OANDA Japan Accelerates MT4 Retirement Amid Security Breaches
OANDA Japan Accelerates MT4 Retirement, Citing Security Breaches and Legacy Vulnerabilities
Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.
OANDA Securities has pulled its Japanese MetaTrader 4 shutdown forward by two weeks, to November 14, 2026, and the broker is unusually direct about the reason. Its client notice cites recent security breaches and mounting cyber risk inside external VPS environments, on top of the platform's age and the end of official support and maintenance from MetaQuotes (Finance Magnates).
For a casual reader, that is a cybersecurity story. For anyone running an expert advisor (EA) on MT4, it is a strategy-continuity story, and that is the lens we use here. We re-implement EAs in MQL5, push each port through our 2026 algorithmic testing framework, and log the exact points where live behaviour drifts from the published specification. In our current review cycle we have also benchmarked against Zephyr AI's adaptive engine, which matters because a forced platform migration is precisely the kind of event that breaks brittle rule-based systems and rewards adaptive ones.
The core problem is easy to state and expensive to ignore. OANDA warns that MT4 EAs, indicators and templates cannot be used on MT5, so an account transfer does not reproduce a customer's trading setup (Finance Magnates). Strategies that ran untouched for years now have to be rewritten, re-validated and re-deployed on a different execution engine, and the November 14, 2026 deadline gives most of them less than a month to do it.
What does the MT4 shutdown actually mean for EA traders?
An expert advisor is a set of rules compiled into a program that the platform executes on your behalf. On MT4 that program is written in MQL4 and calls MT4's order functions. On MT5 it is written in MQL5 and calls a different API. The two are not drop-in compatible, and OANDA says so plainly: MT4 EAs, indicators and templates cannot be used on MT5. That single sentence is the most expensive line in the entire notice.
The consequences stack up in a specific order. First, your EA stops running at the hard deadline. Second, any open positions on the MT4 Tokyo server are scheduled to transfer automatically into MT5 Standard Plan accounts over November 14 to 15, 2026, which means the position moves but the logic managing it does not (Finance Magnates). Third, you have to rebuild the strategy, and rebuilding is not copying. Order-fill logic, slippage assumptions and even how the platform nets or hedges positions can differ between the two environments.
We have seen this pattern across our own ports. When we move a rule-based EA from MQL4 to MQL5, the translation is rarely mechanical, and the areas that break are the ones the original spec never documented: how the EA handles a partial fill, a rejected order, or a requote at the Tokyo open. That is why we treat any MT4 to MT5 port as a brand new strategy rather than a migration, and why we re-run the full validation suite instead of trusting a line-by-line diff.
Why is a VPS breach a trading strategy problem?
Most retail EAs do not run on a home PC. They run on a virtual private server, because a VPS stays online when your laptop sleeps. That convenience is also the attack surface. The OANDA decision follows unauthorised access involving Onamae.com Desktop Cloud, a VPS operated by GMO Internet Group and used by retail traders to host MT4 and MT5 (Finance Magnates).
The scale is the part that should hold your attention. Up to 26,739 servers may have been exposed, which forced major brokerages including OANDA to issue urgent security alerts, place transaction restrictions on affected accounts, and tell clients to change passwords and enable two-factor authentication (Finance Magnates). Five Japanese FX brokers later confirmed that hackers placed trades after the server-host breach (Finance Magnates).
Read that again as a strategy risk rather than a login risk. An attacker with access to the VPS does not need to drain your balance directly. They can place trades, and a live EA will manage those trades according to its rules, because from the EA's point of view an open position is just an open position. That is a strategy-integrity failure, and it is the dimension most migration write-ups skip entirely.
| Item | Detail |
|---|---|
| VPS provider named in the incident | Onamae.com Desktop Cloud, operated by GMO Internet Group |
| Servers potentially exposed | Up to 26,739 |
| Broker response | Urgent security alerts, transaction restrictions on affected accounts, password and 2FA guidance |
| Who secures the VPS | The customer, per OANDA, because its security measures do not extend to third-party servers |
| New login requirement | Google Authenticator OTP-based MFA for MT5 from October 31, 2026 |
OANDA's position on third-party servers is blunt: customers using external VPS services must secure those environments themselves. We treat that as a hard boundary. Your broker secures the trade server, not the box your EA lives on, and the 26,739-server exposure figure shows how quickly that boundary becomes a real loss event rather than a theoretical one.
The migration timeline, date by date
The dates matter because each one changes what your automation can actually do.
| Date | Event | What it changes for EA traders |
|---|---|---|
| March 2026 | Original MT4 retirement announced | First signal that MT4 would end in Japan |
| October 29, 2026 | Announcement linked to the hosting-service incident | Security context behind the accelerated timeline |
| October 31, 2026 | Google Authenticator OTP-based MFA mandatory for MT5 login | Automated logins need a human OTP step |
| November 14, 2026 | MT4 termination date, moved forward by two weeks | Hard stop for MT4 access |
| November 14 to 15, 2026 | Tokyo server positions and balances transfer to MT5 Standard Plan | Positions move; the EA logic does not |
Free Download: OANDA Japan MT4 Retirement: Algo-Bot Migration & Security Checklist
A due-diligence checklist to assess whether your MT4-based bot can safely migrate, stay compliant, and avoid OANDA Japan's legacy platform and security risks.
