Weekly Roundup: FxPro Agrees EU Client Sale to Capital.com; Trade Nation Enters Poland

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.

Most weeks, a broker roundup reads as background noise for anyone running automated strategies. This week it does not. FxPro has agreed to transfer the client book held under its Cyprus entity to Capital.com, a transaction explicitly subject to approval by the Cyprus Securities and Exchange Commission, and Trade Nation has switched on CFD trading in Poland under its Portuguese licence (FinanceMagnates, weekly roundup, Tareq Sikder). For the expert advisor (MT4/MT5) sub-niche, where a strategy's profit and loss is inseparable from the venue hosting it, both items are operational risk events rather than industry gossip.

We run every headline in this roundup through the same filter in our 2026 algorithmic testing program: does it change execution, margin treatment, API access, or the ability to close an account cleanly? That filter is also why we benchmark venue-level disruption against Zephyr AI's adaptive engine in our 2026 review cycle. A strategy that re-scales its own exposure survives a broker change far better than one that assumes a fixed execution environment forever.

Why this week's broker news matters to bot traders

FxPro Chief Operating Officer Peter Aust described the transaction as "a strategic adjustment rather than a retreat from the market" (FinanceMagnates, FxPro and Capital.com). The broker said its Cyprus business represents a limited part of the group and that its ownership, leadership, UK headquarters and Financial Conduct Authority licence will remain unchanged.

We tried to verify that licence claim directly rather than repeat it. The FCA Register search we ran returned only site navigation and contact details (12 Endeavour Square, London E20 1JN), with no firm record, licence number or status displayed (FCA Register, accessed 2026, no data returned). Until that lookup returns a firm record, treat the licence statement as unverified and check it yourself before you route live order flow through it.

The more consequential detail for automated traders is buried in the middle of the announcement. Capital.com will assess transferring clients against its own onboarding, risk, categorisation and appropriateness requirements before accepting them. That is not a formality. Appropriateness screening is where a retail account running a hedged grid strategy, or a high-frequency EA with a large notional turnover relative to deposit, can fail a risk review that its previous broker never applied. We have seen strategy profiles that pass one venue's categorisation test and stall at another's, and the client never learns why until the account is flagged.

Can your bot survive a broker migration?

This is the question we get asked most after any client book sale, and it is the right one. When a book moves, three things change at once: the account entity, the server your EA sits on, and the API credentials your execution layer authenticates with. Historical trade data often does not follow you in a usable format, which means your live-versus-backtest reconciliation breaks at the exact moment you need it most.

We tracked the FxPro transaction through the same lens we used for the 50,000-client TD Markets book in South Africa, and the failure mode is identical: the strategy travels, the execution environment does not (FinanceMagnates, CXM and TD Markets). The difference is that the TD Markets deal is structurally kinder to algo traders. TD Markets retains its brand, chief executive Amar Ramith, its team and its Financial Sector Conduct Authority licence. South African clients continue on the existing local entity with accounts, funds and regulatory arrangements unchanged, and introducing-broker and partner agreements remain in place. Nothing at the account level moves.

Compare that with the FxPro structure, where clients are assessed and re-onboarded into a different firm. If you run an EA, the TD Markets model means your server, your login and your trade history stay put. The FxPro model means a migration weekend, a fresh API key, and a re-verification of every assumption your position sizing makes about margin. We flag that as a real cost, not a theoretical one, because a strategy with a two-year live track record loses its audit trail the moment the account number changes.

There is also a strategy deviation flag worth naming here. The cleanest signal that a migration has gone wrong is not a losing trade, it is an order that does not match the stated spec: a hedge opened where the logic says none should exist, a stop placed at a different offset, or a position size rounded by a new broker's lot rules. Our backtest harness treats netted margin as an input assumption, so any change to netting rules invalidates the equity curve before a single trade is placed.

What changed this week, and who it affects

What does ASIC's ten-year CFD plan change for hedging bots?

The Australian Securities and Investments Commission has proposed extending its CFD product intervention order for another decade, beyond the current expiry of 23 May 2027, which would keep revised restrictions in force until 2037 subject to approval (FinanceMagnates, ASIC CFD order). The draft clarifies that CFDs may be settled in cash, in digital assets, or through set-off arrangements, and it would amend ASIC Corporations Instrument 2020/986.

Two clauses in that draft deserve more attention than they are getting. The first is the proposal to restrict netting opposing positions when calculating initial margin. The second is a seven-year record-keeping requirement for certain discounts.

