Crisis management for forex traders – MT4 outages, fat finger, gaps and tilt

A connection failure, incorrect position size or sharp move after an announcement requires a different response from an ordinary strategy signal. First establish which orders and positions actually exist, then control the confirmed risk using a prepared procedure. This article covers platform, internet and power failures, operating errors, market events and a loss of decision discipline. We explain how to use backup channels, document an incident and confirm that instructions were executed. Historical events and illustrative thresholds help explain the mechanisms, but cannot replace your account's terms. You will finish by preparing an emergency card and a rehearsal plan, so an outage does not force improvisation or increase exposure through repeated orders. The technical preparation is covered in Article 15.1.

Scope: the Polish source mainly concerns retail FX/CFD trading through retail brokers (MT4/MT5), rather than interbank spot FX. Its Polish retail-account example refers to KNF decision DAS.456.2.2019, effective from 1 August 2019 after the temporary ESMA intervention expired. This jurisdictional context is retained; protections must not be assumed identical for every international account. Material differences — negative balance protection, B-Book/ECN model, and rollover hours — are identified where relevant.

1. First 30 seconds – crisis or catastrophe

A crisis is anything that is not in the playbook. A platform outage in the middle of a trade, a gap against you, two phones ringing at once, NFP +199k instead of a +130k consensus, the broker offline, a hand that clicked "Buy 1.0" instead of "Buy 0.10". In every one of these scenarios time is the dominant variable – not price, not analysis, not "what would somebody on a desk do". The first action is always the same: cut exposure first, analyse later.

During an outage there is no time to design a new procedure. Plan the decisions beforehand, then identify the scenario and execute the appropriate loss-limiting steps. In the first 30 seconds you do not invent a solution – you recognise the type of failure and trigger the variant you wrote earlier: close, hedge, phone, mobile or stop the session. Any trade instruction first requires confirmation of the current position state.

Three questions you ask within 5 seconds

1. How much open exposure do I have? Track two separate quantities: nominal exposure (direction and lots) and risk to the stop-loss as a percentage of the account. Add currency exposure with its direction: 0.3 lots long EUR/USD plus 0.2 lots long EUR/GBP gives 0.5 lots long EUR; 0.3 lots long EUR/USD plus 0.2 lots short EUR/GBP gives 0.1 lots net long EUR. Calculate risk separately from the stop distance and pip value. Lots are neither account risk percentages nor margin.
2. Are my SL and TP active? Are they on the broker's server or merely mental levels? Server-side orders remain active when your terminal disconnects. MT4 trailing stops are managed by the client terminal: when it stops running, the last submitted SL remains, but further trailing updates stop. See the MT4 trailing-stop documentation^[8] (checked in the Polish source on 6 September 2026).
3. Which scenario is this? My outage (hardware, internet, power), broker outage (platform, server, regulator), market event (release, intervention, gap), or my error (fat finger, tilt)?

Depending on the answer you pick one of the 8 procedures described in sections 2 – 9. The procedures are written in advance – your job is only to launch them, not to invent new ones in real time.

Decision tree for the first 30 seconds

First action: cut exposure, then analyse

The most common retail mistake in a crisis is trying to recover P&L during the event. The platform unfreezes after 40 seconds, the price shows you are −0.8% under water, so you add a position to "average in". Within 4 minutes you are −2.3% and your hand is shaking. Hard rule: in a crisis you do not add to positions, you close them or freeze them. You come back to play the next day, when you know what happened, why, and whether your thesis is still valid.

Second duty: do not generate the next crisis. A crisis tends to multiply – platform outage – fast manual close – fat finger – tilt – another position. The procedure exists to break this chain. After the first event you stop operationally, regardless of how you feel.

What "having a procedure" actually means

A procedure is a 1 – 2 page A4 document in a visible place (a sheet under the monitor, a PDF on the desktop). For each scenario it contains: (1) a trigger (how you recognise this is the scenario), (2) the order of actions in the first 30 s, 5 min and 1 h, (3) phone numbers, (4) a default decision: if the position state is unknown, send no further orders until it has been confirmed through the second or third channel; use “close all” only for a confirmed position state. A printable template is in section 10.

The mechanism that makes the procedure indispensable is neurobiological: stress hits working memory capacity, so creative thinking five seconds after an outage may be impaired. A pre-written sequence is not a luxury – it is a substitute for the thinking you do not have at that moment.

2. MT4/MT5 platform outage — connection broken, freeze, restart

MT4 and MT5 were designed for a different scale of load from today's. They generally operate steadily in normal conditions, but NFP releases, FOMC or ECB decisions and unexpected interventions can coincide with timeouts, requotes, delayed confirmations and “Trade context busy” messages affecting many clients. The recurring catalogue of symptoms ranges from individual rejections to a complete connection failure lasting tens of seconds. How frequently this happens at your execution provider is unknown here: no verified sample or observation count is available. Measuring it in your crisis journal is the first step. Treat the platform's technical limitations as an operational risk, rather than dismissing them as bad luck; the symptom alone does not establish whether the cause lies in the terminal, connection or broker infrastructure.

Classifying platform failures

Diagnostic mini-checklist: local, broker or market?

Before activating the three-channel procedure, spend 10–15 seconds classifying the problem and selecting the appropriate response.

• Ping/traceroute to the configured server host, if the broker discloses it: for example, ping [server-host-from-MT4-MT5-settings]. No ICMP reply does not necessarily mean an outage: some brokers block ping. Treat it as supporting evidence, not definitive proof.
• Mobile MT over cellular data, rather than the desktop's Wi-Fi: if mobile displays positions while desktop cannot, investigate the local device or connection.
• Prices on TradingView or another broker: stalled comparison quotes during London trading can indicate a wider outage or market disruption; normally moving comparison quotes with frozen broker quotes point toward that broker's feed or connection. These are diagnostic clues, not proof from a single feed.
• The broker's website status and social accounts: consult its official status page and Twitter/X account, alongside services such as Downdetector, for corroboration of a broker outage.
• Timestamp screenshots: save each screenshot with its date and time in the folder or filename, for example 2026-05-04_14-30-12_no-connection.png. This is evidence for a complaint.

