree. It does require doing the steps in sequence.
First, what algo trading in India actually means
Algorithmic trading means your entry, exit and position-sizing decisions are written as explicit rules, and a program executes them by connecting to your broker over an API. That is the whole idea. The rules can be simple.
It does not mean high-frequency trading, it does not mean machine learning, and it does not mean a black box someone sells you. It also is not a licence: as a retail trader placing your own orders through your own broker account, you do not need SEBI registration to automate your trading.
What it does mean, in practice, is that everything becomes measurable. You will know your win rate, your average loss, your worst drawdown and your total costs, because the program records all of it. Most discretionary traders never learn these numbers about themselves.
Step 1: Learn the market before you automate it
This is the step people skip and it is the one that costs the most. You cannot write rules for a game whose mechanics you do not know.
The minimum is genuinely a minimum: what a share is, how an order reaches the exchange, the difference between market, limit, SL and SL-M orders, what margin and leverage do to a position, and how you get taxed on it. If any of those is fuzzy, start with the fundamentals rather than with Python — we wrote a separate walkthrough for exactly this.
Time: three to six weeks of evenings. Cost: zero, if you use free material and are disciplined about finishing it.
Step 2: Open a broker account with API access
Your program has to talk to something. In India only a subset of brokers offer a documented API that retail clients can actually use, and among those, Zerodha's Kite Connect is the best documented with the most mature Python client. Angel One's SmartAPI, Dhan, Fyers and Upstox are the usual alternatives.
Two costs to be aware of, because they surprise people. The trading account itself is free to open. The API subscription is separate and billed monthly by the broker — we broke the current charges down in detail elsewhere, and it is worth reading before you budget.
You need an account before any code runs
Kite Connect sits on top of a Zerodha account. Free to open·Aadhaar OTP, done in minutes
Step 3: Learn enough Python, which is less than you think
You do not need to become a software engineer. A working trading system uses a small slice of the language: variables, lists and dictionaries, conditions, loops, functions, reading and writing files, calling an API, and handling errors so a network blip does not kill your program mid-session.
Add two libraries — pandas for handling price data and requests for HTTP — plus your broker's client library, and that is the toolkit.
Time: four to eight weeks at five hours a week for a complete beginner; a weekend for anyone who already codes. Cost: zero. Python and everything around it is free.
Step 4: Write the strategy as rules, not as a feeling
Take whatever you currently do and force it into unambiguous statements. Not "buy when it looks strong" but: buy when the 20-period average crosses above the 50-period average, on the daily close, in this list of instruments, risking 1% of capital, with a stop at the previous swing low, exiting on the reverse cross or a 3% loss, whichever comes first.
Writing this down honestly is uncomfortable, and it is the most valuable hour in the whole process. Most people discover at this point that they do not have a strategy; they have a set of habits that changed depending on how the last trade went.
Also decide the boring parameters now: maximum positions open at once, maximum loss per day after which the system stops, and what happens if it restarts mid-session holding a position.
Step 5: Backtest it, honestly
Run the rules over several years of historical data and see what would have happened. Anyone can do this; doing it without fooling yourself is the skill.
The three things that turn a backtest into fiction: acting on the same candle whose close you used to decide (look-ahead bias), leaving out brokerage, STT and slippage, and tuning parameters until the curve looks good on the exact data you tuned it on (overfitting). Each of these can turn a losing strategy into a beautiful equity chart.
What you are looking for is not the highest return. It is the maximum drawdown — the worst peak-to-trough fall — because that is the number that decides whether you will still be running the system a year from now.
Step 6: Paper trade, then go live small
A backtest tells you about the past. Paper trading tells you whether your code actually works: whether the API token refreshes, whether orders get rejected for margin, whether your loop survives a disconnection at 11:30.
Run it on live data with simulated fills for at least a month. Then go live with the smallest size the instrument allows, and keep it there until the live results and the paper results stop diverging. If they never converge, the difference is usually slippage or an assumption in your fill logic.
Time: one to two months. Skipping this step is how people lose real money to a bug rather than to the market.
Step 7: Deploy it so it runs without you
A strategy that only runs when your laptop is open and the wifi holds is not automated. Production means a small cloud server that starts your program every morning, restarts it if it crashes, and logs everything it did.
