The State of Retail Algo Trading in India
Foreword
Most conversations about retail trading outcomes run on opinion. We run an algo platform, so we can do better than opinion: we have twelve months of live automated deployments, a research programme that has run hundreds of thousands of backtests on real market data, and the arithmetic of statutory trading charges, which is public. (Brokerage and slippage vary by broker and execution — our cost tables deliberately use best-case values.) We built Tradetron to democratise algorithmic trading, and by that measure we have succeeded. But democratising the ability to trade is not the same as improving the outcome of trading, and we owe our users honesty about the difference.
This report is that honesty. It is built from those three sources — our platform's aggregate live data, cost arithmetic, and our research programme, most of which ended in null results. We publish the nulls too. A finding you can't falsify is marketing; we have tried to write a report, not an advertisement.
One conviction runs through it: a retail trader's biggest enemies are not the market and not their strategy. They are turnover, leverage, and time-frame — three dials the trader controls. Our business does better when you trade more. You do better, on the evidence, when you trade less and hold longer. We have decided to build for the second sentence, because a platform whose users survive is the only platform that compounds.
Four numbers that describe the whole report
What 42,661 live deployments show
We open with the platform's own ledger rather than anyone's opinion. Every live automated deployment started on Tradetron in the twelve months to August 2026 — 42,661 deployments by 16,979 users — was classified not by what its strategy calls itself but by the product type it actually traded: intraday (MIS, squared off the same day by rule) or positional (NRML/CNC, able to hold). Of the 42,661 deployments begun in the window, the 32,893 with recorded positions form the outcome analysis (11,664 intraday, 20,915 positional, 314 mixed); the 9,768 that never recorded a position are excluded. Each deployment is observed from its start date to its last recorded statistic within the window — between a few weeks and twelve months. The analysis is aggregate and anonymised — no individual account was examined — and all profit figures are gross of brokerage and charges.1
Across the full cohort, 53% of positional deployments were gross-positive over their observation period against 29% of intraday ones, with median gross P&L of +₹679 against ₹0. Deployments that mixed both products did best of all (61%). Per user the picture is similar: 45% of positional users were gross-positive against 27% of intraday users.
Two further observations keep this honest, and make it more interesting. First, the same-user test: 1,026 users ran both styles in the window. These tend to be the platform's more serious traders, and their intraday books did far better than the intraday average — 51% gross-profitable. Even so, their own positional books still edged ahead (54% profitable, roughly twice the mean gross P&L). Second, persistence: the median intraday deployment recorded just two position entries before being switched off; the median positional deployment recorded twenty. Whatever else these numbers say, intraday deployments as commonly configured are abandoned quickly; positional ones get used.
| Sensitivity cut | Intraday n / gross-positive / median | Positional n / gross-positive / median |
|---|---|---|
| All with recorded positions | 11,664 / 29% / ₹0 | 20,915 / 53% / +₹679 |
| ≥10 recorded positions | 2,872 / 58% / +₹203 | 13,628 / 63% / +₹5,796 |
| ≥30 days deployed | 9,155 / 24% / ₹0 | 13,387 / 49% / ₹0 |
| Started Aug–Nov 2025 (≥9 months observed) | 3,041 / 14% / ₹0 | 1,270 / 47% / ₹0 |
| ≥10 positions and ≥30 days | 1,579 / 56% / +₹42 | 8,348 / 59% / +₹8,440 |
| Per user, all deployments summed | 6,589 / 27% / ₹0 | 8,271 / 45% / ₹0 |
| Same 1,026 users running both styles | 51% / +₹1 | 54% / +₹1,542 |
The sensitivity cuts sharpen the reading rather than soften it. Restricted to deployments that were genuinely used (ten or more positions, thirty or more days), the win-rate gap nearly closes — 56–58% against 59–63% — which matches the same-user result. What does not close is the magnitude gap: the median actively-used intraday deployment earned ₹42–203 gross; the median actively-used positional deployment earned ₹5,796–8,440 — one to two orders of magnitude more. Much of the headline win-rate gap, meanwhile, comes from abandonment: intraday deployments are far more likely to be switched off after a trade or two at a small loss.
What this ledger does not say is that intraday cannot pay — profitable intraday users are visible in the data. Note also that gross figures flatter both styles, and flatter high-turnover intraday more, because charges accumulate with every trade. What the ledger does say is that on this platform, over these twelve months, patience was paid more often, and paid far more per deployment. The rest of this report is about the mechanics that plausibly drive that gap: costs (Chapter 2), holding periods (Chapter 3), and the return sources that reward holding at all (Chapter 4).
