Short answer: Do trading bots work? As software, usually: a bot does what its code says, bugs included. As a way to make money, no independent, audited figure was found as of 5 November 2026 showing that retail bots are profitable, and the CFTC calls claims of high returns, or of returns that are guaranteed, red flags of fraud. The closest regulator data, SEBI's August 2026 study of Indian equity derivatives, shows individuals flagged as algo users with gross losses of ₹51,728 crore over FY25 and FY26, so the label alone proves little.
Check a win rate against the average win, the average loss and the fees: Win rate and expectancy calculator.
This search hides two questions. Whether the software runs as written has a checkable answer. Whether it makes money after costs depends on the rules, the market, the fees and chance, and many pages answering that question are selling a bot.
This page rates no bot, and Fincode does build trading software to order, as the last section says. Only two sources below separate algorithmic from other trading at all, and neither isolates the bots sold to retail customers, so the rest is evidence about retail active trading rather than a verdict on bots. Automated trading can lose money, backtests do not guarantee future results, and this is general education, not investment or legal advice.
What are trading bots, and what counts as working?
A trading bot is a program that reads market data and sends orders to an exchange or broker through an API under written rules, and whether it works depends on which of three questions you ask.
| Question | How to check it | What a yes proves |
|---|---|---|
| Does it do what the code says? | Tests with simulated faults, testnet or demo runs, then small live size | The software behaves as specified, and nothing about earnings |
| Did the rules make money after costs? | A dated live record net of fees, slippage, funding and subscriptions | That record, for that period and size |
| Will it keep doing so? | No test can answer this in advance | Nothing: markets change, and past results do not carry forward |
Bots can pass the first test, and how to test trading bots sets out a plan for it. The rest of this page is about the second question.
Are trading bots profitable? What SEBI's data shows
No regulator profit figure for retail bots was found as of 5 November 2026, and the nearest data, from the Securities and Exchange Board of India (SEBI), puts the algo label on both the largest gains and large losses.
SEBI's study of equity derivatives, published in August 2026, splits gross results by whether an entity placed at least one algorithmic order in the year.
| Category | Algo entities | Non-algo entities |
|---|---|---|
| Proprietary traders | 89,452 | 985 |
| Foreign portfolio investors | 44,689 | 291 |
| Corporates | 10,825 | 3,417 |
| Individuals | -51,728 | -118,396 |
Three cautions apply to this table.
- One order is enough. An entity counts as algo after a single algorithmic order, and for individuals the flag also catches broker-run auto square-offs, so SEBI says in the same study that its FY26 figure of 15% of individuals using algos overstates genuine use.
- The groups differ. Institutions bring capital, data and speed that the label does not capture: the 10 proprietary traders with the most turnover earned ₹33,124 crore, 74.5% of that category's FY26 gross profit, according to the same study.
- The scope is narrow. These are Indian equity derivatives, where 93% of individuals traded only options in FY26 and 87.7% lost money after costs (SEBI, August 2026), not crypto or retail bot products.
The fair reading is that large firms using algorithms take most of the profits in this market, not that an algorithm earns money for whoever runs one.
One working paper does look inside retail accounts. Ghosh, Li and Zhang (March 2025) use National Stock Exchange of India account data for 2012 to 2019 to compare the same investor's algo and other trades over the same months. Their 10-day market-adjusted return was -0.15% on algo trades against -0.26% on that investor's other trades in the year after adopting. The authors call that better performance; it is still a shortfall against the market. The measure as defined does not mention fees or subscriptions, although the paper says algo trading generally comes with higher costs. It also reports that unprofitable algo users are more likely to quit.
What retail trading studies measure, and what they do not
The best-known loss figures cover retail CFD, crypto and day trading, and none of them tests bots, so they set a base rate rather than a verdict.
