How a retail broker actually works, end to end — the chain it sits inside, what its desk does with your flow, where the price is really made, what it really costs you, the edge that survives the machine, and how to tell an A-book fill from a B-book one, what to check before running an algo there, and the rules that bind it in every major jurisdiction. New to the jargon? Every term is defined in plain English in the last tab. Numbers without a source are marked illustrative.
Four real layers: you, the broker, the Prime-of-Primes, and the tier-1 banks. ("Prime broker" isn't a separate tier — it's a credit service the banks sell; "LP" is just whoever streams price to the layer below.) The broker is one commercial entity — its trading server, bridge and dealing desk are internal, not separate tiers. A-booked flow (the broker passes your trade out to a bigger provider) leaves it for one of the LPs its bridge aggregates — typically a mix of PoPs, banks / non-bank market makers and ECN / MTF venues; B-booked flow (the broker keeps your trade in-house and is the other side of it) stays inside — on the MT5 server or inside the bridge. It isn't binary: brokers often internalise first and hedge only their net exposure, and switch book per account or per trade (see Tab 06).
Orders & risk ↓ Price feed ↑ Credit / buy-side price made spread marked up
Four real layers — and two of these words aren't layers at all. A typical broker aggregates a basket of Prime-of-Primes (PoPs) — e.g. Finalto, IS Prime, iSAM, StoneX — through its own bridge tech (not a separate party). An MTF venue such as LMAX Exchange is a different kind of LP (firm prices, no last look — see the glossary), and many brokers also connect to banks and non-bank market makers directly. Each PoP rents a prime-broker (PB) line from a tier-1 bank; the PB is a service the banks sell — credit + give-up — not a separate company. Those same tier-1 banks — JPMorgan, UBS, Citi, Deutsche, Barclays, plus non-bank market-makers XTX, Citadel, Jump — then make the actual price. So the only distinct company-layers are PoP and tier-1 bank: "LP" is just a role (each layer is the LP of the one below) and "PB" is the banks' credit desk. Hedge funds and asset managers rent their own bank-PB line too — peers of the PoPs, not customers downstream of your broker. Layers can add markup before the price reaches you, which is why a retail account typically trades above the tier-1 floor below. The broker's own markup is set per MT5 account group — a spread markup over the bridge price, plus commission and any IB rebate — so two accounts in the same group see the same prices whichever book they are on.
Where the spread comes from
Illustrative EUR/USD, standard account, pips. Tier-1 raw is the floor; each layer stacks markup. (ECN / raw-spread accounts compress this to near-zero spread + a commission instead.) These figures are made up to show the stacking — they are not measured from any broker.
Every order hits a routing decision: A-book (hedge out to an LP, earn the markup / commission, carry little market risk) or B-book (become counterparty, keep client losses, warehouse risk). A real-time risk-oversight system sits across both, consolidating exposure and flagging sharp flow to switch over to A-book. In practice it is a dial, not a switch: brokers net client flow, hedge only what is left, and set the routing rule per account, group, symbol or size — in the MT5 server or inside the bridge.
A-book — hedged, little market risk B-book — warehoused risk Risk oversight / control
Who gets B-booked — and the trap in winning
Routing isn't random. The classifier leans on a base rate: most retail accounts lose, so many brokers warehouse new flow by default (it varies by broker). You may graduate to the A-book by proving you win — which the desk reads as toxicity.
A brand-new small-deposit accountoften B-booked by default — the base rate says it loses
It keeps winning, with a clean mark-outmay be "promoted" to A-book (hedged out as a threat) — or restricted instead
Many brokers run a C-book (hybrid)a set ratio warehoused, the rest hedged — risk dialled by client grade (terminology varies by broker)
The inversion: you're kept in-house because you're expected to lose, and one common reward for being good is to be moved to the side of the book where the broker no longer profits from your losses. Other outcomes are wider spreads, added delay or account restrictions (see Tab 05 and Tab 10). On B-booked flow, your skill is the broker's cost.
Risk oversight — B-book exposure (intraday)
The broker's own leftover position from B-booked trades, by instrument, in US-dollar terms. It's the mirror image of how clients are positioned — when clients are net long, the broker is net short. The rows add up to the headline net delta (overall size and direction of the bet). Illustrative figures for a broker running a few hundred yards/month.
Net B-book delta
−$36M
net short, USD notional
Open P&L (today)
+$0.42M
B-book mark-to-market
VaR 95% / 1d
$1.9M
limit $3.0M · 63% used
Open alerts
3
1 critical · 2 watch
Net warehoused exposure · USD notionalillustrative · static
Acct #4471 — XAUUSD, +3.2 lots added in 40s, win-rate 91% over 12 trades. Pattern matches latency / stale-quote arbitrage on the gold feed. → recommend reclassify to A-book; breach of B-book per-account delta limit.
14:18:50 · watch
Group "VIP-2" — aggregate BTCUSD long building, net long trending one-way with news. → partial hedge suggested; VaR contribution rising.
13:55:11 · watch
Acct #2207 — behavioural shift: avg hold time collapsed 4h → 90s post-NFP. → flag for review, possible toxic flow.
auto
LP Reconciler — A-book vs LP positions reconciled, 0 breaks. Expected vs actual routing aligned.
The whole retail chain is OTC — bilateral at every link. But the underlying price a CFD references is discovered differently by asset: currencies have no central exchange; gold and most commodities do.
Global market turnover — to scale
Average daily volume, US$ trillion. In each group the top bar is the total; the indented — bars below break it into venues/instruments. FX = BIS Triennial, Apr 2025 (net-net = adjusted for double-counting); the three largest instruments are shown — ~$1.2T of options/other isn't broken out, so they don't fully re-add to the total. Gold = total across both its venues: the exchange (COMEX futures) plus the OTC market (LBMA London spot), Oct 2025 — a record-high month, so typical is lower, which only widens the gap.
Retail broker volumes — how the industry counts it
Brokers measure in yards/month (1 yard = $1 billion of notional). These are the Q4 2025 top five by CFD volume from Finance Magnates (FM), an industry data provider. Read these as monthly activity, not money at risk: CFD notional is leverage-inflated (a 1:500 position books 500× its margin) and double-counted (open and close deals are typically each tallied), so these monthly figures are not comparable to the BIS daily, net-net turnover in the chart above.
Why this matters — retail references price, it doesn't set it. Global FX turns over ~$9.6T net-net every day (BIS, all instruments); the retail CFD volumes above look large partly because they're leverage-inflated and double-counted, so the two aren't comparable. Retail flow never sets price, it only references it — one reason a broker can warehouse some of it without touching a real venue, and why a single large winner is visible against the book.
Where the price is actually made
"What does the market maker watch?" — three answers. Blue = OTC venue · amber = exchange venue. All feed the same OTC CFD at the bottom.
The structural distinction: gold and commodity market makers typically anchor to a real exchange print (COMEX front-month + the EFP basis). FX has no central spot venue, so the reference price is a consensus constructed across OTC platforms; CME FX futures serve only as a secondary cross-check, not the source of price.
Nobody in this chain puts up the full value of a trade — everyone trades on borrowed buying power (leverage), backed by a deposit called margin (retail) or collateral (institutional). Each layer posts collateral to the layer above it. When prices move the wrong way, margin calls cascade back down toward the client.
