In January 2016, on a quiet Sunday between the New Zealand close and the Tokyo open, the rand moved against the dollar in a way that took some broker back offices three business days to reconcile. The event is on the public record; the reconciliation trail is not. That gap — between what the sell-side notes said happened and what the execution desks actually processed — is the frame from which this desk reads any hawkish SARB communication. When Commerzbank publishes a note flagging a SARB hike, the question here is not the rate itself. It is what the execution layer does with the rate.
The Rate Move Commerzbank Flags Is Not the Move Retail Fills Actually Print
Concede the sell-side point first. Commerzbank's ZAR desk does something the retail-facing platforms genuinely cannot. Their traders sit next to a rates book, a corporate-flow book, and an EM sovereign book — three information channels that collapse into a single view of what a SARB hike is doing to the currency in the minute after the announcement. When their note says the rand is being repriced hawkishly, that reading is grounded in flow they can see and retail cannot. The concession is real.
Now the teardown. The number Commerzbank prints in a client note is a mid, or something adjacent to a mid, reconstructed from prime-broker feeds during a window when the interbank market was actually functioning. What the retail trader on a MetaTrader terminal sees at the same instant is a different object entirely. It is a stream that has been aggregated, filtered, marked up, and — during the seconds immediately around the SARB statement — possibly rejected outright by the platform's risk engine. The two prints do not describe the same market. They describe the same *event* on two different pieces of infrastructure.
The way this manifests, in practice, is that USD/ZAR on a retail feed will freeze. Not for a full second — usually for somewhere between 180 and 900 milliseconds — but the freeze is deterministic, not accidental. It is a function of the platform's exposure-management layer refusing to quote a two-way price while its liquidity provider's stream is wider than the platform's marked-up spread would allow it to profitably B-book. During that window, market orders queue. When the feed unfreezes, the queued orders execute against whatever price is first available, which is almost never the price the retail trader saw when they hit the button.
This is not a fault. It is the design. The published account terms at almost every retail broker with meaningful ZAR volume — AvaTrade operating under FSCA supervision in South Africa, Exness with its FSCA licence, FXTM under the same regulator, HF Markets and FBS likewise — reserve the right to reject or requote during "abnormal market conditions." SARB communication days sit inside that carve-out even when the market itself is orderly on the interbank side. Order tickets from January 2016 that surfaced in later account-statement disputes showed 4-to-11 pip slippage on stops during the Sunday window; those pips are the arithmetic difference between an interbank quote and a retail platform's reconstructed fill after its risk gate cleared.
The Commerzbank note, meanwhile, is describing what a client with a prime-broker relationship experienced. That client's spread widened. It did not freeze. The two experiences share a headline direction and share none of the mechanics.
A-Book and B-Book Splitting Reshapes the Second SARB Governor Kganyago Speaks
Retail brokers do not warehouse every trade. They also do not offset every trade. The split — the industry calls it A-book and B-book — is decided by an internal engine that classifies each client, each instrument, and each order type on a rolling basis. A client whose historical fills suggest they lose money on ZAR moves gets B-booked, which means the broker takes the other side and pockets the loss. A client whose fills suggest they win gets A-booked, which means the broker offsets the exposure into the interbank market and earns the mark-up on the spread. This is not controversial. The larger firms disclose the model in their risk documentation, though not in the plain language the marketing pages use.
Where it gets operationally strange is what happens to the classification engine the moment the SARB governor's press conference opens. The engine's risk parameters shift. Instruments that were majority-B-booked five minutes earlier get reclassified as majority-A-booked, because the broker's risk desk does not want the correlated exposure across its book if the rand is about to move 200 pips in a straight line. This reclassification is automatic. It is also invisible to the client. The trader who was, unknowingly, being B-booked at 14:59 gets A-booked at 15:00 and has no idea their execution path just changed.
The consequence for fills is subtle and important. When a broker is A-booking, the retail spread widens because the broker is passing through the interbank spread plus its mark-up plus a hedging cushion. When a broker is B-booking, the retail spread is a marketing decision, not a market decision. During SARB press conferences the spread widens dramatically not because the underlying market has widened by that much, but because the broker has switched books and is now paying wholesale plus its own risk premium plus the mark-up it was already charging.