| Source and date | What was measured | Finding | Limit for this question |
|---|---|---|---|
| ESMA, 27 March 2018 | Retail CFD accounts, from national regulators across the EU | 74-89% of accounts lost money; average loss per client €1,600 to €29,000 | A leveraged product, with no split by tool |
| BIS Bulletin 69, 20 February 2023 | Crypto app downloads in 95 economies, August 2015 to December 2022, with assumed purchases of $100 a month | In nearly all economies a majority of users probably lost money on bitcoin; the simulated median investor lost $431 of $900 | A buying simulation, not observed trading |
| Chague, De-Losso and Giovannetti, working paper, December 2019 | People who began day trading Brazilian equity futures in 2013 to 2015 and persisted for at least 300 days | 97% lost money; 0.4% earned more than a bank teller | Only the abstract was read, and it does not mention automated orders |
| Bailey, Borwein, López de Prado and Zhu, Notices of the AMS, May 2014 | The mathematics of choosing among backtests | High backtest performance is easily reached after trying a small number of configurations | Tests no live bot; explains why backtests prove little |
SEBI's August 2026 review warns that loss rates across products and countries are not directly comparable. Read together, these sources show poor retail outcomes in leveraged, crypto and day trading, and a bot trades in the same kinds of markets and pays the same kinds of costs. That is context, not proof that a given bot loses. Why backtests flatter a strategy is covered in backtesting pitfalls.
Trading bot success rate: why the number misleads
No audited success rate for trading bots was found, vendor percentages are marketing, and a win rate means little without the average loss and the costs.
Regulators treat such numbers warily. The CFTC's advisory of 25 January 2024 says scammers claim AI-created algorithms can yield 100 percent win rates. In a circular of 2 September 2022, SEBI barred Indian stock brokers offering algorithmic trading from citing an algorithm's past or expected returns, or associating with platforms that do. A SEBI adjudication order of 25 March 2026 lists, among assured-return or misleading claims on one such platform, a strategy advertised with a win rate above 70% and monthly returns above 40%.
Hypothetical example, not a setting or a result. A bot wins 70% of its trades, with an average win of 1.0%, an average loss of 3.0% and costs of 0.2% per round trip.
Net per trade = 0.70 x 1.0 - 0.30 x 3.0 - 0.2 = -0.4%. Over 100 trades of equal size that is -40% of one position's size, with a 70% success rate. Its break-even win rate is (3.0 + 0.2) / (1.0 + 3.0) = 80%.
The function below gives the break-even win rate and rejects inputs the formula cannot use (Python 3.11, run on 5 November 2026):
import math
def breakeven_win_rate(avg_win, avg_loss, cost):
# inputs in % per trade; None means no win rate covers the costs
if not (avg_win > 0 and avg_loss >= 0 and cost >= 0):
raise ValueError("need avg_win > 0, avg_loss >= 0, cost >= 0")
num, den = avg_loss + cost, avg_win + avg_loss
if not (math.isfinite(num) and math.isfinite(den)):
raise ValueError("inputs too large or not finite")
return num / den if num <= den else None
for win, loss in [(1, 3), (1, 1), (3, 1), (0.1, 3)]: # hypothetical, % per trade
rate = breakeven_win_rate(win, loss, 0.2)
print(f"win {win}%, loss {loss}%:",
"no break-even" if rate is None else f"breakeven {rate:.0%}")win 1%, loss 3%: breakeven 80%
win 1%, loss 1%: breakeven 60%
win 3%, loss 1%: breakeven 30%
win 0.1%, loss 3%: no break-evenA break-even line says nothing about whether any bot can reach it. To put your own fees and slippage in, use the crypto trading fee calculator.
Trading bot pros and cons, treated as software
A bot's advantages are operational, and each one comes with a matching way to fail.
| Property | What it gives | What it costs or risks |
|---|---|---|
| Consistency | The same rules every time, without fatigue or second thoughts | A bad rule or a bug repeats just as faithfully |
| Speed | Orders go out faster than a person can react | Wrong or duplicate orders can stack up before anyone notices |
| Hours | It can watch a market through the night, including crypto's 24-hour markets | A fault can grow with nobody watching |
| Testability | Rules can be replayed on past data and tested against faults | Tuning until a backtest looks good proves little |
| Dependencies | No manual order entry | Outages, API changes, rate limits and clock drift become your problem |
| Access | An API key lets it trade without you | A stolen key can trade your account, and one with withdrawal rights can empty it |
Limits outside the strategy, a kill switch and alerts reduce these risks without making the rules profitable, as trading bot risk management explains.
Why a bot that runs correctly can still lose money
Automation carries out rules without creating an edge, and every trade pays costs whether the rules are good or bad.
- Costs repeat. Fees, spread and slippage apply to every trade, and subscriptions and any funding paid on perpetual futures add to them, so more trades raise the bar.
- The other side is not idle. In SEBI's August 2026 data on Indian equity derivatives, a few large firms take most of the gross profit, and a retail bot there trades in the same market.