1 Method notes: deployments classified from their recorded positions' product type; figures are gross of brokerage, exchange charges and taxes (which weigh proportionally more on higher-turnover styles); rupee figures are not capital-normalised; users self-select their strategies, so differences describe association, not causation. Deployments with no recorded positions are excluded.
The meter is running: cost arithmetic nobody escapes
Statutory costs are computable in advance and routinely ignored; brokerage and slippage vary by broker and execution — which is why the table below deliberately uses best-case values for both. The table below itemises one round trip — buy and sell one lot of a NIFTY option carrying a ₹150 premium — deliberately at BEST-case economics: a deep-discount broker charging ₹5 per lot per side, tight quarter-rupee slippage, and the rates in force from April 2026 (lot size 65 per NSE's January-2026 revision; options STT 0.15% per Budget 2026). Even at this friendly cost stack, the visible charges come to about ₹35, and slippage adds a comparable amount.
| Cost line (1 lot NIFTY = 65, ₹150 premium) | Per round trip | Basis (Apr 2026 rates) |
|---|---|---|
| Brokerage (deep-discount, both sides) | ₹10.00 | ₹5/lot × 2 |
| STT (0.15% of sell-side premium) | ₹14.63 | ₹9,750 × 0.15% |
| Exchange transaction charge | ₹6.83 | ~0.035% × both sides |
| GST (18% on brokerage + txn + SEBI) | ₹3.03 | — |
| Stamp duty + SEBI fee | ₹0.31 | buy side |
| Visible total | ₹34.80 | |
| Slippage (~¼ rupee per unit each side, est.) | ≈ ₹32.50 | 65 units × ~₹0.25 × 2 |
| Realistic total | ≈ ₹67 |
Set that against the capital involved: the margin for one short NIFTY lot is roughly ₹1 lakh. At three round trips a day — even on this best-case cost stack — a trader must earn nearly half the margin's worth of profit every year before the first rupee of net gain. Difficult is the right word, not impossible: profitable intraday traders exist, but the arithmetic means they need a genuinely large, consistent edge where a positional trader needs a modest one. This entry fee of the time-frame is the single largest, least discussed driver of the outcome gap in Chapter 1.
The clock, not the signal
The most common objection is: "my strategy is different." So we tested one — a live, representative trend-entry option-buying strategy (a moving-average pullback entry on NIFTY, long ATM puts), replayed over 6.3 years of real minute-level market data. We then sorted every completed trade by how long it was held.
The result deserves a close look. Every intraday holding bucket lost money. All of the strategy's profit came from the trades that were held overnight or longer. The signal was not the problem — the same signal, evaluated on slower charts, performed dramatically better: the 1-minute version of the identical rule earned ₹8,094 over the full period; the 15-minute version earned ₹61,457.
We ran the same experiment on the exit side: making the exit signal faster (1-minute and 3-minute charts) produced the worst results in a 44-configuration study; the two slowest configurations produced the best. Speed was never the missing ingredient. For a retail trader, it almost never is — the edges that exist at second- and minute-horizons belong to co-located market makers, and what remains at that horizon for everyone else is noise minus costs.
The practical rule this evidence supports: enter and manage on daily-or-slower signals; touch positions intraday only to cut risk — stops, caps, and square-offs — never to find profit. Intraday is a fine place for defence; hunting profit there is not impossible, but it demands an edge big enough to clear the meter every single day. One strategy and 86 trades illustrate a mechanism — they cannot prove a universal claim, and we present them as a case study, not a census. The platform-scale census in Chapter 1 points the same way.
What actually gets paid: the four durable return sources
Our research programme spent a year trying to find edges — hundreds of thousands of backtest configurations across structures, signals, and asset classes. Nearly all of it produced nulls, and we consider the nulls a result: at the information level available to a retail trader, there is no secret signal. What survives instead are risk premia — payments the market makes for bearing a specific, nameable risk. The four below are not lucky survivors of that search: each is a premium documented in decades of academic research3; our contribution is testing whether an Indian retail trader can actually reach them. Every implementation that survived our tests operated on positional horizons.
1 · Market drift
The oldest premium: Indian equities have compounded ahead of cash over every long window. Costs to harvest it are near zero. Most derivatives traders unknowingly fight it every day they short the market's direction.
2 · Cross-sectional momentum
Owning recent relative winners over recent losers. In our tests on 200+ liquid F&O names, the top-ranked basket beat its own universe by a strong, statistically significant margin — one of the most robust results our research produced.
3 · Option premium (VRP), defined-risk
Options persistently price more volatility than is later realised. The harvest is real — and it is also where retail accounts die, because the tail is real too. On our evidence it is only worth running with defined risk (spreads, flies, hedged structures) and hard exits, never naked.