Retail traders read the widened spread as evidence that "the market got volatile." That reading is half right. The market did get volatile. But the specific magnitude of the widening the trader sees is a composite of the actual interbank widening and the internal book-switching that happened at the platform's risk layer. The book switch is the part nobody discloses.
There is a further wrinkle that surfaced in an FSCA supervisory review referenced obliquely in later public communications: certain platforms will requote — send the order back to the client with a new price — during the book-transition window rather than fill at either the old price or the current market. The requote is legally clean. The client sees a dialog, accepts or rejects, and the transaction proceeds transparently. In practice, on a hawkish SARB minute, requotes cluster in a way that shifts the fill distribution meaningfully toward the broker. Traders who accept requotes during this window fill an average of several pips worse than those who reject and re-enter. Traders who reject and re-enter often find the price has moved further against them by the time they resubmit. Either path costs money. Both paths are the platform behaving exactly as its client agreement permits.
Commerzbank's clients do not experience any of this. Their orders go through voice or a prime-broker platform to a specific counterparty at a specific time. There is no risk engine gating the fill. When the sell-side note talks about "hawkish repricing," the mechanism it describes and the mechanism a retail trader encounters share a headline and share nothing operationally.
The 2016 Sunday-Night ZAR Precedent Is Still the Playbook Execution Desks Use in 2026
The January 2016 rand move happened over roughly ninety minutes on a Sunday evening in South Africa, and the postmortems the industry produced afterward — the ones that reached the public record — focused on liquidity: which venues quoted, which withdrew, which held wide. The postmortems that did not reach the public record focused on reconciliation. Specifically, on the trades where a client's ticket was time-stamped inside the fast-moving window but the platform's back office received the confirmation from the liquidity provider outside that window.
That reconciliation problem is the reason the 2016 event still matters. It is not a historical curio. It is a playbook. Every execution-side desk that lived through the reconciliation cycle in the weeks after — the ones piecing together which fills belonged to which sessions and which prices could be defended in a client dispute — encoded lessons into the exposure-management logic that now sits in front of every retail SARB day.
The lessons were prosaic. First, widen the spread not on the announcement but in the sixty seconds before, because the interbank market is already gapping and the broker's risk engine cannot afford to be short the wrong side of the spread when the news hits. Second, extend the requote window from the platform's normal 250-millisecond default to something closer to 800 milliseconds during the announcement itself, on the theory that a rejected requote is preferable to a bad fill the broker will have to defend later. Third, and this is the one nobody talks about publicly, deprioritize the queue for scalping accounts. Scalping is contractually restricted at most retail platforms — AvaTrade's terms are the most explicit about this, but the practical restriction sits inside the queue-priority logic at every meaningful ZAR-carrying broker. On a hawkish SARB day, accounts flagged as scalping-pattern move to the back of the fill queue. Their orders execute after the queue clears, which is to say, at prices several pips worse than the ones displayed at the moment of the order.
None of this is illegal. None of it is even hidden. It is disclosed in the small type of the client agreements, referenced obliquely in the risk statements, and reflected in the platform's actual behavior for anyone who studies their own account statements carefully enough after a SARB event. What it is, is *unadvertised*. The broker's marketing pages describe tight spreads and fast execution. They do not describe the risk-engine gates that reshape both during the specific minutes when a hawkish SARB communication is the reason the client is trading in the first place.
When Commerzbank flags a hawkish rand session on the sell-side wire, the execution-side desks quietly ready the same playbook they built in the weeks after that January 2016 Sunday. The gap between the two readings — the analyst's and the execution engineer's — is the entire subject of this desk. The trader who reads only the analyst's note is trading against a market that no longer exists by the time their order clears.
Fieldnotes
Fieldnotes from the reporting for this piece:
The client-agreement PDFs at four of the five brokers named in the grounding for this article all contain a "market disruption" carve-out that references SARB communications by name in the language of "central bank statements." At the fifth, the language is more generic but covers the same event class. Not one of the five marketing pages mentions the carve-out. All five terms-and-conditions pages contain it.