Get the MT4 Migration Checklist
The MFA line deserves a second look. OANDA made Google Authenticator OTP-based MFA mandatory for MT5 login from October 31, 2026, following an October 29 announcement linked to the hosting incident (Finance Magnates). Mandatory MFA is good security and awkward for fully hands-off automation. If your workflow assumes a headless login, you now have a manual step in the chain, and you should confirm with OANDA how that interacts with the way your EA reconnects after a restart. That is a two-week gap between the MFA requirement and the MT4 shutdown, which is not much time to test a new login path.
What survives the move from MT4 to MT5?
Not as much as the word "migration" implies. Here is the honest scorecard, limited strictly to what the source material states.
| Dimension | MT4 | MT5 | Source status |
|---|---|---|---|
| EA compatibility | MT4 EAs run natively | MT4 EAs cannot be used | Stated by OANDA via Finance Magnates |
| Indicators and templates | MT4 format | Not transferable | Stated by OANDA via Finance Magnates |
| Official support | Ended by MetaQuotes | Active | Stated via Finance Magnates |
| Alternative access for NY-server clients | MT4 access ends | fxTrade or TradingView | Stated via Finance Magnates |
| Order handling model | Verify with provider | Verify with provider | Not in source data |
| Backtest engine behaviour | Verify with provider | Verify with provider | Not in source data |
Two rows say "verify with provider" on purpose. The source material does not give us order-handling or backtest-engine specifics, and we will not invent them to fill a table. If a vendor tells you the port is seamless, ask for the two documents that prove it: a written order-handling comparison and a side-by-side backtest on the same data window. A vendor who cannot produce both is describing a hope, not a process.
Not sure which AI trading bot fits your strategy? Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026
This link is an affiliate partnership - see our editorial policy for details.
How big is the backtest gap between MT4 and MT5?
This is where the marketing and the mathematics part ways. An MT4 backtest and an MT5 backtest of the "same" strategy are not the same experiment, because the two platforms use different testing engines, different tick modelling and different order-execution assumptions. The source material confirms the compatibility break but does not publish a quantified performance gap, so we will not hand you a fabricated percentage.
What we can say, from our own re-implementation work, is directional and consistent: when we port a rule-based EA between MQL4 and MQL5, the entry logic usually survives and the exit logic usually does not, because exits depend on the platform's fill behaviour. That is a structural gap, not a rounding error. If you want a number for your specific EA, the only honest answer is to run the port yourself on an identical data window and compare, and to treat any vendor who skips that step as selling optimism.
The contrast with adaptive systems is worth stating plainly. A rule-based EA that hard-codes MT4 order assumptions has to be rewritten for MT5. An adaptive engine that sizes positions from current volatility and current fill quality absorbs the platform change as an input rather than a rewrite. In our 2026 review cycle, Zephyr AI's adaptive position-sizing framework is the example we use for that class of design, and it is the reason we treat "platform-agnostic" as a genuine engineering property rather than a marketing label.
What are the drawdown and risk implications?
A forced migration is a drawdown event waiting to happen, and it usually arrives in three ways. The first is gap risk: if your EA is mid-trade when MT4 access ends, the position transfers to MT5 Standard Plan over November 14 to 15, and the logic that would have managed it is no longer running. The second is execution drift: a freshly ported EA trades differently for a while because its assumptions no longer match the venue. The third is operational: a mandatory MFA step from October 31 can leave an EA offline after a restart until a human completes the login.
The source material does not give us drawdown figures for OANDA's client base, and we will not manufacture them. But the risk framing is clear enough. A migration window is the moment to reduce position size, not increase it, and to run the ported strategy in a monitored mode before you let it size up. Rule-based EAs are especially exposed here because their risk logic is fixed at compile time. An adaptive approach that scales exposure to realised volatility gives you a dial to turn during the transition, and that is a concrete difference rather than a slogan. For comparison, our own funded-account testing of adaptive frameworks has consistently shown tighter peak-to-trough behaviour than fixed-lot rule-based EAs on the same strategy class, though the exact figures vary by strategy parameters and should be verified with the provider.
Fees, subscriptions, and the economics of migrating
Migration is not free, even when the broker charges nothing for it. The costs hide in three places. First, developer time: rewriting an EA from MQL4 to MQL5 is skilled work, and if you bought the EA rather than built it, the original vendor may charge for a port or refuse outright. Second, re-validation: a ported EA needs a fresh backtest and a monitored live period before you trust it with real size. Third, opportunity cost: while you are rebuilding, the strategy is not earning.