Collateral posted toward the market ↓ Margin calls cascade toward the client ↑
The leverage asymmetry
The leverage a broker grants its clients and the leverage it receives from the market are two very different numbers — and the gap between them is the business model.
What the client is offered1:30 regulated · up to 1:2000 offshore
Offshore leverage is usually dynamictiered down as equity grows — e.g. 1:2000 under ~$1k, stepping to 1:1000 / 1:500 / 1:200 on bigger balances
What the broker itself gets from its LP / PoPusually far less than clients get — set by its credit agreement (~1:30 is illustrative)
So how can a broker hand out 1:2000 while getting far less itself? It doesn't pass that leverage upstream. B-booked trades are never hedged, so the broker posts no collateral to an LP on them — it just holds regulatory capital and leans on the statistical edge that most retail accounts lose. Only the A-booked slice is funded at the market, where the broker posts institutional margin on whatever terms its credit agreement sets (an illustrative ~1:30) — and only on its net hedged exposure, which is far less than the sum of client positions. High client leverage is therefore largely synthetic — backed by the broker's own balance sheet and risk appetite, not by matching leverage from the banks.
Try it — a gold trade at your leverage
1 lot of XAU/USD = 100 troy ounces — that's the fixed contract size; the lot defines it, not your deposit. Drag the sliders and watch how the leverage you pick decides how little margin holds that gold — and how small a move it takes to stop you out. Gold price is illustrative; set it where you like (it traded around $4,200/oz in October 2026).
$2,000
1.00 lots · 100 oz
1:500
$4,200
Contract size
100 oz
100 oz × lots
Notional value
$420,000
what you control
Required margin
$840
notional ÷ leverage
Margin level
238%
equity ÷ margin
P&L per $1 move
$100
per $1/oz in gold
Move to stop-out
$15.80
against you, at 50%
…as a % move
0.38%
how little it takes
Free margin
$1,160
deposit − margin
Drag a slider to begin.
How a retail margin call actually works
The numbers a retail trader sees on MT4/MT5. Margin level = Equity ÷ Used Margin × 100. The broker watches this percentage. Round numbers for the arithmetic — a $1,640 loss on one mini-lot would be an extreme move; real accounts reach these levels with larger sizes.
You deposit (equity)$2,000
You open 1 mini-lot EUR/USD @ 1.0800notional ≈ $10,800
Required (used) margin at 1:30 leverage$10,800 ÷ 30 = $360
Margin call = the warning (often 100% level). Stop-out = automatic liquidation (often 50% level; both levels are set per broker / MT5 group) so your loss can't run past your deposit. In the EU/UK/AU, retail clients also get negative-balance protection — you can't lose more than you put in.
When the cascade breaks — Swiss franc, 15 Jan 2015. The Swiss central bank scrapped its EUR/CHF floor with no warning; the franc surged — EUR/CHF fell as much as ~30% within minutes. Prices gapped straight through stop-out levels, so brokers couldn't close client positions at the trigger price — clients went deeply negative, beyond what the brokers could collect. But those brokers still owed their own LPs and prime brokers the collateral on the other side. The mismatch sank several firms (FXCM needed an emergency rescue; Alpari UK went insolvent). This is the whole reason every layer demands collateral and watches margin in real time.
An LP doesn't just stream prices — it grades the flow it gets back, trade by trade, and re-prices you on it. Flow that consistently costs the LP money (sharp or toxic) can set off a remediation ladder that runs the whole way down: LP → broker → your individual login.
How an LP grades your flow — the mark-out
The decisive metric is mark-out: the LP's profit or loss on your fill, measured at fixed horizons after it (+1s, +5s, +30s, +1m…). If the price keeps moving your way right after you trade, the LP is underwater — your flow is toxic. Benign flow drifts back toward zero; toxic flow runs against the LP and stays there.
Mark-out is the headline, but the LP scores it alongside other tells: very short hold times (seconds), trades clustered on price updates (latency / stale-quote signatures), one-sided bursts around news, an abnormal win-rate, and the broker's fill ratio & reject pattern.
Sharp vs toxic
Two shades of the same problem for the LP — informed skill versus structural extraction.
Sharp
Informed, skilled flow — news, macro, genuinely good timing. The LP loses, but to skill. Often tolerated if it isn't too fast or too one-sided.
Toxic
Flow that extracts from the LP's own pricing rather than the market — latency / stale-quote arbitrage, last-look gaming, feed-lag picking. There's no genuine edge for the LP to price around.
In practice the LP doesn't moralise the difference: if your mark-out is negative for it, you're priced out either way.
Last look — the reject you never see
The quote an LP streams isn't always a firm promise. Last look is a short window — typically milliseconds — in which the LP, after the broker's bridge has hit its quote, can accept or reject the trade after checking where the market is now. Used as a price check it's a risk control against stale quotes; the question is whether the check is symmetric. Some LPs apply it asymmetrically, and that skews outcomes against the trader even on a genuine A-book.
Symmetric price checkrejects any trade where price moved beyond tolerance — in either direction
Asymmetric price checkaccepts when price moved in the LP's favour, rejects when it moved in the client's
What the trader seesno "last look" label — a reject, a retry, a later fill at a different price, or "Request rejected" / "Prices changed"
Extra latency in fast marketscomes mostly from reject → retry at another LP (each retry is another round trip), not from a longer hold window
Toxic flow getsmore rejects and tighter tolerances at LPs that grade flow
Why it makes your edge fragile: an exchange fill is firm — hit the price, it's yours. A last-look fill is conditional. With an asymmetric check, the cleaner your timing the more your winners get rejected and your losers get filled — the spread you beat on screen is quietly handed back at the point of execution. Source: GFXC last-look report (FX Global Code, Principle 17).
The LP's remediation ladder
What the LP does to the broker's stream as toxicity rises. A typical, not universal, escalation — terms vary by LP and agreement. Each rung is cheaper than the last resort of pulling the line.
1 · Skew & widenworse, asymmetric pricing on that stream
2 · Last lookmore rejects and tighter price tolerance on your orders
3 · Added hold / latencya longer LP-side hold before accept / reject, at LPs that do this — blunts speed edges
4 · Lower fill ratiopartial fills and more outright rejections
5 · Demand taggingbroker must split toxic flow onto a separate, tagged stream
6 · Mark-out chargessome LPs re-rate or surcharge the broker when mark-outs stay negative (terms are commercial and vary)
7 · Yellow card → cutpull the stream and terminate the relationship
What rolls downhill to your login
The broker is now being charged — or cut — for your flow, and can't pass you to an LP cleanly. So it acts on the specific login: hedge it if an LP will still take it, or fall back on a toolkit of execution settings and restrictions. These are things some brokers do, not things every broker does.
execution
Execution delay — a dealer plugin or bridge / group setting adds latency to your orders, so stale-quote and speed edges miss.
fills
Slippage & rejects — asymmetric: negative-only slippage going in, rejects on the trades that would have won. (Requotes only exist in MT5's Request and Instant execution modes — never under Market execution.)
pricing
Spread widening — your login is quoted a wider markup than the public book.
limits
Caps & bans — max lot cut, EAs disabled, news-trading windows blocked.
money
Profit clawback — winning trades can be voided under a "latency / arbitrage" clause in the T&Cs.
account
Withdrawal friction → closure — payouts stalled, then the account is restricted or closed.