4 · Futures trend
Following medium-term trends, long or short, across diversified futures. Modest returns in normal times — and, globally, the one approach with a documented record of paying precisely in crash years. The honest catch: we tested an India-only version (NIFTY, BANKNIFTY, MCX metals and energy, USDINR; 2007–2026) and while the diversification survives (near-zero market correlation, shallow drawdowns), the crash-year payoff does not — India's futures menu lacks the safe-haven legs, notably bond futures, that pay when equities fall. The premium is real; from an India-bound account it is the hardest of the four to reach.
These are the four durable premia our research could replicate — the first three fully reachable from an Indian account, the fourth only partially — and notice what they share: all pay for patience, each has a nameable failure mode that risk management — not prediction — must handle. A portfolio drawing on all four, sized so that no single tail can end the account, is the highest-probability path this report can point a retail trader toward.
3 Momentum: Jegadeesh & Titman (1993). Equity risk premium: long-run index literature (e.g., Dimson–Marsh–Staunton). Volatility risk premium: Bakshi & Kapadia (2003). Time-series trend: Moskowitz, Ooi & Pedersen (2012).
Three popular beliefs, tested on real data
We tested four widely held convictions on up to 13 years of data across 200+ liquid F&O stocks, with corporate actions cleaned and costs modelled. One caveat applies throughout: the universe is today's F&O membership projected backwards, which flatters buy-the-fallen results and understates short-side ones — we therefore lean on relative comparisons, which reduce — though do not eliminate — the bias. We publish the results whichever way they cut.
Across 959 sharp breakdowns (a fresh 20% fall from the 52-week high), the full outcome table over the following year: 54% rallied at least 15% without ever falling another 20%; only 20% kept falling without such a bounce; 18% did both at different points, and 8% did neither. A basket that shorted every ranked downtrend on real single-stock futures lost roughly 14% a year. The falls that continue exist — but they hide among nearly three times as many paths that punish the short.
Deep fallers did rise afterwards — a median +24% over the next year — which is why the belief survives. But the boring alternative rose more: the same rupee in a plain equal-weight basket of the universe beat the washout buys by about 11 percentage points a year, in every era we measured. And where the latest results already showed profits falling, the washouts did worst of all. The comeback story is market beta wearing a costume.
The identical entry-and-exit rule, run on 1-minute, 3-minute, 5-minute and 15-minute charts over 6.3 years: the 1-minute version earned ₹8,094; the 15-minute version ₹61,457. Every step faster added whipsaw and costs, and removed nothing but profit. (Chapter 3 shows the same result from the holding-period side.)
We ran a momentum basket of leading F&O stocks on real single-stock futures — 72 monthly rolls, 2020–2026 — then added the textbook market-neutral overlay: short NIFTY futures against the whole basket, at real contract prices. Long-only earned +11.8% a year with a −28.8% worst drawdown. Fully hedged earned +4.9% with −22.8% — market correlation zero, exactly as designed. The hedge cost more than half the return to remove a fifth of the drawdown, because most of the basket's return was the market: the pure basket-minus-index spread ran about 5% a year with a t-statistic of 0.9. Before paying for market-neutrality, measure how much of your return is simply the market's. Usually, it is most of it.
The whole report in three sentences
The market pays retail traders for patience — through drift, momentum, carefully-hedged option premium, and trend — and charges them for activity, through costs that compound at exactly the speed of their trading. Nothing in our data suggests the average trader's signals are the problem; the holding period and the turnover are. Trade slower, size smaller, define the risk, and let the four things that actually get paid do the work. Every driver behind the outcomes in this report — turnover, time-frame, sizing, risk definition — is a dial the trader holds.
Method, sources & disclaimer
This report draws on three kinds of evidence: (1) an anonymised, aggregate analysis of twelve months of live automated deployments on the Tradetron platform (42,661 deployments by 16,979 users, each classified by the product type it actually traded; profit figures gross of charges; no individual account was examined); (2) trading-cost arithmetic computed from published brokerage, STT, exchange, GST and stamp schedules as in force April 2026 (NIFTY lot 65; options STT 0.15%), with slippage separately estimated and labelled; and (3) Tradetron Research backtests on real NSE market data at minute and daily resolution — corporate actions cleaned, transaction costs modelled, and, where a result could be flattered by survivorship in the universe, reported as a relative comparison against the same universe, which reduces though does not eliminate it. Configuration ledgers behind the research chapters are retained and the summarised claims are reproducible.
This report is educational research, not investment advice and not a solicitation to trade. Derivatives trading involves substantial risk of loss and is not suitable for everyone. Neutrino Trading Pvt Ltd is a technology platform, not an investment adviser; nothing here is a recommendation of any security, strategy, or product. Past performance, simulated or live, does not guarantee future results. Aggregate platform figures are anonymised and descriptive; they are not a promise of any outcome.