Start with the institutional side, because it sets the vocabulary. A fund that needs to buy a large quantity has a problem that has nothing to do with direction: moving that size at once would push the price against itself. So it uses an execution algorithm. VWAP slices the order across the session to track the volume-weighted average price. TWAP slices it evenly across time. Implementation shortfall algos trade more aggressively early to reduce the risk of the price running away. None of these has an opinion about where the market is going. They are cost-control tools.
Retail algorithmic trading is a different animal. There is no size problem to solve, because a retail order fills instantly at the quoted price. What a retail algo automates is the decision itself: when to enter, where to place the stop, when to exit. It is a trading strategy written in code, most commonly as an expert advisor on MetaTrader or as a script driving orders through an API.
What automation actually gives you
Three things, and they are worth having. Consistency: the code takes every signal, including the eleventh one after a losing streak, which no human reliably does. Coverage: it watches instruments and sessions you cannot stay awake for. Speed: it acts within milliseconds of a condition being met, which matters for breakout entries and matters enormously for exits.
What it does not give you is an edge. A rule set that loses money when you trade it by hand loses money faster in code, because the code takes more of the trades. This is the single most common misunderstanding among people buying automated systems, and it is why expert advisors sold with a promise of hands-off returns should be treated as a marketing claim first and a piece of software second.
The failure modes that actually kill retail systems
Very few retail algos fail because the logic was badly coded. They fail for a small set of repeatable reasons.
- Overfitting. Parameters tuned until the historical curve looks smooth describe the past, not the market. Fewer parameters and out-of-sample testing are the only real defences, covered in backtest overfitting.
- Costs ignored in the test. A backtest run on fixed spread and zero slippage flatters short-horizon strategies enormously, because their edge per trade is small.
- Regime dependence. A trend-following rule set tested through a trending decade will lose steadily through a ranging one, and the code cannot tell you which you are in.
- Infrastructure. A dropped connection, an unattended platform update or a machine that went to sleep leaves positions open with no manager, which is why running on a VPS is standard rather than optional.
- No kill switch. Systems need a hard daily loss limit that flattens and stops. Without it, a logic error can trade all night.
The overfitting problem deserves emphasis because it is invisible from inside. A curve fitted to history looks exactly like a discovered edge on the report page. The distinction only appears in data the optimiser never saw, which is why forward testing on a demo or a small live account for a meaningful period is not a formality.
A backtest is a hypothesis. A forward test is evidence. Anyone showing you the first while calling it the second is either confused or selling something.
How much of the market is automated
A large share of volume in major equity and futures markets comes from automated participants, spanning market makers, execution algos and statistical arbitrage. The precise share varies by venue and by measurement method, so treat any single percentage you see quoted with suspicion. What matters for a retail trader is the practical consequence: the counterparty on the other side of a fast move is usually a machine, and the price reacts to scheduled data in the first fraction of a second.
That is why manual news trading at the moment of release rarely works and why spreads behave the way they do around releases, as set out in news trading. It does not mean retail cannot compete. It means retail should not compete on speed. The available ground is longer horizons, positions machines are less interested in, and risk control, which no amount of speed substitutes for.
Worth separating two more terms that get mixed up. A signal service sends you a message and you place the order, so the execution risk and the discipline stay with you. A trade copier links accounts so that orders placed in one appear in another without your involvement. Neither is algorithmic trading in the strict sense, because in both cases a human made the decision. The distinction matters when you are assessing a track record: a copier reproduces the provider decisions and their timing, while a signal service reproduces only the first and leaves the second to you, and the gap between the two shows up as slippage against the published result.
Where an algo makes sense for an individual
The realistic starting point is not full automation. It is partial automation of the parts of your process that you already know you execute badly: moving the stop to break-even at a defined point, closing at a target, sizing the position from a fixed fraction of equity, refusing a trade after the daily limit. These are mechanical, testable and immediately useful, and they do not require any predictive claim.
From there, if you want a fully automated strategy, the honest sequence is: a written rule set with no ambiguity, a backtest on realistic costs, a walk-forward test on unseen data, months of forward testing at small size, then a slow increase. Most abandoned algos never got past the first step, because the rules were never actually written down in a form a machine could follow. That is the same discipline as a trading plan, enforced by a compiler instead of by willpower.
Automated trading carries the same high risk of loss as manual trading, plus the operational risk of running unattended software. The advantage is repeatability, not certainty, and any system marketed on the second rather than the first should be read carefully before any money moves.
"An algorithm removes hesitation from your process. If the process was wrong, you now lose money faster and more reliably."
— Alex Onta, Executive Director, SINGUARD
Key Takeaways
- Institutional execution algos manage cost and market impact; retail algos automate a trading decision.
- Automation delivers consistency, coverage and speed, and delivers no edge at all on its own.
- Overfitting, ignored costs, regime change and missing kill switches account for most retail algo failures.
- Partial automation of stops, targets and sizing is more useful to most traders than a full black box.
Frequently Asked Questions
Is algorithmic trading profitable for retail traders?
Automation does not create profitability. It executes a rule set consistently, so a set with a genuine edge is applied more reliably and a set without one loses more consistently. Trading carries a high risk of loss either way.
What is the difference between an expert advisor and an execution algorithm?
An expert advisor is a retail program that decides when to enter and exit. An execution algorithm is an institutional tool that splits a large order across time or volume to reduce market impact, and it has no directional view.
How long should a system be forward tested before going live?
Long enough to cover more than one market condition and a meaningful number of trades. A high-frequency system reaches that in weeks, a swing system may need many months. Trade count and variety of conditions matter more than the calendar.
About the Author
Alex Onta is an Executive Director at SINGUARD. He built eTrader, the terminal, the mobile apps, eTrader Broker, Copytrading, Business and Community, along with the worldwide clustered-server infrastructure it all runs on, with his brother Roman Onta helping on the design, and he leads that division today. Together with Roman he builds the Prop Firm CRM, the Broker CRM, Scalegram and CopySignals, and the two of them carry worldwide compliance, payment processing and international business structuring side by side. He lives and works in Dubai for most of the year. Meet the executive duo leading Singuard's five divisions.