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.

mBank Quits Forex, Leaving Two Bank-Owned Brokers in Poland's Growing CFD Market

When a bank-owned brokerage walks away from a product line that is still growing, algo traders should pay attention. That is exactly what happened in Poland this autumn, and it matters to anyone running an algorithmic trading platform or an expert advisor across European CFD venues. mBank's brokerage arm stopped signing new mForex agreements on September 3 and, from October 1, will let clients only close existing positions, according to the Polish daily Parkiet, as reported by Finance Magnates. The unit joins a list of bank-affiliated houses that have already exited, including Erste's brokerage, BM PKO BP, BM ING BSK, and Noble Securities.

For our 2026 review cycle, we benchmarked execution venues and automation layers against the Ellington AI trading platform, and this story is a useful stress test of a thesis we keep returning to: the venue is the strategy. An expert advisor or algorithmic trading platform is only as durable as the broker underneath it. When a bank-owned house pulls the plug, every automated strategy parked on that venue has to be re-homed, re-tested, and re-validated. That is a portfolio-level event, not a footnote.

What mBank's exit actually changes

The headline number is the one that makes the exit strange. The Polish Financial Supervision Authority (KNF) counted 369,737 active clients trading OTC derivatives through domestic brokerages in 2025, a 50% rise from 246,826 a year earlier. Polish residents made up 186,372 of those clients, up 59%. Among the European markets compared when the Polish total stood near 175,000 in 2023, only the UK had more active retail traders.

So the market is expanding while a bank-owned incumbent retreats. mBank said it wants to focus on products that help customers build savings and capital for the long term. That is a strategic pivot away from leveraged products, not a statement about demand. For a systematic trader, the practical consequence is venue concentration. If you were running an EA on an mForex account, you now have a forced migration deadline and a strategy that needs to be re-validated on a new feed.

We have seen this movie before. When BM PKO BP and BM ING BSK stepped back, the clients did not disappear. They moved. The question for an algo trader is whether the destination venue preserves the execution characteristics the strategy was calibrated against.

Which brokers are left, and what they offer

Two traditional bank-owned houses remain in the Polish CFD market: Dom Maklerski BOŚ and Alior Bank's brokerage unit. DM BOŚ is the more explicit about staying. Sebastian Zadora, director of its financial instruments sales department, told Parkiet that CFDs are "a profitable business and a segment we definitely want to keep developing." The firm has added MetaTrader 5, widened its crypto CFD range, and expanded leveraged trading in Polish shares, including short selling. BM Alior Banku declined to comment on the future of its forex offering.

That asymmetry is worth noting. One bank-owned broker is leaning in with platform upgrades; the other will not say. For an algorithmic trader, a broker that will not comment on product continuity is a venue risk you have to price.

Table note: Statuses reflect reporting from Parkiet via Finance Magnates. Product availability and platform support should be verified directly with each broker.

The XTB row deserves its own paragraph because the "traditional bank-owned" framing leaves out the largest player entirely. XTB is a KNF-licensed brokerage house and it ran 1.235 million of the 3.03 million securities accounts on the Polish market at the end of August, per Central Securities Depository of Poland (KDPW) data cited by Parkiet. It added 53,600 accounts in August, out of 61,500 for the whole market. mBank's brokerage, second with nearly 577,000, added about 5,500.

But here is the number that should interest anyone running automated strategies: only 7% of XTB's new clients in 2025 made a CFD their first trade, yet the instruments generate about 95% of revenue, according to CEO Omar Arnaout, who said in February he wants to cut that dominance from 95% to 70%. That gap between user intent and revenue concentration is the whole cross-selling debate in one statistic.

Why the cross-selling risk matters to algo traders

Zadora told Parkiet that aggressive telemarketing was the industry's problem a few years ago, and that the bigger risk now is selling CFDs to clients who came for ETFs or IKE and IKZE retirement accounts to build long-term savings. If brokers do not keep CFDs clearly separate from simple investment products, he said, the industry may itself prompt tighter regulation.

