A-book vs B-book: is your broker trading against you?

Your bot puts a stop at 1.0850. Price prints 1.0849, your stop goes, and then the market turns politely back up. That feels personal. Especially the third time in a week.

The question underneath is a good one: does my broker make money when I lose? The answer depends on where your orders end up.

What do A-book and B-book mean at a broker?

On A-book your broker passes your order to external liquidity providers and earns from commission or a markup on the spread. On B-book it keeps your order in house and takes the other side itself. Most brokers do both at once.

That last part is the bit missing from most explanations. This is not a choice between two species of broker. Nearly every large firm runs a hybrid model and splits the order flow: some goes out to the market, some stays inside.

How that split gets made is the interesting question. Often by client behaviour. Small, short-lived positions from clients who lose on balance are cheap to keep in house. Larger and consistently profitable flow is more likely to go out, because the broker does not want that risk on its own book.

Does your broker make money when you lose?

On the part of the flow sitting in the B-book, yes, because there it is on the other side of your trade. On the A-book part, no, because there it earns from volume whether you win or lose.

That sharpens the conflict question more than people think. Under a pure A-book setup your broker mainly wants you to keep trading. On a B-book portion, your loss is its gain, directly.

What you should not conclude is that somebody sits at a console picking off your stop. At a serious firm that does not happen, if only because the fine outweighs the take. What does exist is the incentive. And incentives shape behaviour over the long run, even when nobody makes a decision.

Why B-book is not automatically a scam

Internal execution is legal and common, and for small orders it is often faster and cheaper than routing out. Serious B-book brokers fill against real market prices, with slippage in both directions.

Picture the practice. Your bot sends an order of 0.05 lot. Routing that to a liquidity provider costs time and money and benefits nobody. Handling it internally at the same price is faster for you and cheaper for the broker. Nothing wrong there.

The problem starts when execution gets steered: spreads that widen right around your stop, slippage that structurally falls one way, requotes that only arrive when you are in profit. That is a different story, and it is exactly what you want to be able to see. For the broader trust check, from licence to segregated client money, we have a separate piece: how to check whether your forex broker is safe.

Why this hits a bot harder than a manual trader

Because a bot takes more trades and its edge is thinner per trade. Someone taking two manual swings a week will not notice half a pip. A bot taking fifty setups a week pays that half pip fifty times.

Run it through. Fifty trades a week, half a pip of extra cost per trade, at 0.10 dollars per pip on 0.01 lot. That is 2.50 dollars a week on a small account. Sounds like nothing. Scale it to 1 lot and you are at 250 dollars a week of difference that appeared nowhere in your backtest.

On top of that, bots tend to trade at fixed moments. Around the London open, around news releases, at a set time. That is precisely when spreads move most. A manual trader can wait for calm. Your bot does what the rules say. What account type does to those costs is covered in Raw ECN vs standard account for a trading bot.

How to measure what happens to your orders

Compare requested price with fill price across a few hundred trades in your MT5 history. If slippage almost always lands against you and you never get a fill better than the price you asked for, that is the strongest signal you can get without access to your broker's systems.

This is the core, so let me be precise. Slippage should fall both ways. Sometimes you get a worse price, sometimes a better one. That second kind is positive slippage, and on orders that genuinely reach the market it simply exists. On a flow handled entirely in house and steered, it often disappears quietly.

Three things you can pull from your own history without much effort:

• The share of trades with positive slippage. If that sits near zero across hundreds of orders while negative slippage shows up regularly, you have a one-way street.
• The gap in average slippage between quiet hours and news releases. A large gap is normal. Barely any gap on a bot that trades through NFP is odd.
• The count of rejected or requoted orders, and when they land. If they cluster on trades where you are in profit, you have found a pattern.

Do not do this over twenty trades. Over twenty trades anything can be coincidence. From a few hundred it starts to mean something, which is also why this research so rarely happens: it takes patience, and most people switch brokers before they export their history.

Why your demo always executes better than your live account

Because a demo is almost always handled entirely in house with cosmetic fills. You get the price you asked for, without delay and without requotes. There is no real counterparty on the other side.

There is a handy test hiding in that. Run your bot for a few weeks on demo, then the same stretch on a small live account, and compare the execution. Some difference belongs there, because live is real. A large and persistent difference says something about how your actual flow gets treated.

Be clear about what you are measuring. You are measuring execution, not your strategy. A bot that profits on demo and loses live can have an execution problem or simply a thin edge. Separating those two takes enough trades on both sides, and the method is in how to judge a trading bot's results.

What the rules do and do not guarantee

Article 27 of MiFID II requires investment firms to take all sufficient steps to obtain the best possible result for the client when executing orders. That is a real obligation, and regulators enforce it.

What it is not: a guarantee that your order goes to an external venue. Best execution weighs price, cost, speed and likelihood of execution together. An internal fill that happens fast and at a good price can satisfy it.

The reporting side has also thinned out. The old RTS 28 reports, where firms published their top five execution venues annually, were scrapped in the review because almost nobody read them. In a final report from April 2025, ESMA moved the emphasis to an execution policy a firm must be able to demonstrate on an ongoing basis rather than an annual list. For you as a client that means less paper to inspect, and more value in what you measure yourself.

What to look at when picking a broker

Start with the licence and segregated client money, because without those two the rest is irrelevant. Only then look at the execution model.

Ask the broker outright how it handles orders and whether it runs hybrid. A firm that is open about it gives you a usable answer. A firm that dodges, or shouts "we are 100% ECN" without explaining, has also told you something. Overblown claims are a signal in themselves, and the others are in how to spot a trading bot scam.

What matters in our own setup sits on the same point. The account is in your name, you change your MT5 password after the connection is made, and we cannot withdraw anything. We take 30 percent of your profit and nothing when you lose, so if you lose consistently we get zero. That incentive points the same way yours does, and that is exactly what is missing when a counterparty holds your flow.

So do not go hunting for the broker that promises never to be on the other side. Find the broker where you can check what happens to your orders, then measure it yourself. A few hundred trades and an export of your history say more than any page of promises.