How do trading bots work behind the scenes?

How do trading bots actually work?

A trading bot is a computer program, usually built as an Expert Advisor on a platform like MetaTrader 4 or 5, that reads live price data and opens or closes trades automatically once a pre-written set of numeric conditions is met, without a person clicking the buttons.

On the BWE podcast, trader Matt Bonney describes a bot as essentially "a strategy that runs like an algorithm and then trades that strategy for you." The video says the key difference from manual trading is precision. A discretionary trader can glance at a chart and decide a setup "looks right," but a bot cannot act on a feeling. Matt Bonney says the code needs an exact rule for when to enter, where the stop loss sits, and when to exit, because it cannot interpret a vague instruction like "if it hits this fair value gap, take a trade." He gives his own example: a bot he built for Bitcoin traded a breakout of the prior four-hour candle range. It ran the logic mechanically every time that pattern appeared, whether or not the trade actually worked out.

Because the rules are fixed, the video says a bot removes human emotion from its own execution once it is switched on. Matt Bonney frames that as the main appeal after years of manual trading, where he says he went "full tilt" and revenge-traded after a loss. But he is clear that removing emotion from execution is not the same as removing the trader from the process. A person still decides the lot size, the maximum drawdown the bot is allowed to hit, which sessions it can trade, and when to switch it off. Matt Bonney says his own discipline around those settings improved after he worked on his trading psychology, including daily journaling of trades and a gratitude practice with a mindset coach he refers to as Ty, and he ties his more consistent results directly to that work rather than to the bots themselves changing.

How do traders build and test a trading bot?

The video describes a step-by-step process for turning a strategy into a working bot, moving from a written rule set to a live account only after the idea has been checked against historical data. Here is the sequence Matt Bonney lays out.

1. Write the strategy as strict, mechanical rules. Every entry, stop-loss, and exit condition has to be expressed as something a computer can evaluate, such as a specific candle range, a fixed distance in pips, or a time-of-day filter, rather than a discretionary read of the chart.
2. Code the logic, or pay someone who does. Matt Bonney says he learned some coding through a program called Algo Pro but did not enjoy it and describes it as "not my jam." He says many traders instead pay a developer who specializes in building bots, since that person will typically be more efficient than someone learning to code from scratch. He describes the course creator behind Algo Pro as someone who, in his account, has also built bots for institutional clients and has spoken at UK forex industry events, though this is presented in the video as Matt Bonney's own understanding of that person's work rather than independently confirmed.
3. Download historical tick data. The video says a trader can pull several years of price data, in Matt Bonney's example around five years, to check the strategy against past market conditions.
4. Run a back-test inside the platform. MetaTrader's built-in back-testing tool, accessed through MetaEditor, replays that historical data through the code and reports the results, including the number of trades, win rate, and drawdown.
5. Review the results for an edge. Matt Bonney says this step tells you "if it's profitable or if there's an edge" before any real money is involved.
6. Tweak the parameters and re-test, rather than abandoning the idea. Matt Bonney separately says that constantly switching strategies without giving any one of them enough time made it harder to trust the process. He applies the same caution to bots: adjust the settings and re-run the back-test instead of discarding a strategy after one weak result.
7. Forward-test before going live. The video describes running a tuned bot on a demo account to see how it behaves in current conditions, separate from the historical back-test.
8. Deploy with defined risk settings and keep reviewing. Matt Bonney says he sets a lot size, a maximum drawdown limit (he mentions running one bot at roughly 25 to 30 percent), and filters such as which trading session it runs in or whether it avoids high-impact news, then checks in on performance rather than leaving it running unattended indefinitely.

Matt Bonney is candid that a back-test is not a guarantee. He describes trying and discarding many bots he calls "trash" and says he still takes losses on the ones that do work, because "nothing's foolproof" in trading.

What is a high-frequency style bot, and why did it stop working live?

Matt Bonney describes acquiring a high-frequency trading style bot from another trader that could pass prop-firm funding challenges within minutes by firing a very high volume of trades on an index. He says it worked reliably on demo evaluation servers but broke down once real money and live market conditions were involved.

In his account, the bot took a small live account, around $100, up by roughly 4 percent under normal conditions. He says that changed when the New York trading session opened during a period when the index was moving fast: the spread widened sharply as the bot was entering a trade, the stop-loss level was effectively skipped over, and the account lost money instead of continuing to grow. Matt Bonney also says a broker emailed him directly, warning that the bot was sending an unusually high number of trade requests and that the account would be closed if he kept running it on their live server. He describes this as one reason high-frequency style bots are hard to use safely outside of controlled testing conditions, since a strategy that repeatedly clears a demo challenge in minutes is not automatically viable once real execution, real spreads, and a broker's own trading rules are involved.

What is a martingale or grid trading bot, and why is it risky?

A martingale-style bot, sometimes built as a grid trading bot, increases the size of each new trade after a loss, for example doubling the position, while waiting for the price to reverse back in its favor. Matt Bonney describes it as originating from a gambling strategy before being adapted into automated grid trading tools.

He says he has run this type of bot on prop-firm accounts that do not allow high-frequency style bots but do permit a fixed lot size, since a martingale bot does not need to change its trade size to function in the way an HFT bot does. The version he trades runs during the Asian session, targeting a consolidating price range. Matt Bonney says this style of bot needs wide "breathing room," meaning a generous drawdown allowance, because the position size, and therefore the potential loss, grows each time the trade moves further against it before reversing. He says that with a tight daily drawdown limit of only 4 to 5 percent, the bot cannot be run at meaningful risk without breaching the limit quickly. In his own trading he says he has withdrawn real profit from this type of bot on multiple occasions, and separately says he has blown accounts running it too aggressively, which he attributes to pushing the risk too hard rather than a flaw in the concept itself.