The three-channel procedure: parallel channels, not a choice

The key is to use all three channels in parallel. Do not spend the entire incident choosing between them: having a procedure means activating the prepared alternatives.

1. Channel 1: desktop MT4/MT5. This is the frozen terminal. Do not force-close it immediately; allow 30 seconds for recovery while starting the other channels.
2. Channel 2: call the broker. Use the duty dealer's number, where that service is offered, rather than general customer support. Dial at t = 5 seconds whether or not desktop appears to recover. Say: “Login XXXXX, I have position XX open. Please confirm its status / close it at market.” A recording can provide evidence. Confirm the broker's recording arrangements; if recording yourself is available and lawful, follow the consent requirements applicable to the call. Do not assume every phone supports recording or every jurisdiction has the same rules.
3. Channel 3: the MT4/MT5 mobile app. Use independent internet access, such as cellular data, with a previously authenticated active session. It lets you check the server's positions independently of desktop. Use that confirmed server view, or the dealer's confirmation, until desktop catches up.

Activate all three within the first 15 seconds. A channel that does not work is a separate preparedness problem to resolve outside trading hours, rather than discovering it during an outage.

What the broker can do by phone

The Polish source describes emergency dealing-desk telephone services at regulated EU, UK and Polish brokers. Availability is broker- and account-specific: verify the dedicated number, service hours, authentication and permitted operations beforehand. Depending on the service, the dealer may be able to perform the following:

• Close a position at market, normally the most direct emergency instruction: give your login, position ticket and “close at market”.
• Modify SL/TP, including submitting a server-side stop where you previously had only a mental level.
• Partially close a position, for example “close half”.
• Block further trading: some providers may permit an emergency account restriction, for example after a tenfold size error, until manual release. Procedure to verify: check your own account documentation; do not assume this option exists.

What the broker will not do through an ordinary emergency trade instruction: change your leverage, undo a completed order simply because you dislike the fill, or close at a price from a minute ago. A market-close instruction is subject to the executable market price when processed and confirmed.

Restarting MT4/MT5: when and how

Rule: restart only after confirming the position state on the server through channel 2, the dealer, or channel 3, mobile. Restarting before independent confirmation risks working from a stale view of the account. The Polish source also raises a possible rare buffered-order duplication scenario, but supplies no build number or supporting document: this behavior remains unverified and must not be treated as an established platform feature. Confirm state first, then restart.

Restart procedure:

1. Save a screenshot of the current state in a timestamped “crisis” folder.
2. Check positions through mobile or the dealer.
3. Close the terminal. If it does not respond, end the process in Task Manager at this stage. The method of terminating the client does not itself change the server's positions; stop interacting with the frozen interface.
4. Restart. Log in after channel 2 or 3 has confirmed the expected position state.
5. Compare the reopened terminal's orders and positions against the mobile or telephone confirmation. Call the broker about every discrepancy.

“I clicked, lost the connection and received no confirmation”

This is an easy way to inflict a second loss. You click Buy, the screen flickers and displays “Trade context busy” or “No connection”; now you are uncertain about the attempted instruction. Do not treat the absence of a confirmation as proof of no execution. An instruction whose transmission is uncertain may have executed, been rejected or remain pending. Those states require different responses, but share one rule: send no further instruction until you know the state. The local “Trade context busy” error described above is not itself an acknowledgement that an order reached the server.

1. Stop sending, t = 0–5 seconds. Hands off the keyboard. No repeat click or precautionary close. Repeating an instruction against an unknown state can create duplicate or unintended opposite exposure.
2. Establish the provider's state, t = 5–30 seconds. The broker's server holds the positions; your frozen screen does not establish their status. Check Trade and History in mobile over cellular data. If that route also fails, call the dealer: “Login XXXXX, please list open positions and orders from the last two minutes.”
3. Record the state, t = 30–60 seconds. Save a screenshot with seconds visible, the ticket if available, click time and confirmation time. These are complaint and crisis-journal evidence that cannot reliably be reconstructed later.
4. Decide only after confirmation. Executed once: manage that position under its plan. Executed twice: close the duplicate rather than the whole intended position. Not executed: reassess at the current price, not the price from 40 seconds ago. Still unconfirmed: submit neither an opening nor a closing order; document and wait for confirmation. A server-side SL does not disappear merely because you cannot see it.

The rule applies to exits as well as entries. If an attempted close is unconfirmed, another instruction sent blindly can produce unintended opposite exposure, depending on the order type and account mechanics, instead of finishing the original close.

Broker hotline response time: what to measure

A dealer's number is only as useful as its real response time during high-impact events. Product-page promises are insufficient, and no externally verified average or sample count is available here. Run a routine monthly check that the number is current and a quarterly stress test: call 30–90 seconds after a release treated as lower priority, such as UK PMI or retail sales, and record connection time in the crisis journal. Several observations give your own order of magnitude instead of somebody else's claim. Record the fastest and slowest measured response together with the number of attempts, not just one average. A few tests cannot establish how the line will behave during a future event unlike anything measured.

3. Power and internet failure — when the UPS is no longer enough

The UPS bought time for a controlled response, but power is still out. In this scenario eight minutes of backup runtime remain, your phone has 47% battery, and wired internet failed with mains power because the router was not connected to the UPS, a common preparedness mistake. Cellular data works, but laptop failover is not configured. Your long EUR/USD position has planned risk of 1.2% of the account. The source's opening mixes elapsed and remaining runtime; the decision here uses the measured eight minutes remaining.

This procedure begins exactly where Article 15.1 ends. That article prepares the workstation: equipment selection, measured battery reserve and an independent control channel with the dealing-desk number, account number, PIN and charged power bank. It directs the sequence for an open position here. Once hardware redundancy is exhausted, three resources remain: the phone, mobile MT and the duty dealer.

The decision within 30 seconds of a blackout

Overlapping conditions have a defined priority: an ongoing macro release and risk above 2% override the server-side-SL rows, including for positions with server-side stops. Stops do not guarantee execution price during events or gaps, and without equipment you cannot readily respond to a worse fill. Assess the event and swing context first, then apply the risk bands; the swing row never overrides a higher-priority close condition. These thresholds are the source's illustrative procedure, to be reconciled with your own plan.