That is not an abstract compliance concern for systematic traders. Tighter regulation tends to arrive as leverage caps, marketing restrictions, and onboarding friction. In March, the KNF fined XTB 20 million zlotys, and among its findings the regulator said XTB treated experience with simple instruments as enough to assess clients for CFDs between January 2022 and September 2023. When suitability rules tighten, the pool of clients eligible for leveraged products shrinks, and so does the liquidity that automated strategies rely on.

We logged a similar pattern in our 2026 review period across three European venues where suitability rules changed mid-year. The venues that tightened onboarding saw wider spreads during the first hour of the London session, and our mean-reversion test strategies saw their fill quality degrade before the strategies themselves changed a single parameter. The strategy did not break. The venue did.

How big are the losses, really?

The KNF's own numbers are the most sobering part of this story. On the CFD side, 72.2% of active clients closed 2025 with a loss, within the 70.6% to 79.1% range the KNF has recorded since 2021. Their combined losses of 2.68 billion zlotys were nearly four times what winning clients made.

Free Download: Poland CFD Broker Due-Diligence Checklist: Evaluating Bank-Owned vs. mBank-Exit Alternatives for Your AI Bot
A step-by-step checklist to verify regulatory status, fee transparency, withdrawal flow, and bot compatibility when choosing a Polish CFD broker after mBank's forex exit.
Get the Broker Checklist

Table note: Figures as reported by the KNF and cited by Parkiet via Finance Magnates. Loss rates are historical and not predictive of any individual strategy or account.

That 72.2% loss rate is the number every automated strategy operator should sit with. It is not a claim that automation loses money. It is a claim that the retail CFD population, as a whole, is not profitable, and that a bot does not automatically change that. A poorly specified expert advisor on a leveraged CFD venue is just a faster way to arrive at the same distribution.

Registered account totals also overstate activity. Warsaw Stock Exchange data show 444,100 accounts made at least one trade in the first half of 2026, about double the level three years earlier, per Parkiet. An account that trades once is not an active trader, and a bot running on a dormant account is not a strategy.

Table note: Platform capabilities described from public reporting and provider documentation. Performance figures vary by strategy parameters — consult each platform's published metrics.

Not sure which AI trading bot fits your strategy? Try Ellington — The AI Trading Platform for 2026

This link is an affiliate partnership - see our editorial policy for details.

What the UK comparison tells us about venue quality

Among the European markets compared when the Polish total stood near 175,000 in 2023, only the UK had more active retail traders. That is a useful benchmark because it forces a comparison of regulatory posture. The UK operates under FCA supervision, and the FCA maintains a public register where you can verify any firm's permissions directly. We always check the primary register before we place a single automated trade, and we recommend the same discipline for any venue you are considering.

The Polish market's competitive complaint is worth restating plainly. Jakub Paturalski, managing director of XTB Polska, told Parkiet he expects wider interest in financial markets to bring in investors who want to trade part of their capital more actively, and he called for more consistent interpretation and enforcement of the rules so that Polish firms paying taxes at home compete with foreign providers on product and technology, not on differences between supervisors. He did not name any rival.

The rivals he did not name are easy to identify. Revolut serves Polish investors under a Lithuanian license and sits outside the KDPW count. Trade Republic entered Poland on a single German license in September 2025. Both are examples of passporting, where a firm operates in a market under a home-state license rather than a local one. For an algorithmic trader, passporting creates a regulatory edge case that is genuinely under-discussed: your bot may be executing against a venue whose supervisory regime, dispute-resolution path, and compensation scheme are all in a different country from the one you live in. When something goes wrong at 3 a.m. during a volatility spike, you are not calling your local regulator.

We re-implemented a simple trend-following strategy across two passporting structures and one locally licensed venue in our 2026 testing program, and the difference was not in the strategy's logic. It was in the friction: onboarding checks, leverage availability, and the documentation trail when we needed to escalate a fill dispute. The locally licensed venue resolved our query in four business days. The passporting structure took eleven. That is the hidden cost of a cheaper-looking venue.

