Path A · Spot XAU/USD · Module A4
A1 to A3 established that your broker sets your price, fills your order, and collects your costs. This module confronts what that means at its most uncomfortable — and then turns the obvious conclusion completely on its head.
What you'll learn
Say the hard thing first
A3 ended on a question, and we are not going to dodge it. You now know that XAU/USD is over the counter, that there is no exchange between you and your trade, and that your broker sets the price, fills the order, and takes the costs. So here is the sentence that follows, stated as starkly as it deserves: when your broker holds the other side of your trade, your loss is its gain. Money you lose does not vanish into the market — there is no market, only the firm. It becomes the firm's revenue.
Most sites that earn money sending you to brokers will never put it that bluntly, which is exactly why we are starting there. The conflict is real. It is structural. It does not go away because a broker is regulated or well-reviewed. Pretending otherwise would be the fastest way to lose your trust, and you would be right to withhold it.
But — and this is the whole module — the conclusion almost everyone draws from that sentence is wrong. "My broker profits when I lose" feels like it should mean "my broker is rigging my fills to make me lose." It doesn't. To see why, you have to understand what taking the other side actually involves, mechanically. And when you do, the moral picture turns out to be close to upside down.
The fork, up close
Every module in this path has ended at the same fork: your order reaches the broker, and the broker fills it one of two ways. We have walked past it three times. Now we stop and look directly at it, because this fork is the entire subject of the module.
Your order reaches the broker
Your broker
Holds your order and makes one decision: fill it in-house, or pass it out. That single choice is the fork — and it decides who your counterparty really is.
The fork, zoomed: who actually fills you
B-book — broker fills you
The broker takes the other side itself, off its own price feed. Your order never leaves the firm; the broker is your counterparty.
Fill: instant, at the quote — no external liquidity to consume
A-book — broker hedges you out
The broker isn't your counterparty — it acts as agent, hedging your trade with a liquidity provider. The LP is the real other side.
Fill: clean if the LP accepts the hedge — bad if the LP rejects it
Same fork as A1, A2 and A3 — now read for one thing only: who fills you, and what that does to your fill. The third option, hybrid, isn't a separate branch. It's the broker choosing which branch to send each order down, trade by trade.
Three terms, demystified. On A-book, the broker isn't your counterparty at all — it acts as your agent. It fills you, then immediately hedges itself by placing the mirror of your trade with a liquidity provider: you buy a tenth of a lot of gold from the broker, the broker buys a tenth of a lot from its LP to stay flat. Think of it as the broker copy-trading your position upstream so it carries no market risk — it earns from the spread markup or commission for standing in the middle, and the LP is the real other side. On B-book, the broker is itself the counterparty: it fills you internally, off its own feed, and keeps the risk. If you lose, it keeps your loss; if you win, it pays from its own book. There is no hedge, because the broker is content to be the other side.
Hybrid is what virtually every real retail broker actually runs, and it isn't a third venue — it's a routing rule. The broker segments its flow and decides, order by order, which book to use. The open secret of the industry, and you should hear it plainly: brokers tend to internalise the flow they expect to lose — the bulk of retail — and pass the flow they expect to win out to the market. That sounds cynical until you see why it's also the arrangement that gives most clients their best fills. Which is the turn the next section makes.
The part nobody tells you
Here is the intuition almost everyone carries, planted by a decade of broker marketing: A-book is the honest model — your order goes to the "real market" — and B-book is the shady one, where the broker bets against you in the dark. So you should want A-book. On execution quality, this is precisely backwards.
Think about what each fill physically involves. On a clean B-book, the broker fills you off its own price feed, instantly, at the quote you clicked. There is nothing to hedge and no one to ask — the broker simply takes the other side at the price shown. The fill is immediate and clean. On A-book, the broker has to hedge you out with its LP, and that introduces a dependency that B-book doesn't have: the LP has to accept the hedge. Most of the time it does, and your fill is just as clean. But the LP gets a brief window to look at the trade before accepting it — the industry calls this last look — and it can reject. An LP might reject for any number of reasons: the price has moved in the moment it took to reach them, their own risk limits, the size of the order, or simply not wanting that exposure right then.
When the hedge is rejected, the clean fill is gone and one of two things happens. Either the broker wears the hit — it's now holding your trade unhedged, having failed to offset it, and must either accept that risk or scramble to re-cover it — or the order is passed onward into the wider market, what desks call the Street, where it meets real depth and fills at genuinely worse prices. That's slippage, and it is not a malfunction — it is what touching real liquidity does. The crucial point: this rejection-and-slippage risk exists only because the trade had to leave the firm to be hedged. A clean B-book has no hedge to be rejected, so it has nothing to slip. The A-book downside isn't that every fill is bad — it's that A-book carries a tail the B-book structurally cannot.
Clean B-book
Better execution
Filled off the broker's feed at the quote. No hedge to place and no LP to reject it, so nothing to slip. The fill you see is the fill you get.
A-book / STP
Carries a tail risk
Filled cleanly when the LP accepts the broker's hedge — which is most of the time. The downside is the tail: a rejected hedge the broker must eat, or an order passed to the Street where it slips on real depth.