For Indian brokers there is one extra piece: a static IP. Brokers ask you to register the IP your orders come from, which is part of how the exchange-level algo framework tracks automated order flow. Cloud providers offer static IPs for a few hundred rupees a month.
Cost: roughly Rs 500 to Rs 1,500 a month for a small VPS with a static IP, plus your broker's API subscription.
What SEBI's rules mean for you
Short version for a retail trader automating their own account through their own broker: you can do it. The regulatory framework that came in for retail algo trading places most of the obligations on brokers and algo providers — registering strategies above a certain order rate, tagging orders, and running the approval process.
Your practical obligations are to register with your broker for API use, to place orders from a registered static IP, and to stay within the order-per-second limits your broker sets. If you ever sell or share your strategy with others, the picture changes substantially and you should read the rules properly.
Capital, time and the mistakes that cost most
On capital: you need none to learn, because backtesting and paper trading are free. To go live meaningfully, most people start somewhere between Rs 50,000 and Rs 2 lakh — small enough that a bug is a lesson rather than a disaster, large enough that fixed costs are not eating the returns. Never start with money you need.
On time: six to nine months from zero to a small live system, working evenings and weekends, is a realistic and unexciting estimate. Anyone promising four weeks is selling something.
The four mistakes we see repeatedly: automating before understanding the market; backtesting without costs; increasing size after a good month rather than according to a rule; and running a strategy nobody can explain, which means nobody can tell whether it has stopped working or is merely having a bad quarter.
If you would rather not assemble this yourself
Everything above can be self-taught. Plenty of people do it, and free material in India has got genuinely good. The reason a structured course exists is that the steps interlock badly when you learn them separately — the Python tutorial does not know about STT, the market course does not know about API tokens, and the gap between them is where most attempts die.
Our Algorithmic Trading with Python course at Rs 24,900 runs the same seven steps in order across 50+ modules, ending with a deployed system on a cloud VPS with a static IP. One payment, lifetime access, taught in a Hindi-English mix. There is a free demo on WhatsApp first.
We do not sell tips or calls, do not manage money, and make no claim about returns. Trading carries a real risk of loss.
Disclosure: the account-opening link on this page is under Atul Shrivastava's Zerodha Authorised Person registration (NSE AP Reg: AP2516003481; Zerodha Broking Ltd. SEBI Reg: INZ000031633) and earns a revenue share. TheFinBaba is not a SEBI-registered Investment Adviser — this content is educational, not investment advice.
Frequently Asked Questions
Do I need SEBI registration to start algo trading in India?
Not as a retail trader automating your own account through your own broker. You register for API access with your broker and place orders from a registered static IP. Registration requirements apply if you sell, distribute or manage strategies for other people.
How much money do I need to start algo trading?
Nothing to learn, because backtesting and paper trading cost nothing. To go live, most people start between Rs 50,000 and Rs 2 lakh - small enough that a mistake is survivable, large enough that monthly API and server costs are not eating everything.
Do I need to know coding for algo trading?
You need a small amount of Python: variables, lists, conditions, loops, functions and API calls. That is a few weeks of evenings for a complete beginner. You do not need a computer science background.
Which broker is best for algo trading in India?
Zerodha's Kite Connect is the most widely used because of its documentation and mature Python client. Angel One SmartAPI, Dhan, Fyers and Upstox are common alternatives. Compare the monthly API charge and the rate limits before choosing.
How long does it take to go from zero to a live trading system?
Six to nine months of evenings and weekends is realistic for a beginner - roughly a month on market basics, two on Python, one or two on strategy and backtesting, and one or two on paper trading before going live small.
Is algo trading profitable?
Automation removes hesitation and record-keeping problems; it does not create an edge. A losing strategy automated simply loses faster. Profitability depends on your strategy, your costs and your risk control, and losses are a real possibility.
Related Reading
- How to learn the stock market from scratch
- Kite Connect API in Python - step-by-step tutorial
- Zerodha API charges, broken down
- Backtesting a trading strategy in Python
- SEBI's algo trading rules for retail traders
- Running a trading bot 24x7
- Algorithmic Trading with Python - full syllabus
- All course fees, stated plainly
Disclaimer: TheFinBaba provides educational content only - this is not investment advice. Trading involves risk of loss.