When the phone is your only equipment: prepare these beforehand

Article 15.1 requires the broker's current number in pinned contacts. Here you verify that the entire route works:

• Mobile MT4/MT5 authenticated, with the password available securely or in a password manager. If you cannot authenticate, you do not have a usable session.
• A charged power bank. Capacity and equipment selection belong in 15.1. Phone runtime varies with model, screen brightness and signal. Measure your own runtime with MT open over cellular data and enter it on the incident card: this is a procedure to verify, not a universal battery-life claim.
• A second device with MT. 15.1 establishes whether you need it and which equipment fits. Here the question is whether a drill demonstrates switching to it within a measured time.
• The duty dealer's number first in Favorites, checked this month through a routine test call during London session hours.
• Login and account number in a secure note, such as a password manager, rather than an ordinary unsecured notes app. The dealer needs the account details and required authentication before accepting instructions.

What remains after preventive safeguards fail

Equipment by capital scale, UPS, dual-WAN router, second terminal and power bank are covered in 15.1. This section concerns the actions after those safeguards have failed.

One dangerous misconception needs correcting: marking Wi-Fi as a metered connection does not create cellular failover. It limits background data use. Cellular settings such as “Let Windows keep me connected” or “Use cellular whenever Wi-Fi is poor”, where available, govern the relevant connection behavior. Interface priority involves metrics, for example netsh interface ipv4 set interface … metric=; netsh interface set interface enables or disables an interface rather than setting its metric. These are the Polish source's notes dated 6 September 2026; verify the controls in your Windows version and hardware configuration. Configuring metering proves no failover capability. Test by unplugging the router's WAN cable while Wi-Fi remains connected, because disabling Wi-Fi tests a different failure, and time reconnection to the broker's server. Configure the equipment under 15.1 and record the measured switch time on the incident card.

Power returns: the first hour

Do not immediately resume trading. Follow this sequence:

1. Check that the modem and router restart automatically, allowing the scenario's 10–60 seconds.
2. Check that the laptop loads the correct profile. Windows updates after a restart can delay access; the scenario allows 5–15 minutes.
3. Confirm positions through channel 3, mobile, before logging into desktop MT.
4. Reconcile every order since the blackout against mobile and account history. Call the broker about any discrepancy.
5. Use the first 30 minutes after equipment recovery for observation rather than new trades. Residual stress can impair decisions even after the technical problem ends.

4. Fat finger: wrong side, size or instrument

Fat-finger mistakes affect professionals as well as retail traders. The useful professional response is “break the trade, eat the spread”: promptly remove the mistaken exposure and accept execution costs. Trying to repair it with an additional position can turn one error into two, leaving two trades at unintended prices. The cost includes spread, slippage and any commission, not spread alone.

Classic fat-finger errors

The “eat the spread” rule

The central principle is to remove unintended exposure promptly instead of accepting an open-ended holding period. The Polish source describes immediate closing costs as known and capped, but market execution cannot guarantee that cap: spread, slippage and price movement remain variable. The operational comparison is still useful:

• Estimate immediate execution costs from spread, illustrated as 0.5–3 pips on majors or 5–15 on exotics, plus slippage and commission on commission-based accounts. These are scenario ranges, not guaranteed quotes; ordinary conditions may mean a few pips, while a crisis can cost much more.
• Keeping the mistaken position introduces unknown additional market risk, potentially consuming the account under adverse conditions.
• Time spent debating whether to keep it leaves the position exposed to price drift.
• Holding an error on hope can impose a much larger emotional burden for the day than accepting a prompt corrective close.

Rule: if the executed position differs from the intended one, remove the mistaken exposure at market and enter the intended trade only under its plan. Predetermined exceptions, such as retaining the correctly sized portion, belong in the playbook before entry; do not invent exceptions under stress. The reasoning rests on the exposure mechanism rather than an unsupported statistic: delay and larger size increase the scope for damage, although price need not move adversely every second.

Calculating fat-finger cost

A Close All hotkey: faster than repeated clicks

During a fat-finger, NFP or intervention cascade, every second spent navigating the terminal leaves exposure to price drift. A verified Close All script or EA can request market closure of open positions through a shortcut, where supported. It requires a confirmed position state and does not guarantee simultaneous fills. Configuration:

• Use the item's context menu in Navigator and “Set hotkey” where available. The MT4 Navigator documentation^[8] describes this for indicators; the Polish source checked it on 6 September 2026. Verify availability for your script and terminal version. Otherwise launch the script by double-clicking it in Navigator. MT4 does not provide a shortcut-assignment route under Tools → Options → Charts.
• Free Close All Positions scripts are available in the MQL5 community. Review the code, permissions and behavior with partial positions; test on demo and then small live exposure before relying on one.
• Choose a key combination that does not conflict with terminal shortcuts. F11 activates full-screen mode in MT4 and is therefore a poor Close All choice.
• A broker-side kill switch may block new orders after a daily-loss limit. It is not a standard feature: verify whether your account offers it and how it is reset. Use the loss threshold from the trading plan, 15.2, rather than creating another limit.

A hotkey does not replace the three-channel procedure: it depends on a responsive desktop. With the terminal working, price moving and hands shaking, a tested shortcut can reduce repeated mouse actions compared with clicking through five positions.

Prevention built into the workflow

Five habits reduce opportunities for fat-finger mistakes without making them impossible:

1. Set the default lot size to the minimum, illustrated as 0.01 in MT4/MT5 where the instrument supports it. Increasing size requires deliberate input before the trade.
2. Disable one-click trading if it is unnecessary for your execution method. The ordinary confirmation window provides another two or three seconds to check.
3. Keep a position calculator in a separate window. Enter account-risk percentage and calculate lots. This reduces the “1.0 versus 0.10” opportunity but cannot eliminate a manual transcription error.
4. Use distinct chart backgrounds, for example blue for EUR/USD and orange for GBP/USD. This helps distinguish instruments, but not two charts of the same pair at different time frames.
5. Use an account-level maximum-lot restriction if offered. Some providers may support a hard cap, for example no order above 0.5 lots. Verify it in the account documentation. Otherwise rely on a conservative default and position calculator, without pretending they are a hard broker-side block.

Automation errors, inconsistent data and the wrong account

Three related variants create orders without the intended decision or send them to the wrong destination. Instead of one mistaken click, you see a series of orders or a position in an account where you did not intend to trade.