The asymmetry that ends most arbitrage: you can be consistently profitable and still get throttled or shut down — because your edge is, line for line, the LP's and the broker's loss. The faster and more reliable the edge, the sooner the mark-out flags it, and the sooner the ladder reaches your login. Durable edges tend to be the ones that look less sharp on a mark-out curve.
From the MT5 terminal you can't see which book your order went to. What you can see are execution fingerprints — timing, slippage, rejections, order IDs — and whether they behave like a market (they move with conditions) or like a setting (they don't). This tab shows the route an order can take, why outside observation can't settle the question, and how far a careful test can take you.
A-book path — all the way to an LP B-book — stops at the server or the bridge Book switching / control
The route — and where it can stop
An order always starts at your terminal and always reaches the broker's MT5 server. Whether it goes further is the broker's choice — and a B-booked order can stop at two different places.
Why you can't prove it from outside
Every fingerprint below can be manufactured by a broker's own settings. That is why a test gives a likelihood, never a verdict.
delay
Delay plugins / settings — a fixed or random execution delay can be added on the server or in the bridge. It looks like "LP latency" but doesn't react to the market.
slippage
Slippage plugins — price can be shifted by a configured amount, often only against you. It looks like market slippage but is the same size in quiet and fast markets.
size
Size-based pricing — worse prices as order size rises, by a schedule rather than by real depth. A genuine A-book also degrades with size, so size alone proves nothing.
bridge
B-booking inside the bridge — the bridge's risk module can internalise an order. The MT5 server just receives an ordinary fill, so even a trader who checks the server's behaviour sees nothing unusual.
switching
Book switching — the same account can be B-booked today and A-booked tomorrow, or split trade by trade. Brokers often move consistently profitable traders to A-book, so a test run on a new account can describe a different book from the one a proven winner gets.
A-book vs B-book — the fingerprints
Tendencies, not rules. Read each row as "what would make me more or less suspicious", never as a verdict on its own.
Signal
Genuine A-book tends to look like
B-book (or simulated) tends to look like
Execution time ORDER_TIME_SETUP_MSC → ORDER_TIME_DONE_MSC
Varies with the market. Round trips to the LP add milliseconds; in fast markets the tail grows because rejected orders are retried at another LP (each retry is another round trip).
Either very uniform, or a fixed/random added delay that looks the same in every condition. A server-side fill can also be faster than any LP round trip.
Slippage
Two-sided in a fair feed — some positive, some negative — and wider in news, rollover and thin crosses.
Either none (fills at the requested price) or one-directional, similar-sized in every condition. Zero positive slippage over a large sample is a flag — but see "what results can show" below.
Large orders
Price degrades and partial fills appear as size exceeds what the LPs show at the top of book — and how much depends on the moment.
Full fill at the quote regardless of size, or price worsening on a size schedule that doesn't depend on the moment.
Rejections
LP rejects (last look) bunch up in fast markets — "Request rejected" (10006), "Prices changed" (10020), "No quotes to process the request" (10021) — and the bridge may retry elsewhere.
Rare, or rule-based: triggered by size, symbol, a news window or a pattern, not by the market state.
External ID ORDER_EXTERNAL_ID · DEAL_EXTERNAL_ID
Often holds a bridge or LP reference.
Often empty on server-side fills.
Requotes
Execution-mode dependent, not book dependent. A requote (10004) can only happen in Request or Instant execution mode, on either book. Under Market execution there are no requotes on either book — you get a fill, a slippage or a reject. A requote tells you the account's execution mode, not its book.
External IDs are setup dependent: MQL5 documents the field only as an identifier in an external trading system, so whether it's populated depends on the bridge and server configuration. An empty ID doesn't prove B-book, and a populated one doesn't prove A-book. Sources: MQL5 · trade return codes · MQL5 · execution modes · MQL5 · order properties.
The key tell. Genuine A-book latency and slippage react to market conditions — news releases, the 17:00 New York rollover, thin crosses, large size. Simulated delay and slippage stay statistically the same in every condition. So the question to ask of a data set isn't "is it fast?" or "is it negative?" — it's "does it change when the market does?" A flat distribution across quiet hours, session opens, news and rollover is the single most informative thing a test can show.
How to test it
A paired test removes most of the noise: the market is the same for both legs, so differences belong to the account, not the moment.
1 · accounts
Two accounts of the same account type (same execution mode, commission model and symbol set) — typically one at the broker under test and one at a reference broker, or two logins at the same broker if you want per-account treatment. Anything that differs between them has to be known, or the comparison is polluted.
2 · sending
Identical orders, sent simultaneously by one EA from one VPS, so network path and clock are shared. Close each trade quickly and flat; don't hold across rollover unless that's the bucket you're measuring.
3 · sample
100+ trades as a floor, several hundred if you can, spread across buckets: quiet hours, London / New York opens, high-impact news, the 17:00 New York rollover, and small / medium / large size. Few trades per bucket gives noise, not evidence.
4 · log
For every order: requested price, filled price, ORDER_TIME_SETUP_MSC → ORDER_TIME_DONE_MSC, the client-side round trip (clock before send → result received), ORDER_EXTERNAL_ID / DEAL_EXTERNAL_ID, the retcode, volume, symbol, spread at send and the account's execution mode.
5 · hedge
If you hedge the test positions, do it at a third broker. A hedge at the same broker can be netted internally and corrupt the result, and cross-account hedging may breach the broker's T&Cs.
6 · no demo
Demo accounts can't be used for this — demo servers simulate every fill, so they prove your tooling works, not how a broker routes. The test needs live accounts.
7 · analyse
Compare distributions per bucket, not averages: latency percentiles, slippage sign and size, reject rate. Then ask the key-tell question — did anything move between quiet and news?
What the results can — and can't — show
No positive slippage over a large sampleprice improvement is being kept somewhere on the route — broker settings or a one-sided last look — on either book
Flat latency and slippage in every bucketlooks like a setting, not a market — a strong hint, still not proof
Latency and slippage that swell in news and rolloverconsistent with real LP execution — but a good B-book can imitate it
Looks like B-book, but it's an A-bookpossible: an LP applying last look asymmetrically skews outcomes against you even on a genuine A-book
Real confirmationthe broker's bridge reports, LP confirmations and routing settings — nothing client-side
Does the broker act as principal?its order execution policy says — read it, and ask for the answer in writing
Treat the output as a likelihood. A test can make "this is a genuine A-book" more or less plausible; it cannot confirm it. The only proof is documents from the other side of the glass — which is why the due-diligence checklist in Tab 10 asks for execution reports on specific test tickets. This is general information, not legal or financial advice.
A retail broker has several revenue taps, and for a B-book-heavy broker the biggest is you losing. The mix depends on how much flow it B-books (keeps in-house) versus A-books (hedges out): an A-book / STP broker lives on spread + commission, a B-book-heavy broker on internalised client losses.