Now look at the words brokers actually market with — NDD, STP, ECN — because they're sales terms, not execution models, and they don't even sit on the same axis. NDD ("no dealing desk") just means the fills are automated, with no human dealer touching your order. That's it — and a B-book broker with a slick auto-routing system is fully NDD, so the term tells you nothing about who your counterparty is. ECN is about pricing: raw, tight spreads. But a B-book broker can quote those same raw spreads and fill you against them, so ECN tells you nothing about counterparty either. Only STP — straight-through processing — actually denotes routing: it means A-book, your trade hedged out. Brokers stack all three on a landing page so the bundle sounds like "we're not the other side," when not one of them, on its own, establishes that.
And the one that does carry meaning — STP, the A-book routing — is the one with the hidden cost. Because hedging you out means every fill now depends on an LP accepting the hedge, which is exactly where last look, rejections, and the slip to the Street live. So the honest reading of the marketing inverts it: the label that genuinely signals "not your counterparty" (STP) is also the one that exposes you to the rejection tail, while a clean automated B-book — equally entitled to call itself NDD, equally able to quote ECN spreads — simply absorbs that risk by being the other side. The transparency words are real; they're just not answering the question you think they're answering.
So the ranking your instinct produces is inverted. The routing marketed as honest carries the execution tail; the model marketed as shady, run cleanly, gives the more reliable fill. Hold that — and then ask the obvious question it raises, because it's the right one: if a clean B-book fills me at least as well, why do so many people lose money on B-book brokers?
Not rigged — something worse
If internalising your flow gives you a better fill, then the broker clearly isn't emptying your account through your fills. So where does the money go? The uncomfortable answer is the one this whole path has been building toward: aggregate retail flow loses on its own, without anyone touching it.
The broker running a B-book doesn't need to cheat, because it has read the statistics. The reasons retail loses are the ones we've already named, module by module. The leverage from A2 — a position sized so a one-percent move erases the margin — does the work unprompted. The swap from A3 — the long carry bleeding tens of thousands a year on a held position — does more of it. And the behaviour we'll confront in How to Trade: overtrading into spreads, cutting winners, letting losers run, revenge-trading a bad morning. Put a large book of retail accounts together and, in aggregate, they lose to these forces reliably enough that the broker taking the other side is simply collecting the result. No stop-hunting required. No rigged fills required. The house edge is structural, legal, and behavioural — not a conspiracy against you specifically.
This connects straight back to F7's hard look at the reality of retail trading. The conflict of interest exists, exactly as the opening sentence said. But it is not the mechanism that takes your money. You are — or more precisely, the predictable patterns you share with every other undercapitalised, over-leveraged retail account. And that is genuinely the more useful news, because it points at the only variable you actually control. A rigged market would be hopeless. A structural edge built out of leverage, carry, and your own behaviour is a problem you can work on — which is the entire purpose of How to Trade.
The one thing worth checking
Everything so far defends the clean B-book. Now the other edge of the blade — because not every B-book is clean, and the specific way a dishonest one robs you is concrete enough that you can look for it. It isn't betting against you. As we've seen, that's just the model. The dishonest move is subtler: manufacturing the A-book problem inside a B-book that doesn't have it.
Recall the inversion: a clean B-book has no slippage to pass on, because it never hedged anything — it filled you at the quote off its own feed, and no LP ever got the chance to reject it. So a B-book broker that slips your fills is charging you for a rejection that never happened, on a hedge it never placed. And the tooling makes it trivial. The bridge software sitting between the platform and the broker's book — the layer A1 named in the chain — exposes slippage, markup, and execution delay as configurable settings. A broker can dial in a rule that slips fills by a set amount once a trade crosses a size threshold, mimicking the market impact you'd have suffered on a real exchange, and bank the difference. Some desks call this depleting or warehousing against synthetic liquidity; the plain description is better — they are inventing a cost to charge you for, then keeping it.
The distinction that matters
The question was never "A-book or B-book?" It's "are my fills clean, or synthetically degraded?"
This reframes the entire problem of choosing a broker. The naive trader asks "is this an A-book or a B-book broker?" — and the answer barely matters, because a clean B-book is fine and an honest A-book just costs you real slippage. The right question is whether the broker is giving you clean fills or quietly degrading them. You can probe it: watch your fills against the quote on ordinary-sized orders away from news, compare execution on a demo versus a live account, read independent reports of slippage that only ever runs against the client and never in their favour. None of this is foolproof from the outside — which is exactly why the next module exists.
Where this leaves you
Pull the threads together. The conflict of interest is real — your loss can be the broker's revenue, and no amount of marketing erases that. But a clean B-book gives you the more reliable fill, not the worse one; the transparency words — NDD, STP, ECN — don't establish who your counterparty is, and the one that does mean A-book routing quietly carries the rejection tail a B-book would have absorbed by being the other side; and the reason retail loses isn't rigged execution but the structural and behavioural edge that aggregate retail flow hands the house for free. The genuine danger isn't that your broker takes the other side. It's that a dishonest one manufactures slippage it has no real reason to charge.
Which means the question that decides whether a broker is safe to fund was never about the book model at all. It's whether the broker is constrained to behave — by a regulator with real teeth, by a jurisdiction that can hold it to account, by a reputation it can't afford to burn. A firm that can configure synthetic slippage with a few clicks will or won't, depending almost entirely on what stops it. That constraint — jurisdiction, regulation, and what to verify before you transfer a cent — is the whole of A5.
Carry this into A5