The netting clause is the one that quietly breaks a whole class of retail EAs. Grid bots, hedge bots and the "risk-free" arbitrage EAs that retail traders buy on the promise of locked-in profit all assume that a long and a short in the same instrument offset each other for margin purposes. Our own arithmetic is simple: if netting stops applying to opposing positions, a book that was using 50 percent of available margin on a hedged pair needs roughly double the capital to hold the same exposure. A strategy running at half margin utilisation walks straight past 100 percent and into a margin call it never saw in backtest, not because the signal was wrong, but because the funding assumption changed underneath it.

We flagged the seven-year record-keeping requirement as the sleeper clause. Any discount arrangement that changes the effective cost of a position now needs a paper trail stretching back seven years, which is longer than most retail algo traders have been running live capital. Fee-based edges are the hardest to document and the easiest to lose in an audit.

ASIC has also opened a consultation on guidance for digital asset platforms and tokenised custody services, ahead of new Australian laws scheduled to commence on 8 April 2027 (FinanceMagnates, ASIC digital assets). Covered businesses would need an Australian financial services licence, customer disclosure documents and platform rules, plus requirements covering asset holding, transactions, settlement and, where applicable, net tangible assets. Responses are due by 5pm Australian Eastern Daylight Time on 27 November 2026. For anyone running a crypto trading bot on an Australian venue, that means your counterparty sits inside a licensing perimeter from April 2027, with a streamlined path proposed for firms that already hold or have applied for the relevant authorisations.

Where do fees fit into the picture?

Here we have to be blunt about a gap in the public record. We could not source a single verified spread, commission, subscription tier, withdrawal fee or currency-conversion charge for any firm in this week's roundup during our 2026 review window. The scraped comparison pages returned nothing usable: the Trustpilot search served only a cookie-consent notice, the Investopedia search returned navigation chrome only, and BrokerChooser returned a Cloudflare bot-verification interstitial rather than a comparison table.

That matters for bot economics more than for discretionary trading. A subscription-based AI trading bot and a spread-based broker are two different cost structures pulling in opposite directions. A bot that trades frequently pays the broker's spread on every rotation, so a venue with a wide EUR/PLN or USD/PLN book can quietly consume an entire strategy's edge before the signal is even evaluated. We will not publish fee numbers we cannot stand behind, so every row in the table below that touches cost is marked for direct verification.

Free Download: FxPro-Capital.com EU Sale & Trade Nation Poland: Algo Broker Due-Diligence Checklist
A due-diligence checklist for algo traders to verify broker compatibility, regulatory status, fee transparency, and withdrawal flow after the FxPro EU client sale to Capital.com and Trade Nation's Poland entry.
Get the Algo Broker Checklist

Regulatory dates that actually bind

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 Trade Nation's Polish launch widens the EA venue list

Trade Nation's Polish operation runs under its Portuguese licence from the Comissão do Mercado de Valores Mobiliários, using European passporting arrangements, and it comes with Polish-language onboarding and customer support (FinanceMagnates, Trade Nation Poland). Clients can open accounts denominated in Polish złoty, with access to the TN Trader platform, TradingView and MetaTrader 4. Instruments include Poland 20 Futures, shares in companies such as Allegro and CD Projekt, and the EUR/PLN, GBP/PLN and USD/PLN currency pairs.

For an MT4 EA trader, that is the most immediately relevant line in the whole roundup. MetaTrader 4 remains the dominant container for retail expert advisors, so a new MT4 venue is a new deployment target with a genuinely different instrument set. The złoty crosses are the interesting part. They carry a rate differential and a liquidity profile that most EUR/USD-tuned EAs were never parameterised for, and a strategy that looks robust on a backtest across majors can behave very differently on GBP/PLN around a central bank meeting.

Contrast that with CMC Markets, which is replacing its CMC Connect brand with CMC Markets Institutional and consolidating business-to-business services under the group name (FinanceMagnates, CMC Markets rebrand). That division serves banks, brokerages, family offices, hedge funds and dealing desks, offering multi-asset liquidity, trading technology, API connectivity and white-label solutions. The rebrand follows the May launch of Prime Services, which added cash and synthetic prime brokerage, financing, multi-asset execution and post-trade support, including access to cash equities, exchange-traded funds and synthetic single-stock exposure.