Retail accounts that lose
74–89%
ESMA 2018 analysis · broker-disclosed
Biggest tap (B-book-heavy)
B-book
internalised client losses
Spread vs tier-1 floor
≈6×
illustrative · see Tab 01
Held positions bleed
nightly
swaps — see Tab 08
The revenue taps
Illustrative, not measured: a hypothetical share of a retail broker's gross revenue, split by mechanism. The exact split is the business model — a pure A-book / STP broker lives on spread + commission; a B-book-heavy broker lives on client losses. Note: B-booked trades also earn the spread-markup slice above, so the B-book's true contribution to revenue is higher than the client-losses bar alone.
Why losing clients are the product
On B-booked flow the broker is your counterparty, so your loss is its revenue directly — no spread required. It leans on the published statistic that most retail accounts lose over time.
You trade a B-booked accountbroker is the counterparty
You lose $1,000+$1,000 to the broker's book
You win $1,000−$1,000 from the broker's book
Across a large book of clients74–89% lose (ESMA, 2018) → positive expectancy
Sharp / toxic winnersoften hedged out to A-book, or restricted (see Tab 02, Tab 05)
The acquisition machine — why they spend so hard to win you
A B-booked loser has high lifetime value — so brokers pay heavily to acquire you, and the marketing you can't escape is just that economics running downstream.
Introducing Broker (IB) rebatea share of the spread / commission on every lot you trade, paid out of the broker's own take — for as long as you trade
Affiliate / CPA deala flat bounty per funded account delivered (typically hundreds of dollars; varies widely)
Deposit bonus"credit" you can't withdraw until you trade many times its value (banned onshore in the EU, UK and Australia)
Why the spend is rationalon B-booked flow your losses are the revenue — acquisition pays for itself
The funnel feeds the book. Aggressive marketing, IB armies and bonus offers all make sense the moment you see the model: the broker is your counterparty, the average acquired client is net-negative, so every dollar spent winning you is recovered from the same losses the B-book is built on. The bonus that "boosts" your account usually just locks you into the trading volume that grinds it down.
The structural conflict: when your broker B-books you it profits when you lose and pays when you win — it is on the other side of your trade, not a neutral venue. That isn't necessarily fraud (it's a disclosed, regulated model), but on that flow "your broker" and "your interests" are opposed. Always know who your counterparty actually is.
The spread is the cost you see. It's rarely the cost you pay. A held position bleeds through several separate channels — and the one nobody quotes you, the overnight swap, compounds every night you hold.
The all-in cost of a round trip
Illustrative, not measured: 1 standard lot EUR/USD ($100k notional), standard account, held 5 nights. Your real numbers vary by broker, account type and instrument.
Spread — 1.2 pips × $10/pip$12.00
Commission (a raw account would add)$0 standard / ~$7 raw
Hold past the daily rollover (usually 5pm New York) and you pay or receive a swap: the interest-rate difference between the two currencies, plus the broker's markup. Charged every night — and, for most FX pairs, tripled on Wednesday to cover the weekend's value date (other instruments differ).
Where it comes fromrate differential of the pair ± markup
Long the higher-yielding currencyyou may receive (positive carry)
Short it / hold the low-yielderyou pay (negative carry)
The broker's markupoften makes both sides net negative
Triple-swap day (usually Wednesday for FX)3× the nightly charge
Why it matters for holding strategies: a swing or carry position held for weeks can pay more in swaps than in spread — and a strategy that's break-even before financing is a loser after it. Price the swap before you size the trade.
Beyond FX — financing on index, share & crypto CFDs
FX charges a swap. A CFD on an index, share or crypto charges overnight financing instead: a daily rate on the full position value — not your margin. Because it scales with notional, it's the cost that quietly dwarfs the spread on anything you hold.
How it's charged(benchmark rate + broker markup) × full notional ÷ 360 (365 for some currencies), every night
$2,000 margin → $100k index CFD at 1:50 (offshore-style leverage; EU/UK/AU cap indices at 20:1)financing on the $100k, not the $2k
~8%/yr all-in (illustrative rate + markup)≈ $22 per night — on a $2,000 stake
Hold it a month≈ $660 — a third of your deposit, before the price moves
Short positionsmay receive or pay, but the markup usually tilts it against you
The leverage trap in reverse: leverage lets a small deposit hold a large position — and financing is billed on that large position. The same 1:50 that makes the trade feel cheap to open makes it expensive to hold. Day-trade and you barely see it; hold for weeks and financing, not the spread, becomes your biggest cost.
Tab 05 showed how the LP and broker grade and kill profitable flow. So which edges actually survive that machine? Rule of thumb: the faster and more mechanical the edge, the sooner the mark-out flags it and the ladder reaches your login. Durable edges look less sharp.
Edge types by kill speed
How quickly each tends to get throttled, skewed, rejected or banned. Not advice, and not measured data — a heuristic map of where the broker's incentives point.
dies fastest
Latency / stale-quote arb — mechanical, one-sided, screaming mark-out. Often flagged within days or less; commonly met with A-booking, delay or restriction.
often banned
Two-broker locking / hedge arb — opposite positions across brokers to harvest swap or bonus. Commonly prohibited in T&Cs; profits can be voided.
throttled
News-spike trading — bursty and obvious around NFP / CPI. Slippage, rejects (and requotes, in Request / Instant execution modes) and news-window blocks can get applied.
slower to flag
Swap / carry arb — the edge accrues quietly over nights, not ticks. Harder to mark out, but capped by swap markups and dynamic leverage.
survives longest
Diffuse directional skill — modest hit-rate, held over time, spread across venues and accounts. Looks like ordinary flow on a mark-out curve.
What "durable" actually looks like
Mark-out signatureflat / noisy, not a clean one-way ramp
Hold timelong enough not to scream "latency"
Footprint per accountsmall relative to the broker's book
Distributionspread across brokers, accounts, symbols
Why spreading thin defeats the mark-out
The same profitable edge, booked two ways. Concentrated in one broker it draws a clean toxic mark-out and gets killed; diffused across many it dissolves into each book's noise and survives. This is the structural problem — and why a distributed edge tends to last longer.
The uncomfortable truth: the cleanest, fastest, most "obvious" edges are the ones broker risk systems are built to detect. Edges that look diffuse across venues and time tend to survive longer — though how much any one broker does about it varies, and none of this replaces checking what a given broker actually does (Tab 06, Tab 10).
The case: a vendor whose MT5 EA runs inside its clients' own broker accounts, and who partners with a broker — an introducing-broker (IB) deal, or a white-label arrangement where the vendor's brand sits on the broker's platform and licence. The vendor doesn't control execution; the broker does. This tab is what to negotiate, where it lives in MT5, where the incentives collide, and what to verify before a single client is onboarded. It is broker-neutral — it describes the structure, not any one firm.
What's negotiable — and where it lives
Almost everything that shapes an algo's results is a setting the broker controls. Knowing where each one lives tells you what to ask for in writing.
Lever
Where it lives
What to ask for / watch
Spread markup
MT5 account group — a markup on top of the bridge price, per symbol
The group's markup per symbol, in writing. Two accounts in the same group see the same prices.
Commission
Account group (per lot / per million, round turn or per side)
All-in cost = spread + commission, not either alone.