Feed versus code matters operationally. A feed problem may stop when valid quotes return; a logic bug can recur at the next signal. Keep the algorithm disabled until the cause is known, rather than restarting merely because activity has quietened down. Preventive measures include one terminal per account or clearly different profile colors, and a maximum simultaneous-position count enforced within the EA. A code-level limiter can react faster than a human.

5. Unexpected macro data — NFP, CPI and FOMC disorder

NFP consensus +130,000 versus a +199,000 release; year-on-year CPI expected at 2.9% but reported at 3.4%; an FOMC dot plot implying two cuts rather than four. These illustrate surprises outside the expected range. The source's working definition uses a surprise above 1.5 standard deviations, which requires an explicitly measured historical surprise distribution; the quoted numbers alone do not establish that threshold. Its illustrative reaction sequence is an initial 30–90-second panic move, a possible double-top or double-bottom flush over the following 5–15 minutes, and a new price range after approximately 30 minutes. This is a scenario, not a guaranteed market sequence.

Retail traders should not assume they have an edge under these conditions. The source's stress scenarios include spreads at 5–15 times normal and slippage of 5–30 pips on majors or over 100 on exotics; these are not verified frequency estimates. Platform overload and strong emotions can further impair execution. Procedure: trade the release only under an event playbook written beforehand.

Three decisions before the release

Make these decisions 15 minutes before publication, not during it:

1. Do I have exposure contrary to the scenario I am preparing for? If so, apply the predefined reduction, illustrated as 50%, or close it. Holding through the event requires reserve for execution risk beyond the directional thesis: a first-minute SL can fill at the next available quote tens of pips beyond its level.
2. Am I planning an event trade? Only with a written playbook specifying (a) the setup, for example fading the first move after 90 seconds if the defined volume measure is below x; (b) sizing at 50% of normal in this scenario; (c) a stop distance calibrated to event volatility rather than ordinary chart structure, with methods and thresholds in the strategy specification, 15.3, and regime filters in 15.6; and (d) explicit acceptance of the modeled 10–30-pip slippage range. Here you check that the stop distance was defined before the release; acceptance of a range does not cap actual slippage.
3. Do I have a plan B? Specify the response to a frozen platform, failed internet or Off quotes. Without a backup procedure, do not enter.

The first 30 seconds during the release

The market is already moving a second after publication. If you did not plan an event trade, use the following sequence, extending through the first completed minute:

1. t = 0 seconds: check exposure. If account risk exceeds the procedure's illustrative 0.5% threshold and the position opposes the move, execute the predefined immediate close after confirming its state.
2. t = 5 seconds: check that stops are on the server and match the predefined event-risk plan. Do not improvise a wider stop to postpone a loss.
3. t = 10 seconds: keep the broker's number ready; dial if the failure procedure is triggered.
4. t = 15 seconds: observe the initial reaction without treating it as a completed signal. If the release occurs at the minute boundary, its M1 candle closes at t = 60 seconds; at t = 15 it is still forming.
5. t = 30 seconds: begin a discretionary review of reducing or closing an existing position. This waiting point does not override the immediate protective exit at t = 0 or any earlier predefined risk trigger.
6. t = 60 seconds: the first M1 candle closes. Only now is that candle complete; any new position still requires the event playbook written before publication.

Desk rule: during the first minute, execute previously planned decisions rather than inventing new trades. A setup not planned before publication is not an entry in that first minute. Execution conditions and market microstructure can differ sharply from ordinary trading, beyond merely higher volatility.

Event playbook: a binary pre-release checklist

Every required condition must be explicitly satisfied. One missing YES means no event entry.

Example: NFP +199,000 versus +130,000 expected

In this illustrative scenario, the surprise strengthens USD and EUR/USD falls 50–80 pips in the first minute. A trader long EUR/USD faces three possible outcomes:

• Server-side SL 25 pips below entry: triggered with eight pips of slippage, producing a 33-pip loss instead of 25. This is the scenario's accepted execution overrun, not guaranteed protection at 25 pips.
• Mental stop 25 pips below entry: platform lag prevents a close during the first 15 seconds. After 90 seconds price is 70 pips below entry; with eight pips of slippage, the 78-pip loss is approximately three times planned.
• No stop, 0.3 lots and PLN 8,000 equity: an 80-pip adverse move costs 0.3 × 80 × PLN 38 = PLN 912, or 11.4%. Holding for a recovery leads to 110 pips against the trade after 12 minutes: PLN 1,254, or 15.7%. At 1:30 leverage, 0.3 standard lots require about EUR 1,000 margin, using EUR 3,333 per standard lot. With a separate scenario assumption EUR/PLN = 4.30, margin is approximately PLN 4,300. Margin level falls from about 186% to 165% and then 157%, reducing the buffer to the broker's applicable warning and close-out levels.

A server-side SL is protection that remains active without your terminal; it is not the only risk control. Smaller size and exiting before publication reduce exposure when stops fill tens of pips beyond their levels. A mental stop depends on the trader, keyboard, connection and terminal reacting in time. A planned 1% loss can therefore become 3–5% in an adverse scenario; the source supplies no probability estimate for that outcome.

6. Weekend gaps and the Sunday open

Retail FX generally closes on Friday around the 17:00 New York rollover, usually 23:00 Polish local time outside US–EU daylight-saving mismatch weeks, and reopens on Sunday evening. The exact session depends on your broker and instrument: the Polish source lists examples of 22:00, 22:30 or 23:00 Polish time, all requiring verification against the account specification. During approximately 48 hours, elections, geopolitical developments, regulatory decisions and political statements continue. The source's gap scenarios range from 30 to 500 pips. A stop-loss does not guarantee price across a gap; it can execute only against available liquidity, potentially tens or hundreds of pips beyond the stop.

Friday at 18:00: reduce weekend exposure

The source reviews weekend holdings around 18:00 Polish time, approximately five hours before a 23:00 close. Its default is a 50% reduction in swing-position size. This is an illustrative assumption, not a market standard; derive your own reduction from the gap loss your account can withstand. In this example, normal planned risk of 1.5% per position becomes 0.75%. Translate the review time to your own location and the actual broker close.