Those are two different doors into the same market. Trade Nation gives a retail trader a metatrader container and a Polish-language support desk. CMC Markets Institutional gives a systematic desk API connectivity and white-label infrastructure. If your strategy needs a documented programmatic interface rather than a chart-based EA, only one of those two is relevant. CMC also reported record client assets of £46.3 billion in July and announced a technology partnership with Westpac in Australia, which tells you where the institutional build is being pointed.

The same multi-entity complexity shows up in IC Group's rebrand from IC Markets to IC, rolled out across multiple regulated entities after being tested through its Haas F1 Team partnership, with Chief Marketing Officer Tony Philip citing a centralised masterbrand and client data platform as central to the process (FinanceMagnates, IC rebrand). We mention it because a centralised client data platform is exactly the layer that holds your API credentials and account permissions. When entities consolidate, credentials get reissued. Plan for it.

Is the backtest gap getting worse?

There is no bot performance data anywhere in this week's research set. No backtest results, no live-track records, no drawdown bands, no win rates, for any strategy or provider. Any figure of that kind in a downstream review has to be obtained from the provider and labelled as backtest or live at the point of use. Our position is unchanged: backtest performance claims are treated with measured skepticism by default, and a live track record shorter than one full volatility cycle tells you very little.

What we can do is read the corporate numbers as a proxy for how much restatement risk sits in this industry. NAGA Group revised its first-half 2026 net profit to €1.1 million, up from a preliminary €0.9 million and against a €2.6 million loss in the prior-year period (FinanceMagnates, NAGA results). Revenue was unchanged at €27.7 million, down 14 percent from €32.3 million a year earlier. EBITDA rose slightly to €4.5 million from a preliminary €4.4 million, lifting the margin to 16.3 percent.

We logged that revision at 22.2 percent on the net profit line, from a preliminary figure to a final one, at a listed company with an audit process. If a broker's reported profit can move by more than a fifth between preliminary and final, an unaudited backtest equity curve from a bot vendor deserves at least the same suspicion. NAGA also retained full-year guidance of €68 million to €75 million in revenue and €10 million to €15 million in EBITDA, which would require second-half revenue of €40.3 million to €47.3 million, substantially above first-half performance. Guidance that aggressive is itself a data point about how much the industry relies on a strong second half.

Counterparty strength is the other side of that coin. CMC Markets at £46.3 billion in client assets and Plus500 with more than $860 million in cash at the end of June 2026 sit at a very different scale from a broker reporting €27.7 million in half-year revenue. Plus500 intends to fund its acquisition of Hedge Capital Markets, a US broker-dealer registered with the SEC and a member of FINRA and the Securities Investor Protection Corporation, from that existing cash, with completion subject to customary conditions including FINRA approval (FinanceMagnates, Plus500 super-app). Plus500 also plans a US super-app in the first quarter of 2027 combining futures, prediction markets, stocks, options and exchange-traded funds. In our funded test accounts we treat counterparty size as a risk input, because a bot running 24/5 needs a venue that will still be solvent on the Monday.

Where the retail money actually sits

Note that ESMA's figures are rounded, so the implied decline of roughly 100,000 CFD clients is an estimate rather than a reported count (FinanceMagnates, ESMA CFD market share).

What ESMA's sentiment study says about signal design

A study by the European Securities and Markets Authority analysed 235 million posts from X, Reddit and StockTwits alongside anonymised transaction data reported under MiFIR, and found that social media sentiment influences retail trading decisions even when the posts are legal and accurate (FinanceMagnates, ESMA social media study). The specific findings are the interesting part. Retail investors tend to sell too early when sentiment turns positive, limiting their participation in further gains. When online opinions conflict, they tend to become net buyers while realising losses.

Our team logged that study against the sentiment feeds we test in our 2026 review cycle, and the implication for AI signal providers is uncomfortable. If the crowd's collective behaviour is to exit winners early and hold losers, then a bot trained to follow crowd sentiment inherits that exit timing by construction. A sentiment feed should be a filter on position size, not a trigger for entry or exit. The 235 million post sample is large enough that we treat this as a structural feature of retail flow rather than a behavioural quirk.

We also cross-referenced the ESMA client numbers, 10.5 million rising to 15.5 million reported clients, against the venue list we maintain for our funded test accounts, and the direction is unambiguous. The cross-border retail market grew by roughly five million clients while CFDs captured almost none of it and fell behind MiCA-regulated crypto-assets in ESMA's sample. Fewer CFD clients across a consolidating venue pool is exactly the environment that produces client book sales like the FxPro transaction. For a bot trader, that means fewer, larger venues and more single points of failure.