IB rebate
Group / IB agreement
Paid out of the broker's spread + commission — see "rebate economics" below.
Swap and leverage
Account group, per symbol
Swap markup and leverage tiers; matters for any strategy that holds overnight.
Execution mode
Symbol / group setting
Market execution gives no requotes (fill, slippage or reject instead); Request/Instant can requote (10004). Know which the EA's accounts get.
Server location
Where the MT5 server and bridge are hosted
Servers in major FX data centres (e.g. Equinix NY4 / LD4) alongside the bridge, and a VPS in the same data centre, cut round-trip time legitimately.
Symbol settings
Per-symbol contract spec
Stops level, freeze level, filling modes, minimum lot step — an EA that places tight stops can be broken by a stops level it never checked.
Account structure
IB vs white-label agreement
Who holds the licence and the client money, who is the counterparty, and who can change the group settings after launch.
"Adjusting latency" — legitimate, or a red flag?
The phrase can mean two very different things. Ask which one a broker means.
Legitimate
Server and VPS location — co-hosting the VPS in the same data centre as the broker's server and bridge. Fewer hops — a faster bridge, a direct cross-connect, an LP closer to the bridge. Measurable — you can see it as a lower round trip on every order, in every condition.
Red flag
A configurable execution delay on a group or account — a number of milliseconds someone can set. A configurable slippage setting that only moves against the client. Different delay for different clients in the same group, set by a risk rule. Not volunteered — if the answer to "do you apply any execution delay or slippage settings?" is evasive, dig further before going on.
The conflict of interest
If a broker B-books a consistently winning algo, every winning trade is the broker's loss — it is the other side. A broker that is paid when the algo trades but loses when the algo wins has two incentives pulling in opposite directions.
spreads
Wider spreads — the group markup for the EA's accounts creeps up after profits show.
delay
Added delay — an execution delay set on the account or group, so fast entries miss.
slippage
Negative-only slippage — fills drift against you and never for you (see Tab 06).
limits
Account restrictions — lot caps, EA bans, "abusive trading" clauses invoked.
money
Withdrawal friction — payouts stalled, then reviewed against the T&Cs.
best case
Flow moved to A-book — the broker hedges it out, so a win isn't a loss to the broker. This is the outcome to negotiate for in writing, because a broker that A-books the flow no longer has a reason to degrade it.
Rebate economics
What the client paysspread markup + commission, both set in the MT5 group
What the IB rebate isa share of that — it comes out of the spread / commission either way
So a bigger rebate usually meansa wider spread or higher commission, or a thinner broker margin
If the broker B-books the flowthe rebate is paid by a counterparty who profits when the clients lose and loses when they win
What to compareall-in cost per lot (spread + commission, net of rebate) against an equivalent direct account
Due-diligence checklist
Before a vendor sends a single client to a broker. Each item is something you can ask for or check.
1 · regulator
Verify the licence on the regulator's own register, not on the broker's footer. Offshore registrations such as St. Vincent & the Grenadines' FSA do not regulate forex — the FSA registers companies and says it doesn't license forex business. A Cayman "registered person" is a lighter status than a full CIMA licence (it is aimed at sophisticated / high-net-worth or intra-group business), so check which one it is.
2 · entity
Which legal entity holds client money and is the counterparty? Is the money segregated, where, and at which bank? Match the entity on the account agreement to the one on the register.
3 · policy
Read the order execution policy. Does the broker act as principal? Does it use last look, execution delay or slippage settings? Ask for the answer in writing, not on a call.
4 · bridge + LPs
Ask for the bridge and the LPs by name (or at least by type: PoP, bank, ECN), and for the server and bridge data centre.
5 · evidence
Request bridge / LP execution reports for specific test tickets — time, LP, fill price. Pick the tickets yourself, after the trades. A broker that A-books genuinely can produce these.
6 · T&Cs
Read the clauses on "latency arbitrage", "abusive trading" and EAs, and on the right to void profits. Get any concessions written into the partner agreement.
7 · withdrawal
Run a test withdrawal early — and again once the account has made money. Friction shows up at the second one, not the first.
8 · test
Run the paired test from Tab 06 on live accounts before onboarding clients, and again after profits have accumulated.
What a vendor can and can't rely on. A broker's assurance that it is "A-book" is a claim, not evidence — it is checkable only through reports and confirmations the broker holds. Put the commitment in the agreement, keep the right to audit the execution on test tickets, and plan for the day the flow gets moved or restricted. This tab is general information and names no firm; it is not legal or financial advice.
Many brokers are several companies wearing one brand. The regulated onshore entity runs the marketing and the trust badges; a separate offshore entity holds your account, your money and the 1:2000 leverage. Same logo — very different rulebook.
Same broker, two rulebooks
What changes when your account sits with the onshore entity (FCA, ASIC, CySEC / ESMA) versus an offshore one (St. Vincent, Seychelles, Mauritius, Vanuatu).
Onshore
Max leverage — 1:30 (ESMA / FCA / ASIC retail). Negative-balance protection — mandatory. Client funds — segregated and audited. Compensation — e.g. FSCS up to £85k if the broker fails. Recourse — a real regulator + ombudsman. The cost — lower leverage, more friction.
Offshore
Max leverage — 1:500 to 1:2000+. Negative-balance protection — usually none. Client funds — often not segregated. Compensation — none. Recourse — minimal; the regulator may not cover forex at all. The cost — your protection.
The entity shell game
You see the brand & its FCA licencethe trust signal
You sign up for 1:1000 leverageonboarded to the offshore entity
Your money is actually held byan SVG / Seychelles company
It goes wrong → you call the FCA"that entity isn't ours"
"Registered" is not "regulated"
St. Vincent & the Grenadines (SVG) FSAregisters companies; it says it does not license or regulate forex business
Cayman "registered person" (CIMA)a lighter status than a full SIBA licence — aimed at sophisticated / high-net-worth or intra-group business
How to check any of itthe regulator's own register, by licence number — not the broker's footer
Check the entity, not the logo. The leverage you were sold is a strong hint: 1:30 is regulated retail; 1:500–2000 usually means an offshore entity that typically carries no negative-balance protection, no segregation guarantee and no compensation scheme. (Onshore professional-client accounts can also get higher leverage — by giving up the retail protections.) The footer of the broker's own site lists which company actually holds your account — read it before you deposit.
The United States — the market that's "banned" and trades anyway
The clearest case of the offshore split: an entire country pushed off the regulated grid. Retail CFDs — gold, indices, stocks, crypto — are effectively prohibited for US persons; the SEC / CFTC don't allow off-exchange retail CFDs at all. Spot forex is allowed, but only through an NFA-registered, CFTC-regulated dealer, capped at 50:1 (majors) / 20:1 (minors) under the CFTC / NFA retail-forex rules. That regime pushed many foreign brokers out — most just stopped accepting Americans rather than register.