What happens after a bot passes a prop-firm evaluation?

Passing a funded-account evaluation with a bot does not guarantee the payout gets paid. Matt Bonney says two prop firms he traded through withheld money he expected after his bots had already passed their evaluation rules, which he treats as a real, separate risk from whether the bot itself worked.

He describes receiving payouts from some firms without issue, but says two other firms disputed larger payouts he was owed, in one case an amount he estimates around US$9,000, citing a rule breach such as an IP address issue, despite saying he had traded only from his phone. In his experience, smaller payouts, roughly under a thousand dollars, tended to get paid without a problem, while larger payout requests were more often where firms pushed back. He describes this as a pattern he has seen affect other traders too, not only himself. Matt Bonney's broader point is that a funded account can become "a bit of a trap" for newer traders who assume that passing the challenge is the hard part, when navigating a firm's rules and its payout process afterward can matter just as much to whether a bot trader actually sees the money.

Matt Bonney also points out that some prop firms offer funded accounts scaling into the millions of dollars, but says the temptation for a bot trader is to keep pushing lot size and risk the same way they did on a small account. He argues that is a mistake: an account managing a large amount of capital does not need outsized daily percentage swings to be worthwhile, and a bot running on a bigger funded account should generally be sized down rather than up, so a single losing streak does not breach the firm's drawdown rules and undo the work of passing the evaluation in the first place.

How does bot trading compare with manual trading?

The video draws a consistent contrast between the two approaches across several practical points, based on Matt Bonney's own experience running both styles side by side. Neither approach is presented as safer or more profitable than the other; the video frames them as carrying different kinds of risk that suit different traders and different stages of a trading account.

| Aspect | Manual trading (per the video) | Bot or algorithmic trading (per the video) | |---|---|---| | Who decides each trade | The trader reads the chart and enters in real time | Pre-written code executes fixed, pre-set conditions | | Strategy type needed | Can use discretionary reads, such as market structure or liquidity concepts | Must be reduced to precise, numeric rules the code can evaluate | | Emotion in the moment | Matt Bonney says he went "full tilt" and revenge-traded after losses | The video says a running bot removes human emotion from execution | | Testing before risking money | Limited to journaling trades and reviewing charts afterward | Can be back-tested against years of historical tick data first | | Skill required to build | No coding needed | Coding skill or a paid developer, per Matt Bonney's account | | What went wrong, per the video | Discretionary errors and revenge trading, by Matt Bonney's own account | Widened news-time spreads, a broker server warning, and martingale drawdown, all described by Matt Bonney |

Matt Bonney's account of institutional trading comes from his own observation rather than published industry data. Before trading full time, he worked as an electrician servicing large finance-sector office buildings in Victoria, including trading floors for major funds, and he says the traders he saw there worked across multiple screens and appeared to weigh broader fundamentals rather than reacting to a single small-timeframe candle pattern. He also says that, in his understanding, large funds running algorithms target a small percentage return relative to a very large amount of capital, which is a different objective to a retail trader trying to grow a small account quickly through larger percentage swings. He offers this as his own impression from that period and from conversations with other traders, not as confirmed industry data.

Matt Bonney says his own preference is still manual trading day to day, and that he stepped back from heavy algo use for a period to work on psychology before returning to bots more recently, using them alongside manual trading rather than instead of it.

How do scammers fake profitable trading-bot results?

Matt Bonney says fabricated results are common in online trading, and he points to specific methods rather than a vague warning. He says trading platform displays, including MetaTrader account histories, can be edited so the entries and trade history shown on screen do not match what actually happened, something he says he was shown firsthand during a course.

He also describes brokers, in his account, sometimes providing traders with an account server labeled as live when it is functioning as a demo, so screen recordings look like real-money trading without real capital at risk. Beyond platform manipulation, Matt Bonney and the podcast host both describe softer marketing tactics they say are common: renting luxury cars or watches for short photo or video sessions to imply a lifestyle funded by trading profits, and unsolicited social media messages offering signals or mentorship from accounts impersonating known traders. Both speakers say the practical defense is the same one they used themselves: check whether a person posts genuine live trading calls and interacts with a real community, rather than judging credibility from images of cars, watches, or screenshots alone, and treat any bot or signal service promising fast, easy payouts with the same suspicion.

What do independent statistics say about the forex market and retail trading risk?

Two figures outside the video help frame the scale of the market these bots trade in, and the odds facing the individual retail traders using them, separate from any single trader's personal results.

The foreign-exchange market that forex-focused trading bots operate in is enormous by global standards. The BIS 2025 Triennial Survey reported that average daily OTC foreign-exchange turnover reached US$9.6 trillion in April 2025, a scale where any individual retail account, bot-driven or manual, is a negligible fraction of daily volume.

Against that backdrop, retail results skew heavily toward losses rather than gains. ASIC Report 828: Risky business records that 68% of Australian retail CFD investors lost money in the 2024 financial year. That figure applies to CFD trading broadly, not specifically to bot users, and it was not cited in the video itself. It is included here as independent, verified context for anyone weighing the personal wins and losses Matt Bonney describes, whether from manual trades, an HFT bot, or a martingale bot, against the documented outcome for the average Australian retail CFD investor over a full financial year. Read together, the two figures make the point from opposite directions: the market a trading bot operates in is enormous, but that scale does not, on its own, improve the odds for the individual retail account running it, automated or not.