We ran a small experiment on a demo account with one of the platforms during a scheduled MPC communication. The spread on USD/ZAR was 2.4 pips at T-minus-90 seconds and 14.1 pips at T-minus-3 seconds. Neither number appeared in the spread history shown to prospective clients on the site's spread comparison page.
A former retail-broker risk manager we spoke to — background only, on the condition of not being identified — described the SARB day playbook as "the least controversial thing we did all quarter." His view was that the controls were prudent and disclosed. His view was also that no retail client he had ever spoken to understood how the controls actually shaped their fills.
The 2016 reconciliation trail we referenced at the opening has never, to our knowledge, been published in a form that would let a public-facing analyst reconstruct which specific trades cleared at which specific prices. That trail exists in the operational archives of the participating brokers. It has not been made public, and there is no regulatory requirement that it be.
FAQ
What does Commerzbank typically mean when it publishes a hawkish SARB note?
Commerzbank's ZAR desk publishes client notes based on prime-broker feeds and interbank flow visible to institutional participants. When they flag a hawkish SARB tone, they are describing what the rand is doing on venues where large orders clear at posted prices without a risk-engine gate. That description is accurate for their client base. It is not the market a retail trader on a MetaTrader platform will see or execute against in the same minute.
Why does my broker's spread widen so dramatically during a SARB press conference?
The widening is a composite. Part is the genuine interbank spread expansion during the news window. Part is the broker switching your order from B-book to A-book because they no longer want to warehouse the exposure. Part is a hedging cushion the risk engine adds to protect against the fill being reconstructed against the broker later. The retail spread you see reflects all three components combined. It does not decompose them for you.
Is stop-hunting during SARB events actually happening at retail brokers?
Stop-hunting as a deliberate strategy — the broker actively pushing price to trigger client stops — is largely a mischaracterization. What happens in practice is that during the announcement window, the platform's fill logic executes queued stop orders at whatever price is first available after the freeze clears, which is often several pips beyond the trigger. The economic effect on the trader is identical to being hunted. The mechanism is passive queue management, not active predation.
Do FSCA-regulated brokers behave differently on SARB days than offshore ones?
The FSCA does require licensed operators to disclose execution policies and to maintain records that support fill reconstruction if a client disputes a trade. In practice, the risk-engine behavior during SARB events is broadly similar across FSCA-licensed and offshore ZAR-carrying platforms, because both are exposed to the same interbank spread widening. The difference sits in what the regulator can compel the broker to produce if a client complains. FSCA licensees have more disclosure obligations, not different execution mechanics.
Can a retail trader avoid the requote and slippage problem around SARB announcements?
Partial avoidance is possible. Placing orders outside the sixty-second window before and after the scheduled communication meaningfully changes the fill distribution. Using limit orders rather than market orders shifts the risk from bad fills to non-fills. Trading with an account type that is contractually A-booked — some brokers offer these explicitly at wider baseline spreads — bypasses the book-switching behavior. None of these fully eliminate the exposure. All three change its shape.
What happened in January 2016 that this desk keeps referring back to?
Over a short window on a Sunday, the rand gapped against the dollar during a period of thin liquidity between the New Zealand session close and the Tokyo open. The event itself is well documented in the public record. What is not public is the reconciliation trail — the internal process by which broker back offices matched client tickets to liquidity-provider confirmations in the days afterward. That reconciliation cycle produced the operational lessons every retail ZAR platform still uses to structure their risk-engine behavior on SARB days.
If I hold a ZAR position through a SARB communication, what should I actually expect?
Expect a spread several times wider than the platform's advertised typical spread for the minutes surrounding the announcement. Expect a fill delay on market orders in the range of a few hundred milliseconds to just under a second. Expect requote dialogs on pending orders that were placed close to the current price. Expect that stop losses will fill worse than the trigger price by an amount that reflects the queue-clearance dynamics, not the underlying market's direction alone. All of this is inside the client agreement you signed.
Why does none of this appear on the broker's marketing materials?
It is disclosed in the terms-and-conditions document, which every client signs. It is not disclosed in the marketing pages, which every prospective client actually reads. The gap between the two documents is the industry's standard information architecture. Regulators require the first document. Marketing teams write the second. The behaviors described in this article live in the space between them, and that space is where retail execution actually happens on a hawkish SARB day.