A US trader wants 1:500 gold CFDsno onshore broker can legally offer them
So they open an offshore accountSVG · Seychelles · Vanuatu · Belize
The broker's T&Cs say "no US clients"enforcement at sign-up varies
It goes wrong → they call the CFTC"that firm isn't registered with us — we can't help"
The trade isn't the crime — the exposure is. US persons trading offshore CFDs are stepping around the rules; the regulatory focus is generally on the firm that solicits them rather than the customer (this is not legal advice, and tax rules still apply). But they've stepped entirely outside the system built to protect them — no NFA recourse, no fund segregation, no compensation scheme. If the money disappears, no US regulator has a lever to pull.
Every other tab is a cross-section of the machine. This one runs a single trade straight through it — one EUR/USD long, from the quote you see to the moment your winning streak becomes a problem. Each step links back to the tab that explains it.
One EUR/USD long, end to end
01 · the quote
You see EUR/USD 1.08000 / 1.08012 — a 1.2-pip spread. The interbank floor is ~0.2 pip; the rest is PoP + broker markup stacked down the chain. → Tab 01, Liquidity Chain.
02 · the routing
You click buy 1 lot. The broker's rules classify you: small deposit, no toxic history → likely B-booked (on the server or inside the bridge — your ticket won't say which). The broker is now your counterparty; nothing touches a real venue. → Tab 02, the A/B-book desk.
03 · the margin
Your $3,600 margin holds a $108,000 notional position at 1:30. The leverage is real — and so is the stop-out level sitting just below you. → Tab 04, Margin.
04 · the carry
You hold 3 nights. Each night a swap is debited — tripled on Wednesday — so you're down ~$13.50 (an illustrative ~$4.50 a night) in financing before the price has done anything. → Tab 08, What It Costs.
05 · the win
EUR/USD runs your way; you're +$800. On the B-book that's −$800 to the broker. The risk desk's mark-out flags your win-rate and hold pattern. → Tab 05 + Tab 02 oversight.
06 · the reclassify
You may be moved to A-book — hedged out — and your login tagged (or restricted instead). One common reward for winning is to be managed as toxic flow. → Tab 02 + Tab 09, the edge that survives.
07 · the exit
You close the winner and meet slippage on the way out; the next clean signal may get a requote — but only if the account is in Request or Instant execution mode (under Market execution it would be a reject or more slippage instead). The withdrawal clears — but the account is now on a watchlist. → Tab 05, what rolls downhill.
What the trade actually cost
The cost you sawthe 1.2-pip spread — ~$12
Spread + 3 nights swap + slippage≈ $30–40 — roughly three times what you saw
The moment you started winningyou stopped being a customer and became a risk to manage
The whole machine in one trade: the spread is marked up before you click (six-fold in the illustrative stack of Tab 01); you're likely warehoused because you're expected to lose; you bleed financing just for holding; and once your edge shows up on a mark-out curve, a broker that was happy to be on the other side of your trades may hedge you out, or restrict you. None of it is hidden — it's just spread across the tabs. This is the one that connects them.
Retail CFDs are a regulated product, and the rulebook changes at every border. The same instrument that's capped at 30:1 with guaranteed loss limits in one country is effectively banned in another, or sold at 500:1 with no safety net in a third. This is the map — who regulates it, how much leverage retail clients can get, what protection comes with it, and the link to the actual rule. Strictest at the bottom.
The retail-leverage map
Each card is the onshore, licensed rulebook for that country. The offshore arm of the same broker (Tab 11) is bound by none of it — that's the entire point of the shell.
🇪🇺 European Union — ESMA + national NCAs
The EU-wide floor every other strict regime copies. ESMA's 2018 product-intervention measures were made permanent by each national regulator (BaFin, AMF, CySEC…). Retail CFD leverage is capped by asset: 30:1 major FX · 20:1 minor FX / gold / major indices · 10:1 other commodities & minor indices · 5:1 single shares · 2:1 crypto. Mandatory negative-balance protection, 50% margin close-out, no trading bonuses.
Mirrors the ESMA bands and made them permanent from Aug 2019 (PS19/18): 30:1 → 2:1 by asset class, mandatory negative-balance protection, 50% margin close-out. Client money must be segregated, and the FSCS compensates eligible clients up to £85,000 if an FCA-authorised broker fails.
Product Intervention Order in force since 29 Mar 2021, extended five years in 2022. Same banding as the EU: 30:1 major FX · 20:1 minor FX / gold / major index · 10:1 other commodities · 5:1 shares · 2:1 crypto. Mandatory negative-balance protection and margin close-out; binary options banned outright.
CFDs and leveraged FX are regulated capital-markets products under the Securities and Futures Act; the broker must hold a CMS licence. Retail leverage is capped indirectly through margin requirements — a 5% minimum margin (≈20:1) on FX CFDs for retail clients, versus 2% (≈50:1) for accredited / institutional investors, who get fewer protections. Other asset classes differ — check MAS's current schedule.
No single statutory leverage cap — limits are set by margin rates per instrument (CIRO, formerly IIROC; provincial CSA oversight). In practice around 50:1 on major FX pairs and lower elsewhere — check CIRO's current margin schedule. Crypto CFDs are effectively off-limits to retail. Member dealers are generally covered by the CIPF, which protects eligible client property (cash, securities) if a dealer fails — it doesn't cover trading losses, and how it applies to leveraged FX / CFD positions should be checked directly.
Derivatives issuers must be FMA-licensed under the FMC Act 2013, but no hard leverage cap currently applies — some issuers cap CFDs at 30:1, others advertise up to 500:1. The FMA consulted in 2024 on limits of 30:1 / 20:1 / 10:1 / 2:1 by asset class and has signalled caps are likely, but its standard licence conditions still contain no leverage limit as of the latest check (September 2026).
The strictest regime of all. Off-exchange retail CFDs are effectively prohibited under the Dodd-Frank Act — they may only trade on a registered exchange. Spot forex is allowed only through an NFA-registered FCM/RFED, capped at 50:1 majors / 20:1 minors. Most offshore brokers simply refuse US clients rather than register.
Read the entity, not the badge. Every rule above binds only the licensed entity in that country. The leverage you were offered is the tell: 1:30 is an onshore, regulated book; 1:500–2000 means your account sits with an offshore arm bound by none of these — no leverage cap, usually no negative-balance protection, no segregation guarantee, no compensation scheme. Figures reflect a review in October 2026 and change often — check the regulator's own page before you deposit. The same broker can also offer higher leverage to clients classed as professional. This is general information, not legal or financial advice.
Plain-English definitions for every term used in the other tabs. No prior knowledge assumed.
Units & pricing
Pip
The smallest standard step a price moves. For EUR/USD it's the 4th decimal — 0.0001. Profit and loss are usually counted in pips.
Lot
A standard trade size. 1 standard lot of a currency pair = 100,000 units of the first currency (mini = 10,000, micro = 1,000). For gold (XAUUSD), 1 lot = 100 troy ounces.
Notional value
The full market value of what you control — not what you put down.
1 lot of EUR/USD at 1.08 ≈ $108,000 notional, even if your deposit was only a few hundred dollars.
Leverage
How much notional you control per dollar of your own money. Amplifies gains and losses equally.
The gap between the price you can sell at (bid) and buy at (ask). It's the built-in fee on every trade — on most accounts the broker's markup (set in its MT5 account group) is part of it.
Yard
Trader slang for one billion (from the French "milliard"). Brokers report monthly turnover in yards.