Why consider a 50% reduction? Weekend gaps can have heavy tails: the source contrasts ordinary 5–25-pip scenarios with extreme 100–500-pip cases. It provides no verified sample, gap definition or frequency distribution, so those ranges cannot establish a probability or validate 50% as sufficient. Size for a severe stress case, not just an average.

Weekend reduction schedule

Instruments requiring additional care

Sunday review: 30 minutes before your broker opens

Start approximately 30 minutes before your broker's specified opening time. The source's 22:00, 22:30 and 23:00 Polish-time examples are broker-dependent, not interchangeable. Checklist:

1. Which weekend news can affect your pairs? Review elections, G7 statements, Asian central-bank announcements including RBA, RBNZ and BoJ, and relevant political statements or social posts.
2. What do available Asian futures indicate? Nikkei 225, Hang Seng and ASX contracts can inform sentiment when their own sessions are open. Verify timestamps; do not treat stale or not-yet-open markets as current evidence.
3. What spread does your broker show at reopening? The source models spreads at 3–5 times normal during the first 5–15 minutes. Avoid entries in that unstable window under this procedure; measure actual conditions.
4. Do Friday's stops still fit the plan? A 30-pip favorable gap may justify a structural stop update under the existing rules, not an improvised increase in allowed loss.
5. What is the first-hour plan? The Sunday opening hour can have thin liquidity and unusual execution. The procedure normally calls for observation rather than a new trade.

7. A central bank against your position — interventions, peg breaks and communication shocks

Direct currency intervention, abandoning an exchange-rate floor and unexpected central-bank communication can invalidate ordinary execution assumptions. The SNB's January 2015 floor removal and Japan's MoF/BoJ operations in 2024 are different mechanisms with a shared operational risk: prices can skip levels. The source models a 50-pip stop filling another 800 pips away and margin level dropping from 500% to negative equity within 90 seconds. Communication shocks, including an unexpected rate decision or RBA statement, are illustrated by 80–200-pip moves over one to three hours. These are stress scenarios, not established timing distributions.

Three variables during intervention

1. Resources matter. Major central banks can command foreign-exchange reserves on the scale of hundreds of billions of dollars. A retail position of 0.3 lots has no capacity to oppose that flow; size and exit preparation matter more than confidence.
2. Time matters. A scheduled statement or press conference may give preparation time; an intervention or policy change may be unannounced. The source's 5–60-minute warning window and claim that 50–80% of a move occurs within 90 seconds are not supported by a supplied sample. Do not rely on either as a forecast or assume a 15-second response guarantees a tolerable fill.
3. Broker liquidity matters. Your broker is the final execution link. If liquidity providers withdraw quotes, modeled here for 30–60 seconds, a market instruction can meet the next executable price 100–500 pips away. These ranges are illustrative; the available quote, not the screen's last price, determines the possible fill.

SNB, 15 January 2015: case study

The Swiss National Bank had maintained a EUR/CHF floor at 1.2000 since 2011, supported by purchases of foreign currency against CHF. The Polish source gives reserves of roughly 80% of Swiss GDP as an order of magnitude but supplies no verification for that figure; it is not a precise sourced estimate here. At 09:30 GMT, 10:30 CET, on 15 January 2015, the SNB unexpectedly announced the floor's removal.

The following price path is synthetic and illustrates speed, not a quotation record. The source identifies the announcement time and a daily low around 0.85 as historical reference points, but has no minute data establishing the intervening sequence. In particular, the table does not establish that the low occurred in the third minute.

Consequences described in the Polish source:

• FXCM, then listed on the NYSE, reported approximately USD 225 million of negative client balances in announcements dated 15–16 January; the source explicitly notes that this number is not in the cited 8-K. It obtained a USD 300 million two-year Leucadia loan, initially at 10% annually, increasing by 1.5 percentage points per quarter up to 17%^[4]. The source records an approximately 87% share-price fall and resumption of share trading on 20 January.
• Alpari UK announced insolvency on 16 January 2015, the day after the SNB decision.
• Retail traders long EUR/CHF could lose their accounts and incur negative balances despite treating the defended floor as a safe premise. The source describes thousands affected but supplies no count methodology. Where applicable protection or contractual relief was absent, negative balances could remain debts owed by clients.

A seemingly safe thesis does not make a position safe. A EUR/CHF long with a 1.1900 stop could be sized to an apparent 0.15% planned account loss; that figure requires the position and account assumptions and is not implied by the stop level alone. A discontinuous break of the floor could instead consume the account and create a negative balance. The stop did not guarantee execution at its specified price.

Japan's MoF and BoJ in 2024: USD/JPY interventions

Japan's Ministry of Finance, MoF, decides foreign-exchange intervention; the Bank of Japan acts as its operational agent. The 2024 operations supported the yen, with official records published by the MoF^[5]. The source's 300–600-pip moves in five to 15 minutes have no supplied price-series verification. Operations need not be confirmed contemporaneously: statements and later reports provide different information. The following are illustrative observation thresholds from the 2024 regime, not current-market levels or official intervention triggers:

• USD/JPY above 158: use 50% of normal JPY exposure in this example and recognize intervention risk.
• USD/JPY above 160: use 25% of normal exposure and review MoF/BoJ statements daily.
• USD/JPY above 162: the scenario adopts a heightened intervention-risk posture; some traders would avoid JPY entirely. This threshold does not establish an intervention probability.

An operational intervention alert: what to calibrate

The MoF publishes intervention records; the BoJ executes the operations. The Polish source describes periodic records arriving with a one-to-two-month delay as historical calibration material. Check the actual release type and schedule rather than using a delayed report as a live trigger. Alternative operational clues include:

• A TradingView alert on one-minute USD/JPY bars for an absolute move above JPY 1.50, or 150 pips, over five M1 bars. In the source's Pine v5 syntax: math.abs(close - close[5]) > 1.50. This symmetric condition detects either direction; it does not identify the cause.
• Relatively low one-month implied USD/JPY volatility, illustrated as below the low teens in volatility percentage points, alongside a large five-minute move. Treat this as a candidate diagnostic feature to test, not proof of intervention.
• Morning statements in Japan. JST is UTC+9, eight hours ahead of Polish winter time and seven ahead of summer time: 09:00 JST is 01:00 CET or 02:00 CEST in Poland. Live monitoring therefore requires a night/early-morning schedule or alerts. The source's suggested one-to-five-day interval between verbal and market intervention is unverified and must not be treated as a countdown.