- Rules age. A rule tuned on one period can stop fitting the next, and nothing in the software warns you.
- Some bots are not bots. In a case announced on 30 June 2022, the CFTC charged that the defendants falsely claimed trading was done by a proprietary bot; trading bot scams covers such cases.
If you are weighing a subscription, the cost side is in crypto trading bot price, which also carries the short form of this answer.
How to judge a claim that a bot works
Ask for evidence that could have failed: a dated live record, net of costs, showing the losses and how many versions were tried.
- Live or simulated? In the US, simulated results of a commodity pool operator or trading advisor must carry a prescribed statement that they have inherent limitations (17 CFR 4.41(b), read 5 November 2026).
- Net of what? Fees, slippage, funding and the subscription itself all count.
- Can you verify it? A screenshot or a seller's dashboard shows only what its owner chooses to show.
- How many versions were tried? A result picked from many attempts needs a much higher bar.
- Who holds the money and keys? Software does not need your funds or withdrawal rights.
- Is a return promised? The CFTC's advisory of 25 January 2024 calls claims of high returns, or of returns that are guaranteed, red flags of fraud.
Whether you may run a bot at all is a separate question, covered in are trading bots legal; rules depend on your jurisdiction and the platform's terms. For readers in Turkey: Article 99/A of Capital Markets Law No. 6362 treats activity by foreign crypto-asset platforms directed at Turkish residents as unauthorized (consolidated text read 5 November 2026). None of this is legal advice. Fincode builds software from rules the customer wrote and cannot say whether those rules will make money.
Frequently asked questions
Do trading bots actually work?
As software, they can, if built and tested well. Whether they make money depends on the rules, the costs and the market, and no audited figure for retail bots was found as of 5 November 2026.
Is bot trading profitable?
No independent figure was found showing that retail bots are profitable as a group. In SEBI's August 2026 data on Indian equity derivatives, individuals flagged as algo users had gross losses over FY25 and FY26, though that flag is broad and is not a test of bots.
What is a good trading bot success rate?
No number is good on its own. In the hypothetical example on this page a 70% win rate still loses money, because break-even depends on the average win, the average loss and the costs.
Do professionals use trading bots?
Yes. SEBI reports that in FY26, 86% of foreign portfolio investors and 47% of proprietary traders in Indian equity derivatives traded through algorithms. That shows use in one market, not that an algorithm creates returns.
Is bot trading better than manual trading?
It removes some human errors and adds software ones. One working paper on Indian retail accounts found algo trades behind the market by less than the same investors' other trades, which is not evidence of profit.
Sources
- SEBI: Profitability of Individual Traders in the Equity Derivatives Segment (FY25-FY26), August 2026 (opens in a new tab)
- Ghosh, Li, Zhang: When Human Meet Algorithm: the Adoption and Impact of Retail Algorithmic Trading (working paper, 15 March 2025) (opens in a new tab)
- ESMA: ESMA agrees to prohibit binary options and restrict CFDs to protect retail investors, 27 March 2018 (opens in a new tab)
- BIS Bulletin 69: Crypto shocks and retail losses, 20 February 2023 (opens in a new tab)
- Chague, De-Losso, Giovannetti: Day trading for a living? (working paper abstract, 11 December 2019) (opens in a new tab)
- Bailey, Borwein, López de Prado, Zhu: Pseudo-Mathematics and Financial Charlatanism (preprint; Notices of the AMS 61(5), May 2014) (opens in a new tab)
- CFTC Release 8854-24 (25 January 2024): Customer Advisory: AI Won't Turn Trading Bots into Money Machines (opens in a new tab)
- CFTC Release 8549-22 (30 June 2022): charges against a South African pool operator and CEO (opens in a new tab)
- SEBI circular of 2 September 2022: Performance/return claimed by unregulated platforms offering algorithmic strategies (opens in a new tab)
- SEBI adjudication order of 25 March 2026 on a stock broker's association with an algo platform (opens in a new tab)
- 17 CFR 4.41: Advertising by commodity pool operators and commodity trading advisors (Cornell LII) (opens in a new tab)
- Capital Markets Law No. 6362, consolidated text, Article 99/A (Turkish) (opens in a new tab)
This guide is general information about software and markets, not investment, legal or tax advice. Automated trading can lose money, past or backtested performance does not guarantee future results, and the rules that apply depend on your jurisdiction and your exchange's terms. Sources, checks and corrections: see the editorial policy.