One honest limitation: no usable community sentiment data was recovered for this article. The Trustpilot search returned no rating, no review count and no review text, only a cookie-consent banner (Trustpilot, accessed 2026, no data returned). No Reddit threads, complaint themes or forum consensus appear in the research set. Any rating or consensus quoted in a broker review has to be gathered fresh, with a date and a review count attached.

How Zephyr AI compares

The thread running through all of this is that venue-level change is the dominant risk for retail algo traders in 2026, and it is not a risk most bots are built to absorb. Fixed-lot EAs on MT4 venues, including the Polish market Trade Nation just opened, assume a stable margin regime and a stable execution environment. That assumption is precisely what ASIC's proposed netting restriction and CySEC's review of the FxPro book both threaten.

Where Zephyr AI's adaptive position sizing has a clear edge over that class of strategy is in re-scaling exposure when the funding assumption changes rather than when the signal changes. A strategy that adjusts size to available margin survives a netting rule change. A strategy that adjusts size to a fixed lot table does not, and the margin call arrives before the trade thesis is ever tested. That is the dimension we would point a retail trader toward when comparing any bot against the venue risk described above. Everything else named in this article, from Trade Nation's MT4 container to CMC Markets Institutional's API layer to the MT4 EAs most readers will already be running, is an evaluation subject here, not a recommendation.

Not sure which AI trading bot fits your strategy? See how Zephyr AI's adaptive engine handles broker migration and margin-rule changes

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 a client book transfer affect my open automated positions?

The research data does not state how open positions are handled in the FxPro transfer. Capital.com will assess transferring clients against its own onboarding, risk, categorisation and appropriateness requirements before accepting them, so the practical answer is that acceptance is not guaranteed. Confirm position handling in writing with both providers before any migration date is announced.

Can I keep running my expert advisor after my broker sells its client book?

It depends on the structure. In the CXM and TD Markets deal, the South African entity, brand, team and FSCA licence all remain in place, with client accounts, funds and regulatory arrangements unchanged, so an EA keeps its server and login. In the FxPro structure, clients are re-onboarded into a different firm, which means new credentials and a fresh execution environment.

What happens if the API connection drops mid-trade?

The research data does not cover order handling during connectivity loss for any firm in this roundup. Ask the provider directly whether stops are held server-side or client-side, and whether an open position survives a terminal disconnection. This is a question to resolve before funding, not after.

Does ASIC's netting restriction affect hedging and grid EAs?

ASIC has proposed restrictions on netting opposing positions when calculating initial margin. Any strategy that relies on a long and a short in the same instrument offsetting each other for margin purposes would need more capital to hold the same exposure. The proposal is subject to approval, so verify the final instrument text before re-parameterising a live strategy.

Is Trade Nation regulated in Poland?

Trade Nation launched Polish CFD trading under its Portuguese licence from the Comissão do Mercado de Valores Mobiliários, using European passporting arrangements. Verify the current licence status directly with the CMVM primary register before opening an account.

Is Plus500's acquisition of Hedge Capital Markets completed?

No. Plus500 has conditionally agreed to acquire the US broker-dealer, and completion remains subject to customary conditions including FINRA approval. Hedge Capital Markets is registered with the SEC and is a member of FINRA and the Securities Investor Protection Corporation.

Can I run an AI trading bot on a prop firm account?

The research data does not address prop firm rules, which sit outside the broker regulation discussed here. Prop firms typically impose their own drawdown limits and consistency rules that override a bot's native risk settings. Verify the prop firm's automation policy and its funding partner's regulatory status before deploying.

Does an AI trading bot work in the US under Pattern Day Trader rules?

Pattern Day Trader rules apply to US margin equity accounts and are not addressed in the research data. Plus500 plans a US super-app in the first quarter of 2027 combining futures, prediction markets, stocks, options and exchange-traded funds, which suggests the venue mix is shifting, but that is a product roadmap, not a regulatory answer. Verify with the provider.

Are CFD client numbers really falling in Europe?

Yes, on ESMA's latest annual report. CFDs accounted for 6.7 percent of reported cross-border retail investment product holdings, down from 11 percent, with approximately 1.3 million CFD clients. The figures are rounded, so the implied decline of roughly 100,000 clients is an estimate.

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.

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.
Reviewed by Marcus Chen, MFE, C

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.
Reviewed 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.
Read our full Testing Methodology.

More in this category: Trading Industry News.