Is Algo Trading Profitable in India? What SEBI's Data Says

Algo trading can be profitable in India, but the algorithm isn’t what makes the money. SEBI’s own data shows that most of the trading profit in equity derivatives goes to proprietary desks and foreign investors, almost all of it through algorithms, while 87.7% of individual traders lost money in FY26. As far as I can tell, those institutions win because of what sits behind their algorithms: research, speed, scale and lower costs per trade. A computer pressing the button is the least important part.

I build automation software for a living, so I’d rather start with the uncomfortable numbers than with a pitch.

Last updated: November 2026.

What SEBI’s studies actually show

SEBI has now published several studies of profit and loss in equity futures and options. The share of individual traders who lost money, after costs:

Period	Individuals who lost money	Source
FY22	89%	SEBI study, January 2023
FY22 to FY24, combined	93% of over 1 crore traders, average loss about ₹2 lakh each	SEBI PR 22/2024
FY24	91.1%	SEBI study, August 2026
FY25	90.9%	SEBI study, August 2026
FY26	87.7%	SEBI study, August 2026

FY26 looks like an improvement, and SEBI itself warns against reading too much into it: the number of active traders fell by around a fifth that year, so a lower loss rate doesn’t necessarily mean outcomes got better.

Now the other side of the trade. SEBI’s September 2024 release found that in FY24, proprietary traders and foreign portfolio investors booked gross trading profits of about ₹33,000 crore and ₹28,000 crore, while individuals and others lost over ₹61,000 crore, all before transaction costs. Then the line that matters for this post: 97% of FPI profits and 96% of proprietary trader profits came from algorithmic trading.

The August 2026 study repeated the exercise for FY25 and FY26 and the pattern got stronger. In FY26, proprietary traders made about ₹44,000 crore in gross trading profit and FPIs about ₹14,000 crore, and 99% of both groups’ profit came from “algo entities”, meaning entities that placed at least one algorithmic order during the year. SEBI also spells out the mechanics plainly: “Derivatives trading is largely a zero-sum activity before transaction costs, so one participant’s loss broadly corresponds to another’s gain.”

So yes, algorithms make money in Indian markets. Just not for most of the people who’d search for this article.

The algorithm isn’t the edge

It’s tempting to read those numbers as “institutions use algos, institutions win, so I should use algos”. That gets the cause backwards.

Look at who uses algorithms and still loses. SEBI’s FY26 data shows only 15% of the 84.2 lakh individual traders were flagged as algo users, and SEBI notes even that overstates real use, because it counts clients whose intraday positions the broker squared off automatically near the close. Across FY25 and FY26, individuals lost money in aggregate whether or not they carried the algo flag.

Then look at how concentrated the winners are. In FY26 the top 10 proprietary traders, 2.3% of the category, earned ₹33,124 crore of the category’s ₹44,483 crore gross profit. That’s about 75%, and SEBI notes it roughly tracks their share of trading volume. Those firms aren’t winning because they automated a moving average crossover. Desks at that scale typically have research teams, years of clean data, servers placed close to the exchange, capital to run many uncorrelated strategies at once, and a far lower cost per unit traded than a retail account.

An edge is a repeatable reason your trades should make money after costs. The algorithm is just the delivery mechanism. A retail algo that buys when RSI crosses 30 isn’t competing with other retail traders clicking buttons. It’s sharing a market with those desks.

What automation genuinely does for you

None of this makes automation useless for a retail trader. It does four things well, and they’re worth having.

It enforces the plan. The stop you wrote down gets placed. There’s no “let me give it ten more points”, no doubling up to win back the morning’s loss.

It executes the same way every time. The rules that run on a quiet Tuesday also run on budget day, without hesitating and without getting tired by the afternoon.

It makes your idea testable. Explicit rules can be replayed over years of history with costs deducted. “I trade on feel” can’t be backtested, which means it can’t be honestly evaluated either.

And it gives you your day back. A strategy that runs without you watching every candle is easier to stick with for the months it takes to know whether it works.

What it can’t do

It can’t create an edge. A strategy with no edge, automated perfectly, loses money with excellent discipline.

It can’t notice that the market has changed. Your algo will keep trading the rules that suited last year’s volatility long after the conditions behind them are gone. Deciding when a strategy is broken is still your job.

It can’t fix bad sizing. If you risk 5% a trade, automation just gets you to the drawdown faster.

And it can’t beat costs on a tiny edge. This one deserves its own section, because it’s where most retail algos actually die.