RBA: communication shock rather than the same mechanism as an FX intervention

For the retail trader, the Reserve Bank of Australia example concerns rate decisions, a changed statement or the Governor's remarks, rather than treating every AUD move as a direct currency operation. The source models 80–200 pips over one to three hours after communication and moves two or three times ordinary daily ATR; no frequency evidence is supplied to call those magnitudes typical. The operational lesson remains that a central-bank statement can invalidate a tight ordinary-session stop. Reduce exposure before the known central-bank window according to the plan and avoid unplanned entries during the press conference.

Negative balance protection: what the crisis procedure needs to know

The practical question is whether a gap can leave this account below zero. The answer depends on client classification, broker entity and jurisdiction, not the platform display. Section 13 covers eligibility, professional-client reclassification and the Polish framework: KNF decision DAS.456.2.2019 took effect on 1 August 2019 after the temporary ESMA measure expired^[2]. See retail-client protection and retail versus professional status. Record only the checked result on the incident card: “negative balance protection: yes / no / to verify”, with the relevant account/entity. Where applicable, protection limits liability by reference to the funds in the covered account, not merely the margin allocated to one position. Margin is collateral tied up by exposure, not a loss limit. Do not export the Polish or EU rule to an account elsewhere without checking its own framework.

8. Broker outage — when nobody responds

You can prepare alternatives for your own equipment or connection failure. When the broker's server is offline, its web platform is inaccessible and the phone goes unanswered, your control is minimal. Preparation starts before the incident, with infrastructure diversification.

A second broker account: its role during an incident

Article 15.1 covers backup-provider selection, including jurisdiction, liquidity provider, execution model and funding. Here the issue is what the account does, and does not do, during an incident:

• Purpose: a temporary hedge, not another strategy, and only against a confirmed open position. If phone or mobile confirms the main position's direction and volume, an opposite, exposure-matched position at the backup broker can reduce directional risk. That offset exists only while both legs remain open; confirmation is a snapshot, not a guarantee for the next several hours.
• What it cannot replace: knowledge of the main account's state. If its server-side stop has already executed, opening a precautionary “hedge” creates a new unhedged position in the opposite direction, during the same move that triggered the original loss.
• Readiness condition: the backup account must have an authenticated active session before the incident and sufficient funds for the exposure actually held at the primary broker.
• What can break the hedge without a visible frontend: the primary SL or TP can execute; either leg can be stopped out for insufficient margin; different contract multipliers or trading hours can make nominally equal lots unequal. The procedure therefore requires margin reserve at the backup, repeated confirmation of the main position while the hedge is held, and a prewritten response to losing one leg: ordinarily close the confirmed remaining leg rather than adding a third position.

Broker-outage symptoms

Hedging through the backup broker

Suppose the main broker is offline and you believe you hold 0.5 lots long EUR/USD. The procedure starts by determining whether opening anything is justified.

1. Step zero: confirm the primary position. Ask the duty dealer for open positions and volumes, check Trade through mobile, or obtain a sufficiently current account report through a trusted channel. A hedge is permitted only if the position is confirmed open with known volume. An old emailed snapshot is not current confirmation.
2. This is substantive, not a formality. The server-side SL may already have executed, leaving a new opposite position if you hedge blindly. A partial stop fill changes remaining exposure; hedging the original full amount can reverse it.
3. If confirmation is impossible, open nothing. Record screenshots, attempted-contact times and a timestamped message to the broker, then continue seeking confirmation. Lack of visibility does not cancel an existing server-side stop, although a stop still cannot guarantee price or execution while required liquidity is absent.
4. Against a confirmed position, open the opposite exposure at the backup broker. Confirmed long 0.5 standard lots corresponds to short 0.5 only with matching contract specifications; a confirmed remainder of 0.2 requires 0.2, not 0.5.
5. Record the timestamp, execution price and how you confirmed the primary state. This documents a decision made against known exposure rather than a guess.
6. While waiting, repeatedly confirm the main position through an independent channel and monitor backup margin. When the main broker returns, reconfirm before unwinding both legs as closely together as executable, or retaining the main leg under its plan. If one leg has disappeared, close the confirmed remaining leg rather than adding a third.
7. Budget the full operation's costs. The source's simplified allowance is two spreads plus 1–3 pips of inter-broker price drift; actual slippage, commissions and financing must also be included. This expense purchases useful control only while the state of both legs is known.

Contract and evidence: some broker agreements contain force-majeure or erroneous-price provisions that may be invoked to adjust disputed executions after an outage. Review the actual terms and applicable rights; do not assume every reconstruction is valid or necessarily adverse. Screenshots, timestamps, account records and lawfully obtained recordings support a dispute. Do not rely entirely on the broker's recordings. Request copies through the applicable access or complaints procedure; your own contemporaneous evidence is the material you control and can produce immediately.

Basis risk: the second broker is not a perfect offset

Matching opposite exposure reduces directionality but does not remove execution and basis risk. Differences affecting the combined result include:

• Spread: providers can diverge sharply during an event. The source models a 1–5-pip difference, without supplying a distribution that would make it a universal typical range.
• Execution: ECN versus market-making arrangements, different liquidity pools, last look and asymmetric slippage.
• CFD specification: similarly named symbols can have different multipliers, settlement or roll arrangements and trading hours. Check the contracts rather than matching the symbol text alone.
• Trading hours: one leg may be tradable while the other market or broker session is closed.
• Overnight swap: differing financing rates can erode the combined result over several days.

The hedge offsets the dominant directional component only while its conditions hold. The source allows 5–15 pips of inter-broker noise in its outage scenario. That is an illustrative cost budget, not a guaranteed cap or perfect hedge.

Record the incident and hand over the evidence

Documentation is part of the crisis procedure and begins during the incident. Complete at least the following within 24 hours:

1. Time: start and end of the symptom to the second in one time zone, preferably UTC, plus attempted-contact times.
2. Account state: account number, tickets, instrument, side, volume, entry and the planned SL/TP at the incident time.
3. Independent evidence: timestamped quotes from another provider, platform screenshots showing seconds and mobile logs where available.
4. Requested remedy: one clear sentence requesting an explanation, execution correction or compensation.

