Is Algo Trading Profitable in India? The Math, SEBI's Data, and What a Bot Taught Me

"Is algo trading profitable" is rising on Google Trends in India, usually typed by someone who has just seen an ad for a ₹999 "AI trading bot". I build and run trading bots on Zerodha's Kite Connect API, so here's the honest answer: a bot doesn't create an edge. It only executes one, if you have it, without emotion, and it loses money faster if you don't.

This post is about how to find out which case you're in before it costs you.

Not investment advice. I'm not sharing my strategies or P&L.

What SEBI's data says

SEBI has studied retail traders twice in recent years:

Intraday equity (FY23): about 7 in 10 individual intraday traders in the cash segment lost money. Loss-makers placed more trades than profit-makers, and their trading costs were 57% of their losses. Among traders under 30, 76% lost. (Business Standard)

F&O (FY22–FY24): about 93% of over 1 crore individual F&O traders lost money, around ₹2 lakh each on average, and ₹1.81 lakh crore in total. Only about 1% made more than ₹1 lakh after costs over three years. (Outlook Money)

The cost line matters most for algo traders. Bots trade more often than people. If costs were already over half the losses for manual traders, a bot that trades ten times a day makes that worse unless each trade has a real edge.

The break-even win rate

Every strategy has a win rate it needs just to break even, after charges:

break-even win rate = net loss per losing trade ÷ (net win + net loss)

On Zerodha, a ₹2,00,000 intraday (MIS) round trip costs about ₹118 in brokerage, STT, exchange fees, stamp duty and GST. (The full breakdown is in my Kite Connect bot guide.)

Small targets with wide stops look great in a backtest because they win most days. They need a win rate almost nobody sustains. One bad morning wipes out weeks.

Before you write any strategy, put your target, stop and position size into this formula. If the answer is above about 70%, be very sceptical.

Charges are a fixed tax on every trade

A few things people miss:

Brokerage is capped at ₹20 per order, so it's cheap on large orders. STT (0.025% on intraday sells) and stamp duty (0.003% on buys) scale with size and never cap.

GST (18%) applies on top of brokerage and exchange charges.

Slippage isn't on the contract note, but it's real. A market exit in a fast tick can cost more than all the charges combined. In my first live trade, a market sell at the target filled ₹2.16 a share below it, which cost about 60% of the planned profit.

Kite Connect costs ₹500/month for market data, plus a server with a static IP (₹400–1,000/month). That's ₹1,000–1,500 a month before your first trade.

At 10 trades a day and ₹118 each, charges come to about ₹26,000 a month. Your edge has to beat that before you earn anything.

Why backtests make algo trading look profitable

Things I learned running replays and paper bots for weeks:

1. Fills are optimistic. A backtest fills at the candle's price. A real stop-loss in a falling stock fills lower. A real limit order at your target sometimes doesn't fill at all.

2. Timing is fragile. On one recorded day, shifting re-entries by less than a second turned a profitable rule into a losing one. If a strategy is that sensitive, it isn't an edge.

3. Results follow the market's mood. In a year-long backtest of a simple "fade the top movers" rule over the NIFTY 500, shorts won for three months and lost for five; longs did the reverse. The rule had no edge. It just rode whichever way the market was going that month. (Full write-up: NIFTY 500 top losers screener.)

4. Backtests and paper trading disagree. Some rules that lost in backtests did fine on paper, and the other way round. Live fills are the final judge. Paper trading is the closest cheap substitute.

5. Simple rules often break even before charges. That's the trap: a rule with zero edge looks "almost working", but charges turn it into a steady loss.

So when is algo trading profitable?

When all of these are true:

You have a rule with a measured edge after charges and realistic slippage

It holds across different market months, not just one trend

Your win rate and payoff are comfortably past break-even

It survives at least a few weeks of paper trading on live prices

Position size is small enough that a bad week doesn't end you

What a bot does give you, even without a big edge:

Discipline. It takes the stop-loss every time. Most manual traders don't.

No missed trades. It's at the screen at 10:00 AM even when you're in a meeting.

Records. Every price, order and fill is logged, so you learn from data instead of memory.

Those are real benefits. They just aren't profit by themselves.

Ignore these "profitable algo" claims

"90% accuracy" bots: see the break-even table. A 90% win rate with a bad payoff loses money.

Telegram tip bots and copy trading promising fixed monthly returns. SEBI has acted against unregistered advisers repeatedly.

Backtest screenshots without charges and slippage. Ask for the contract notes.

"AI trading bot India free": if it were reliably profitable, nobody would give it away.

A four-week test for your own algo

1. Week 0: write the break-even maths. Put charges and a slippage estimate in the code.

2. Weeks 1–4: paper trade on live prices, recording every tick so you can replay changes.

3. Each week: compare paper fills with what the order book showed. Note the gap.

4. After four weeks: if net P&L after charges is positive and not from one lucky day, go live with the smallest size Zerodha allows. Watch the gap between live and paper.

5. If it's negative: change one thing, replay the recorded days, and paper trade again. Don't go live to "see if it's different".

Intraday profit is speculative business income, taxed at your slab rate, and you may need a tax audit at high turnover. Talk to a CA before you scale.

The honest answer

Algo trading in India is profitable for a small minority who have a measured edge, size positions sensibly, and treat charges as a design constraint. For everyone else, a bot is a faster way to join SEBI's 70%.

Building your own? Start with how to build an intraday bot on Kite Connect, or get in touch if you want one built and paper-tested on your own rules.