"500 yards/month" = $500 billion of notional traded.
ECN / raw-spread account
An account type that passes near-raw market prices through — tiny or zero spread — and charges a separate commission, instead of widening the spread to make money.
Swap / rollover
A daily charge or credit for holding a position overnight — the interest-rate difference between the two currencies, plus the broker's markup. Tripled on Wednesday to cover the weekend value date.
Carry
The yield you earn (or pay) simply for holding a position, from the rate differential between its two currencies. Positive carry pays you to hold; negative carry bleeds you.
Inactivity fee
A monthly charge some brokers levy on dormant accounts — a non-trading fee that quietly erodes a balance you've stopped watching.
Overnight financing (CFD)
The daily charge for holding a non-FX CFD (index, share, crypto) past rollover — a benchmark interest rate plus the broker's markup, applied to the position's full notional, not your margin. The CFD equivalent of a swap.
Market structure — who's who
OTC (over-the-counter)
A private, bilateral deal between two parties — not traded on a public exchange. Most FX and all retail CFDs are OTC.
Exchange
A central, regulated venue where many participants trade standardised contracts (e.g. COMEX for gold futures). The opposite of OTC.
CFD (Contract for Difference)
A cash-settled bet on a price moving. You never own the euro or the gold — you hold a contract with your broker that tracks its price.
Liquidity Provider (LP)
Any firm that streams live buy/sell prices a broker can trade on — a role, not a tier. Tier-1 banks, non-bank market makers, MTF/ECN venues and prime-of-primes are all LPs, at different depths. A retail broker's own LP is usually a PoP.
Prime of Prime (PoP)
A "broker's broker": it packages big-bank liquidity and credit and resells it to retail brokers too small to face the banks directly. A PoP is itself an LP to the broker — one sitting on top of deeper LPs. Examples: Finalto, IS Prime, iSAM Securities.
Prime Broker (PB)
The prime-brokerage service a tier-1 bank sells: it lends its balance sheet and credit so a client (a PoP or a fund) can trade the interbank in the bank's name and settle later (give-up). It's a service, not a separate tier — the PB is the same tier-1 bank, wearing a credit hat.
Tier-1
The top banks and non-bank market makers (JPMorgan, UBS, Citi, Deutsche, XTX, Citadel Securities) where the real underlying price is made. These same banks also run the prime-broker desks that bankroll the PoPs.
Interbank
The wholesale market where tier-1 banks and non-bank market makers trade currencies with each other — a decentralised network of bilateral relationships and primary venues (EBS, LSEG Matching), not one central exchange. Where the real FX price is made.
STP (straight-through processing)
Passing a client order straight out to an LP for execution, with no dealing-desk intervention — the A-book ideal. The opposite of a B-book that holds the trade in-house. "STP" is also a marketing label: it doesn't prove a broker has no dealing desk or never B-books.
Spot / spot FX
A trade for near-immediate settlement at the current price — the everyday FX price, as opposed to a future or forward. Retail CFDs reference spot.
MTF venue
A regulated, multi-participant marketplace with a central order book and firm, no-"last-look" pricing (e.g. LMAX Exchange) — an ECN-style venue. A different kind of LP from a PoP: a neutral marketplace, not a credit-reselling intermediary.
Bridge
Software (oneZero, PrimeXM, Gold-i, Centroid) that connects a broker's platform to its LPs and decides where each order goes. Many bridges include risk / internalisation modules, so B-booking can happen inside the bridge as well as on the MT5 server.
MT4 / MT5
MetaTrader 4 and 5 — the dominant retail trading platforms; the screen most retail traders actually use.
Market maker (MM)
A firm that continuously quotes both a buy and a sell price and stands ready to trade either side, earning the spread. Banks and non-banks (XTX, Citadel) make markets in FX; the broker itself makes the market to you.
Single-dealer platform (SDP)
One bank's own pricing portal — its proprietary stream of quotes, not a multi-dealer venue. Brokers tap several SDPs to build a price.
Loco London
Gold quoted for settlement as bullion held in London vaults — the OTC benchmark location for spot gold, run through the LBMA.
MT5 & execution
Execution mode
The MT5 setting that decides how a market order is priced. Request: the price is requested first and the broker confirms or rejects. Instant: the order is sent at the price shown and the broker accepts it or requotes. Market: the broker fills at its own price with no requote; slippage is possible. Set per symbol / group — it says nothing about A- vs B-book.
Requote
A broker's counter-offer at a new price when the price you asked for is no longer available (MT5 retcode 10004). Only possible in Request and Instant execution — never under Market execution — and on either book. A requote shows the execution mode, not the book.
Retcode
The numeric result an MT5 EA gets back for a trade request: 10004 "Requote", 10006 "Request rejected", 10020 "Prices changed", 10021 "No quotes to process the request". ("Off quotes" is MT4 wording.)
Account group
An MT5 server setting that bundles an account's trading conditions: spread markup over the bridge price, commission, IB rebate, swaps, leverage, margin-call / stop-out levels and execution mode. Two accounts in the same group see the same prices whichever book they are on.
External ID
MT5 order and deal fields (ORDER_EXTERNAL_ID / DEAL_EXTERNAL_ID) documented as an identifier in an external trading system. A bridge may fill them with an LP reference — whether it does is setup dependent, so an empty field doesn't prove B-book.
Demo account
A practice account on a demo server. Demo servers simulate every fill — they are never A- or B-booked — so they can't show how a broker really routes orders.
Order execution policy
The document in which a broker describes how it executes client orders — including, where it's required, whether it acts as principal (your counterparty) or passes orders on to third parties.
Booking & risk
A-book
The broker passes (hedges) your trade out to a real LP — often as a net position across many clients. It earns its markup / commission and carries little market risk, so it has no direct stake in whether you win (it still bears execution and credit risk).
B-book
The broker keeps your trade in-house and becomes your counterparty — on the MT5 server or inside the bridge. You lose → it keeps the money; you win → it pays. Profitable if most retail flow loses, but risky if flow is one-way.
C-book (hybrid)
A middle setting between A and B: the broker warehouses a set ratio of a client's flow and hedges the rest, dialling its risk by client grade instead of routing each trade fully one way or the other.
Internalisation
Matching opposing client trades against each other (or warehousing them) instead of sending them to the market. Can happen on the MT5 server or inside the bridge; brokers often hedge only the leftover net exposure.
Sharp / toxic flow
Clients who consistently win — often arbitrage or fast news traders. Brokers want this hedged out of the B-book quickly.
Give-up
An institutional setup where a trade is executed with one party but "given up" to your prime broker to clear and hold.
Slippage
Being filled at a different price from the one requested — in fast or thin markets, or because of broker settings. It can be negative or positive; slippage that is only ever negative across a large sample is a flag.
Hedge
Taking an opposite, offsetting position so a price move can't hurt you. When a broker A-books, it places an offsetting trade with an LP (often for the net of many clients) — its risk largely cancels out and it keeps the markup.
Counterparty
The party on the other side of a trade. In a B-book the broker is your counterparty: when you lose it gains, and when you win it pays.
Netting
Combining many offsetting trades into one net figure, so only the difference is settled or hedged — not every individual ticket.