The complaint letter, escalation route and limits of the broker's obligations belong in Section 13's broker-dispute procedure. This crisis procedure supplies a complete evidence package rather than inventing a separate complaints process.

Collect evidence promptly because second-level screenshots, comparison quotes and complete order history may be difficult to reconstruct a week later. No verified sample or success rate is available here for how completeness affects complaint outcomes. Check applicable response deadlines and escalation routes in the broker-dispute article and the relevant account/jurisdiction.

March 2020 COVID stress: an ecosystem under pressure

The week of 9–13 March 2020 provides historical context for extreme FX conditions and unstable cross-asset relationships. The following describes types of retail-infrastructure stress; the numerical magnitudes are illustrative thresholds, not measured observations. The source provides no verified sample, count or data source for these magnitudes:

• CFD order-flow surges modeled at 5–8 times ordinary levels, with intermittent platform delays of tens of seconds during London and New York hours.
• Exotic-pair spread dislocation, including USD/MXN, USD/ZAR and USD/TRY, modeled at 8–15 times normal and possible temporary unavailability.
• Stop and exit execution disruption, modeled from a few to several dozen pips even on major pairs. Distinguish stop-market slippage from the applicable limit/TP order terms rather than assuming all order types permit the same adverse fill behavior.
• Index CFDs, including SPX, NDX and DAX, with illustrative minute-long Off quotes windows during New York trading.
• Margin changes: the source describes some retail providers raising requirements at short notice, requiring additional collateral or position reduction. Account-specific notices establish what actually applied.

System lesson: extreme market stress can pressure the platform, broker, liquidity providers and even information services simultaneously. The source's illustrative stress triggers are VIX above 35, an S&P 500 daily move above 3%, or bond-market liquidity disruption. Its procedure halves normal size and treats each macro release as a potential extreme-execution event. Record those triggers in the plan rather than presenting them as universal safe thresholds.

9. Your own tilt as a crisis — the stop protocol

A retail trader's own tilt can be a recurring operational crisis alongside platform failures and gaps. After losses, a trader may add exposure, distort stop rules, take off-playbook entries or chase P&L. The source has not measured the shape or frequency of that deterioration, so it is neither an established exponential decline nor necessarily your most common incident. Test the hypothesis in your journal: record each trade's sequence number, isolate trades after the day's first loss and calculate their results separately. Your own sample is the evidence, not a universal story about the third trade.

Tilt concerns behavior that can produce decisions with poor expected outcomes, rather than merely feeling an emotion. After two consecutive losses, judgment alone should not be the only decision filter in this procedure. That does not prove the next trade's expectancy is negative for every trader; it means you should not assume a positive value without evidence while decision conditions may be worse. Kahneman discusses broader decision mechanisms in the domain of losses^[7], but “trading tilt” and your third trade are not findings attributed to that book.

Tilt is an operational crisis

It shares several practical characteristics with a platform incident:

• The first 30 seconds matter: do you stop initiating trades, secure existing exposure and close the interface, or click another order?
• Procedure before improvised analysis: stress can distort risk perception. If the broker offers a verified hard daily-loss restriction, use the limit already specified in the plan, 15.2, and check how the restriction works and resets. A new promise of willpower is not an external control.
• Consequences can be disproportionate: an hour of uncontrolled trading can undo a month of disciplined execution.

Early warning signs of tilt

The stop protocol after two and three losses

The source's two-and-three rule specifies a 60-minute pause after two consecutive losses and the end of the session after three. It is an operational rule to put in the plan, not proof that a particular trade count predicts failure.

1. Two losses: record both trades; close or confirm server-side protection for all remaining exposure before closing MT and leaving the room. Set a 60-minute timer. Walk, eat or shower to change context. Then return to the plan and check whether the setup remains valid. If tilt persists, end the session.
2. Three losses: end the session regardless of remaining market time. The procedure does not depend on whether open P&L is positive: close positions or confirm that each has a server-side SL before stepping away. A profitable position without protection is still unattended exposure. Take the specified 24-hour pause and complete the day's post-mortem.
3. Daily limit breached: end the session and take a day off. The number comes from the trading plan, 15.2; this procedure enforces it rather than setting another threshold.
4. Weekly limit breached: stop for the week. Return requires a journal review and resolution of process breaches, not recovering the money first.

Tilt versus a justified system review

A losing streak does not automatically mean tilt. A strategy can meet an unfavorable regime while the operator follows every rule. Use the following illustrative diagnostic patterns:

• Operator-tilt pattern: two or three losses in under two hours, increasing size or weakening stop discipline, decisions within 60 seconds of a signal and journal phrases such as “must win it back” or “I feel this move”. The pattern matters; the durations are not a diagnosis by themselves.
• Possible system/regime pattern: two or three losses over one or two weeks, each within the plan, with server-side stops, intended size, no unauthorized changes and no emotional markers. This alone does not justify changing a setup after one streak.
• Observation count: no trade count automatically turns losses into proof. State n, but assess uncertainty relative to the decision threshold. Keep, observe and disable criteria belong in the strategy audit, 15.10. One losing day is not proof of failure; neither the thirtieth nor three-hundredth trade settles it merely by its ordinal number.
• Operator first: short, emotional sequences call for immediate operational controls, breaks and procedure repairs. Pausing for operational risk does not wait for statistical significance. Calm, plan-compliant losses over a longer period belong in the strategy audit.

Tilt requires a practical external limit, not a judgment about character. The rationale is uncertainty about decision quality in the moment, not a claim to know the expectancy of your third trade. A control that interrupts further orders is more dependable than assuming willpower alone will settle the issue.