The source material does not publish OANDA's fee schedule for the MT5 Standard Plan transfer, so treat any specific number you see elsewhere as unverified. What the source does confirm is the structure: Tokyo server positions and balances transfer automatically, balance-only customers are asked to move funds before shutdown, and there is a transfer fallback for remaining balances (Finance Magnates). That looks like a no-fee transfer on the surface, but the real bill is the rebuild, and the rebuild is where the money quietly goes.
This is where subscription-based AI platforms and one-off EA purchases diverge. A one-off EA purchase gives you a file you now have to maintain yourself. A subscription platform that keeps the execution layer under its own control absorbs platform changes for you, at the cost of a recurring fee. Neither is automatically better; the question is whether you want to own the maintenance burden or rent it. Zephyr AI sits in the second camp, and the relevant comparison point is fee-structure transparency: a recurring fee you can cancel cleanly versus a sunk cost in a file that just stopped working.
Can you actually exit a legacy EA cleanly?
Disengagement is the most under-tested part of any automation, and this event is a live stress test of it. The clean exit has a defined order: flatten or consciously carry open positions, remove the EA from the chart, confirm no pending orders remain, and only then move funds. OANDA's transfer of Tokyo server positions into MT5 Standard Plan means the platform will carry your positions for you, but it will not carry your EA's exit rules.
We track disengagement as a deviation metric, and across our 2026 test cohort the friction almost always sits in the same two places: pending orders that nobody cancelled, and a strategy that keeps re-entering after you thought you had stopped it. The question we ask of any system is simple. If the platform disappears tomorrow, how many manual steps does it take to get flat and stay flat? A rule-based MT4 EA scores badly here, because its exit logic is welded to a platform that is being retired. A hosted system that exposes a one-click stop and a documented withdrawal path scores well, and the difference shows up on the day you actually need it.
Is OANDA Japan regulated, and does it matter for your EA?
Regulatory status matters for two reasons in an EA context: whether your funds are protected if the broker fails, and whether the broker is accountable for the security promises it makes. OANDA operates across multiple jurisdictions, and the source material does not give us a specific licence number for the Japanese entity, so we will not assert one. Verify the current status directly with the provider's primary regulator, and use the primary registers rather than a secondary summary.
For our due-diligence checklist we run the FCA Register and the ASIC registers as the two primary lookups, and for a Japanese account the relevant oversight is domestic. The honest takeaway is not "OANDA is or is not regulated." It is that a broker telling you to secure your own VPS is drawing a liability line, and you should know exactly where that line sits before you trust it with an automated strategy.
How Zephyr AI Compares
The MT4 retirement exposes a structural weakness in rule-based EAs that has nothing to do with OANDA specifically. When the execution platform changes, a fixed-rule EA has to be rebuilt, and the rebuild is where strategies quietly die. The concrete dimension where an adaptive framework wins is platform independence. Where a ported MT4 EA needs a fresh backtest and a monitored live period before it is trustworthy again, an adaptive engine that reads current volatility and current fill quality treats the new venue as an input and keeps sizing positions on the same logic.
That is the comparison we would put in front of a trader facing this migration. One path is a file that stopped working on November 14, 2026, plus a rebuild you have to fund and validate yourself. The other path is a system that keeps running while the plumbing underneath it changes. We are not claiming the second path is risk-free, because it is not, and its own performance figures should be verified with the provider. We are claiming it is structurally better suited to a world where brokers retire platforms on two weeks' notice, and that structural fit is the thing worth paying attention to.
Not sure which AI trading bot fits your strategy? Explore Zephyr AI's adaptive trading engine for 2026
This link is an affiliate partnership - see our editorial policy for details.
Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026
Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026
This site contains affiliate links. We may earn a commission if you sign up through our links, at no extra cost to you. This does not affect our editorial independence.
Frequently Asked Questions
Does this MT4 shutdown affect traders outside Japan?
The source material describes OANDA Securities' Japanese MT4 service specifically, with a termination date of November 14, 2026. Traders on other OANDA entities and other brokers should check their own platform notices directly, because the timeline here is Japan-specific.
Can I keep running my MT4 expert advisor after November 14, 2026?
On OANDA Japan, no. MT4 access ends on November 14, 2026, and the broker states that MT4 EAs, indicators and templates cannot be used on MT5. You would need to rebuild the strategy in MQL5 or move to a different broker that still supports MT4.
Will my open positions survive the transfer to MT5?
OANDA says Tokyo server positions and balances are scheduled to transfer automatically into MT5 Standard Plan accounts over November 14 to 15, 2026. The positions move
Written by Marcus Chen, MFE, CMT - MFE (UC Berkeley Haas, 2018) and CMT (Levels I-III, 2020). Six years quantitative researcher at a Chicago prop firm before joining BTR to lead algorithmic-strategy review.
Reviewed by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.
Read our full Testing Methodology.
More in this category: MetaTrader Expert Advisor Reviews.