VaR (Value at Risk)
A standard risk gauge: the most a book is expected to lose over a set period, at a given confidence level.
"VaR 95% / 1d = $1.9M" → on a normal day it shouldn't lose more than $1.9M, ~19 days out of 20.
Delta / net delta
The size and direction of a position's exposure — how much it gains or loses as the price moves.
"Net short $36M" = the broker profits if prices fall, loses if they rise.
Mark-to-market
Revaluing an open position at the current market price to show its running profit or loss before it's actually closed.
Mark-out
The LP's profit or loss on a fill, measured at fixed times after it (+1s, +5s, +1m…). Persistently negative mark-out means the flow is picking the LP off — the core toxicity gauge.
Last look
A brief window in which an LP can accept or reject a trade after the broker hits its quote — a price check meant to avoid being filled on stale prices. Symmetric checks reject moves in either direction; some LPs apply it asymmetrically (reject when price moves against the LP, accept when it moves in its favour), which skews outcomes against the trader even on a genuine A-book. Extra latency in fast markets comes mostly from rejects and retries to other LPs.
Fill ratio
The share of a broker's orders an LP actually fills. LPs cut it for toxic flow — more rejects and partial fills.
Flow tagging / segmentation
Splitting a broker's order flow into labelled streams (by client profile) so an LP can price, or refuse, the toxic part separately from the rest.
Locking / hedge arbitrage
Holding offsetting long and short positions — often across two brokers — to harvest a swap, bonus or pricing gap rather than a market move. Explicitly banned in most retail T&Cs.
Skew
Tilting quotes or risk to one side — e.g. a desk that's net-long shades its price to attract sellers, or an LP widens one side against flow it expects to lose to.
Margin & collateral
Margin
The deposit you post to open and hold a leveraged trade. Initial margin opens it; maintenance margin keeps it open.
Margin level
Your equity divided by the margin in use, as a %. The number the broker watches.
Equity $2,000 ÷ used margin $360 = 556%.
Margin call
A warning (often at 100% margin level) that your equity is getting too thin to support your open positions.
Stop-out
The broker automatically closes positions (often at 50% margin level) so your loss can't exceed your deposit.
Negative-balance protection
A rule (mandatory for EU/UK/AU retail) that you can't lose more than you deposited.
Collateral / variation margin
What an institution posts to its LP/PB to back open positions — topped up as positions move against it (that top-up is variation margin).
Dynamic leverage
Offshore brokers tie max leverage to account equity, lowering it as the balance grows. Small accounts may get 1:2000+; larger ones are stepped down to 1:500, 1:200 or less — limiting the broker's tail risk on big positions.
$500 account → 1:2000; same client at $50k → maybe 1:200.
Credit line
Permission to trade now and settle later, up to a limit, instead of paying the full value upfront. PoPs and prime brokers extend credit lines down the chain so each layer can trade on collateral rather than cash.
Segregated funds
Client money kept in a separate bank account from the broker's own, so it can't be used as working capital and is protected if the broker fails. Mandatory onshore; often absent offshore.
Compensation scheme (FSCS)
A statutory backstop that repays clients up to a cap (e.g. £85k for investment claims under the UK's FSCS) if a regulated broker becomes insolvent. Offshore entities typically have none.
Data & benchmarks
BIS Triennial Survey
The Bank for International Settlements polls central banks every three years (each April) to measure global FX turnover — the authoritative size of the market.
2025: $9.6 trillion traded per day.
LBMA
London Bullion Market Association — runs the OTC London gold market and sets the twice-daily LBMA Gold Price benchmark.
COMEX
The New York gold & metals futures exchange (part of CME) — the main exchange-traded gold price.
EFP (Exchange for Physical)
A deal that swaps an OTC gold position for a COMEX futures position — the plumbing that keeps London spot and New York futures in line.
Finance Magnates / QIR
An industry news & data provider; its Quarterly Intelligence Report (QIR) publishes the broker volume rankings used in Tab 03.
NFP (Non-Farm Payrolls)
The big monthly US jobs report — a scheduled news event that sharply moves FX prices and draws fast "news" traders the B-book wants to avoid.
CME
Chicago Mercantile Exchange — the largest US futures exchange; runs COMEX (gold & metals) and the listed FX futures that cross-check OTC spot.
CPI (Consumer Price Index)
The monthly inflation report — like NFP, a scheduled news event that jolts FX prices and draws fast news traders.
Regulation & jurisdiction
CFTC
Commodity Futures Trading Commission — the US federal regulator for futures, options and retail forex. Off-exchange retail CFDs fall outside what it permits, so they can't be lawfully offered to US persons.
NFA
National Futures Association — the CFTC-overseen self-regulatory body that registers and supervises US forex/futures dealers. A US-facing retail forex broker must be an NFA member; offshore brokers taking US clients are not.
Dodd-Frank Act
The 2010 US financial-reform law. For retail forex it brought NFA/CFTC registration, and the CFTC / NFA rules cap leverage at 50:1 (majors) / 20:1 (minors) — terms strict enough that many foreign brokers stopped accepting Americans rather than comply.
FCA
Financial Conduct Authority — the UK regulator. Caps retail CFD leverage at 30:1 down to 2:1, mandates negative-balance protection, and backs clients with the FSCS. Among the strictest retail regimes.
ESMA
European Securities and Markets Authority — set the EU-wide retail-CFD floor (30:1→2:1, negative-balance protection, margin close-out) that the UK and Australia closely mirror. Enforced by each national regulator.
ASIC
Australian Securities & Investments Commission — caps retail CFD leverage at 30:1→2:1 (in force since 2021) and bans binary options outright.
CySEC
Cyprus Securities and Exchange Commission — the EU regulator (ESMA rules) that many brokers passport their licence from. Onshore; a CySEC licence can be passported across the EEA.
MAS
Monetary Authority of Singapore — regulates CFDs and leveraged FX under the Securities and Futures Act. Retail FX CFD margin is at least 5% (≈20:1); accredited investors 2% (≈50:1).
CIRO
Canadian Investment Regulatory Organization (formerly IIROC) — sets leverage via margin rates (~50:1 on major FX) rather than a fixed cap; crypto CFDs are effectively barred for retail. CIPF protects eligible client property if a member dealer fails.
FMA
Financial Markets Authority — New Zealand's regulator. Licenses derivatives issuers but (as of 2026) sets no hard leverage cap; a 30:1 standard condition was consulted on in 2024.
SEC
US Securities and Exchange Commission — the federal securities regulator. Together with the CFTC, it's why off-exchange retail CFDs can't be lawfully offered to US persons.
Low-regulation island domiciles where many brokers register the entity that actually holds your account. Typically no leverage cap, no negative-balance protection, no fund segregation and no compensation scheme.
Acquisition & incentives
Introducing Broker (IB)
A partner who refers clients to a broker in exchange for a cut of the spread or commission on every trade those clients make — an ongoing rebate, not a one-off.
CPA / affiliate
Cost Per Acquisition — a flat bounty paid to a marketer for each new client who registers and funds an account, regardless of how they then trade.
Deposit bonus
Promotional credit added to your balance that can't be withdrawn until you trade a large multiple of its value — it incentivises volume (and the costs that come with it), not winning.