Mechanical controls

After the procedure's two-loss trigger, do not assume you are a neutral evaluator of the next decision. Prepared controls can provide the interruption another resolution may fail to deliver:

• A broker-side daily-loss restriction, if offered. Verify availability, account coverage and reset mechanics. Set it from the daily limit in 15.2, not a separate number.
• A deliberate manual login before a new session. Avoid automatic reopening if a short authentication step helps interrupt impulsive re-entry. Preserve secure emergency access to existing positions; this is not a reason to make the backup channel unavailable during an incident.
• Scheduled time off-screen, for example Tuesday and Thursday evenings without trading regardless of perceived signals.
• A Pomodoro rhythm, illustrated as 25 minutes of work and five minutes away, to avoid four hours of fixation on one setup. Secure or appropriately monitor open exposure during breaks.
• Physical disconnection in extreme cases: after breaching the plan's daily limit, turn off the router and leave the phone in another room only after completing the exposure-first sequence below.

10. Pre-mortem and emergency checklists — prepare in 60 minutes

Gary Klein described the pre-mortem technique in Harvard Business Review in 2007^[6]: assume the catastrophe has already happened, then work backward to identify what went wrong. The article relates it to prospective hindsight; it does not establish an origin in medicine or aviation. For a retail trader, a quarterly hour of preparation can reduce the consequences of a real crisis. The BIS guidance provides broader institutional context for business continuity and incident management^[1], rather than a tested promise of a specific retail saving.

A 60-minute pre-mortem, step by step

1. Step 1, ten minutes: scenario table. List what could cause severe damage technically, such as equipment or broker failure; operationally, such as a wrong order or tilt; and through the market, such as a gap, central-bank surprise or macro release. For each, record the safeguard and when it was last tested.
2. Step 2, 15 minutes: classification. Assign each scenario to A, my infrastructure; B, broker failure; C, market event; D, my operational mistake; or E, tilt.
3. Step 3, 15 minutes: a procedure for each category. Write three to five action points covering the first 30 seconds, five minutes and one hour.
4. Step 4, ten minutes: missing resources. Identify what prevents execution: a pinned dealer number, backup account, charged power bank or available mobile credentials. Turn the gaps into a concrete resource list.
5. Step 5, ten minutes: implementation. Choose three to five tasks to complete this week and put deadlines in the calendar.

The nine-field incident card

Complete a card for every incident, including those that end favorably. Each of the nine fields answers a different question; omissions can make the entry uninterpretable a month later.

Emergency-checklist template: scenario reference

This expands the card across scenarios. The state-confirmation column is a required condition: do not proceed with dependent trade instructions until it has an answer. Use the shorter print version below during stress.

Testing the procedure: drill measurements

An untested procedure remains a hypothesis. 15.1 sets the monthly emergency drill's full scope and schedule; do not create a duplicate frequency here. Add these incident-response measurements to that drill:

• Safety condition: test with live positions closed or on demo. Simulating an outage over real open exposure creates another incident.
• During a quiet session, unplug the router's WAN cable while Wi-Fi remains active and start a timer. Turning Wi-Fi off tests a different failure.
• Measure time to (a) authenticated mobile access, (b) reaching the duty dealer and (c) completed closure of a hypothetical 0.01-lot demo position.
• Measure the often-missed fourth interval: time to independently confirmed position state. It determines whether the prerequisite for a backup hedge can be met.
• Use a full-procedure target below 60 seconds, ideally below 30, only as the source's illustrative goal, not an industry standard.
• List failures: missing password, stale dealer number, exhausted cellular-data allowance or another gap. The repair list, not just the stopwatch result, is the drill's output.

Reaction time: targets to test yourself

Reaction time and loss: conceptual illustration

The chart is a qualitative schematic without observations. It illustrates a possible accelerating increase in loss with response delay, not an empirically estimated function or a guarantee that loss increases monotonically. There are no measured points or uncertainty bands because the source has no sample from which to calculate them. Build your own record using time to the first effective action and incident cost in R and PLN, or your clearly identified account currency. A dozen records may suggest a direction but cannot establish a function's shape; scatter at each response time can be wide. Pair, event, broker execution and the fastest usable channel all affect the outcome. Report every estimate with observation count and range.

A ready-to-use crisis-journal template

After every incident, including minor ones, record in a separate file or journal section:

• Date, time and scenario, using pre-mortem categories A–E.
• Trigger: exactly what revealed the problem.
• Reaction time: seconds to each key action.
• P&L impact: the loss attributed to the incident under the stated method.
• What worked: procedure elements completed effectively.
• What failed: missing resources and procedure gaps.
• Corrective actions: concrete infrastructure or procedure changes and deadlines.

The crisis journal is a separate data layer, not a replacement for the trade journal. It identifies whether losses arise from equipment, broker, execution, tilt or missing procedures. After a year, you can rank your own scenarios rather than adopting somebody else's ranking. Repair the two leading causes instead of expanding an eight-page checklist that is unusable under stress.

Printable A4 checklist: the stress version

The reference table above is useful for learning but too long for an incident. Put this shorter version under the monitor: six time-ordered actions and six fields completed beforehand. The timings are targets, not reasons to delay a confirmed urgent protective action.

• □ t = 0–5 seconds: check account-risk percentage and server-side stops through desktop if responsive.
• □ t = 5–15 seconds: open authenticated mobile and dial the dealer in parallel, using phone and independent cellular internet.
• □ t = 15–30 seconds: classify local, broker or market failure; apply the operational-error or tilt procedure if relevant.
• □ t = 30–60 seconds: confirm position state; only then issue dependent close/hedge instructions or carry out the relevant phone/mobile/session-stop action. Act earlier if state is already confirmed and the plan requires it.
• □ t = 1–5 minutes: save timestamped screenshots and start the crisis-journal entry.
• □ t = 5–30 minutes: reconcile position state across all three available channels.
• □ Broker/dealer phone number: ___________________
• □ Primary account login identifier: ___________________
• □ Backup account login identifier: ___________________
• □ Backup-connection switch time measured in the last drill: ___________________
• □ Daily stop-loss limit from the plan: ___________________
• □ Default: “If position state is unknown, send no trade instruction; confirm through channel 2 or 3.”

Prevention ROI: when infrastructure pays for itself

Historical source scope: the SNB announcement of 15 January 2015 confirms discontinuation of the 1.20 minimum-rate policy, not the table's synthetic minute path. FXCM's 19 January SEC exhibit confirms USD 300 million in financing; it does not establish a FXCM bankruptcy. The Bank of Japan describes its agency role for the Ministry of Finance. Procedure alarms require their own calibration.