The cost arithmetic that sinks scalping algos

Start with Nifty futures. Buy and sell one lot (65 units) at about 23,780, the same price in and out, at Zerodha’s published rates as of September 2026 (the brokerage calculator uses the same figures):

• STT at 0.05% of the sell value: ₹773
• Brokerage, ₹20 each way: ₹40
• Exchange transaction charges: ₹57
• Stamp duty: ₹31
• GST and SEBI fees: ₹21

That’s about ₹921 for a trade that made nothing. Divide by 65 and the strategy needs about 14 Nifty points per round trip just to break even, before any slippage. A futures scalper aiming for 10 points loses money on its winning trades.

Options look cheaper per trade, so here’s a hypothetical options scalper. Every number below is an assumption, not a result:

Assumption	Value
Instrument	One lot of an at-the-money Nifty weekly option at ₹150
Target and stop	+4 points and −4 points
Win rate	55%
Trades	10 a day, 20 days a month
Charges per round trip (Zerodha rates)	about ₹70
Slippage	half a point each side, so ₹65 per round trip

Gross, the strategy has an edge. It wins 0.55 × 4 − 0.45 × 4 = 0.4 points per trade on average, or ₹26 a lot. Over 200 trades a month that’s about ₹5,200.

Net, it’s a different strategy. Each trade earns ₹26 and pays ₹70 in charges and ₹65 in slippage, so it loses about ₹109. Over a month that’s roughly −₹21,800. The edge was real. It just wasn’t bigger than the costs.

To break even with those costs, this strategy would need to win about 76% of its trades on a symmetric 4-point target and stop. Very few strategies of any kind do that reliably.

SEBI’s numbers show the same effect across the whole market. Individuals paid about ₹25,000 crore in transaction costs in each of FY25 and FY26. Before costs, 82.1% of individual traders lost money in FY26; after costs, 87.7% did. Costs alone turned a slice of small winners into losers.

Lower frequency changes this arithmetic. A futures strategy targeting 300 Nifty points pays the same ₹921, which is under 5% of its target rather than more than its entire profit. That isn’t a promise that slower strategies work. It just means the costs stop deciding the outcome for you.

A checklist for a retail algo process that has a chance

I can’t tell you what returns to expect, and nobody honestly can. What I can describe is what a sound process looks like. If your setup doesn’t tick most of these, the safest assumption is that it will join the majority in SEBI’s tables.

1. You can state the edge in one sentence, and say who is likely on the other side of your trades and why they’d accept the loss.
2. The backtest includes brokerage, STT, exchange charges, GST, stamp duty and a slippage assumption. The strategy still makes money with costs doubled.
3. Average profit per trade is several times the round-trip cost, not a fraction of it.
4. There are at least 100 trades in the backtest, spread across different market conditions.
5. Shifting each parameter by 10–20% doesn’t wipe out the result, and a held-back period you never optimised on still works. Our post on overfitting explains why this matters more than the headline return.
6. Position size comes from the most you’re willing to lose, with a daily loss cap and a kill switch set before going live.
7. You paper traded until live behaviour matched the backtest’s expectations.
8. You went live at minimum size and kept a log of expected versus actual fills.
9. You wrote down, in advance, the drawdown at which you’ll switch it off and review.
10. You judge results over months, after costs and tax, not over a good week.

Points 2 and 5 are where tooling helps most. We build INDfolio, so weigh this accordingly: our backtests include the full Indian cost stack, and the platform adds Monte Carlo and walk-forward checks for the “is this real?” question. Any tool that models costs honestly and lets you test out-of-sample will do the same job. If someone sells you a bot with a monthly return figure instead of a process like this one, read our guide to spotting AI trading bot scams first.

If you’re just starting, our 30-day plan for beginners puts this checklist into an order you can follow.

Frequently asked questions

Is algo trading profitable in India? It can be, but SEBI’s data shows that most individual F&O traders lose money and most of the profit goes to institutional desks that trade through algorithms and have better research and lower costs. Automation helps you execute an edge; it doesn’t supply one.

What percentage of retail algo traders make money in India? SEBI doesn’t publish a clean figure for retail algo users, partly because its algo flag also catches broker auto square-offs. Its data does show individual traders lost money in aggregate in FY25 and FY26 whether or not they were flagged as algo users.

Can algo trading give consistent monthly returns? No strategy, automated or manual, guarantees a monthly return, and anyone promising one is a red flag. Sound algo strategies still have losing months and drawdowns that a backtest can estimate but not prevent.

Is algo trading better than manual trading? It’s better at execution: it follows the plan without hesitating, and its rules can be tested on history. It isn’t better at finding an edge, so a weak strategy does no better automated than by hand.

Why do proprietary traders and FPIs make money in F&O? SEBI’s studies show their profits are concentrated in large, algorithm-driven firms, with the top 10 proprietary traders earning about 75% of the category’s FY26 gross profit. In my view, scale, research, infrastructure and low cost per trade explain far more of that than automation alone.