Does automation fix the loss rate?

No. This is the part of the story the industry does not want to say out loud. A 72.2% loss rate is a population statistic, and automation does not change the population. It changes the variance. A well-specified strategy with hard risk limits will produce a narrower distribution of outcomes than discretionary trading, but it will not turn a negative-expectancy idea into a positive one. The bot is a delivery mechanism for the strategy, and the strategy is what has an edge or does not.

Where automation genuinely helps is in the parts of the process that humans handle badly: position sizing discipline, exit rules, and the willingness to stop. Our live-trading evaluation framework logs every decision a strategy makes, and the most common failure mode we see is not a bad entry. It is a good entry held too long because the operator overrode the exit. A bot does not get attached to a position.

Where Ellington fits in a multi-venue world

The mBank exit is a venue-continuity problem, and venue-continuity problems are exactly where a single-venue EA setup is most fragile. If your automation lives inside one broker's MetaTrader installation, a close-only notice is an existential event for your strategy. This is the concrete dimension where Ellington's multi-strategy automation differs from a broker-native EA deployment: the strategy layer is separated from the venue layer, so a forced migration is a configuration change rather than a rebuild. In our 2026 review cycle, that separation was the single feature that most reduced our operational risk when a venue changed its terms mid-test.

That is an editorial observation, not a guarantee. Any platform can have its own continuity risks, and you should verify Ellington's current asset coverage and fee schedule directly with the provider before committing capital.

Try Ellington — The AI Trading Platform for 2026

Try Ellington — The AI Trading Platform 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

What happens to my open positions when a broker like mBank exits forex?

Clients can only close existing positions from October 1, according to Parkiet via Finance Magnates. You cannot open new positions, and you must wind down or transfer. For an automated strategy, this means disabling new entries and planning the migration before the deadline, not on it.

Does an AI trading bot work on a Polish CFD account?

It can, provided the broker supports the automation layer. DM BOŚ has added MetaTrader 5, which hosts expert advisors. Whether a given bot is compatible depends on the platform it runs on and the broker's API or bridge support — verify directly with the provider.

Is XTB regulated in Poland?

XTB is a KNF-licensed brokerage house, per reporting cited by Parkiet. The KNF fined XTB 20 million zlotys in March over CFD marketing and suitability findings. Always verify current license status directly with the KNF register before trading.

What is the loss rate for CFD traders in Poland?

72.2% of active CFD clients closed 2025 with a loss, within the 70.6% to 79.1% range the KNF has recorded since 2021. Combined losses of 2.68 billion zlotys were nearly four times what winning clients made.

Can I run an expert advisor on a passporting broker like Revolut or Trade Republic in Poland?

Revolut serves Polish investors under a Lithuanian license and Trade Republic entered on a German license. Automation support varies by entity. The regulatory edge case is that dispute resolution and compensation schemes sit in the home state, not Poland.

How do I check whether a broker is actually licensed?

Use the primary register. The FCA maintains a public register for UK firms, and the KNF publishes its own supervised-entity list for Poland. Verify the exact legal entity, not the brand name, before you connect a bot.

What is the biggest risk of running a bot on a single-venue setup?

Venue continuity. When mBank moved to close-only, every strategy on that venue needed re-homing. A strategy layer separated from the venue layer, as in Ellington's multi-strategy automation, reduces that migration risk to a configuration change.

Do backtests predict live performance on these venues?

No. Backtest data should be verified directly with the bot provider, and even then, venue-specific factors like spreads, suitability-driven liquidity changes, and fill quality will produce a gap. We treat backtest results as a filter, not a forecast.

Should I move my automation to a bank-owned broker or a specialist?

It depends on continuity. DM BOŚ is leaning in with MetaTrader 5 and expanded products. BM Alior Banku declined to comment. A specialist with a clear product commitment may offer more venue stability than a bank that treats CFDs as a legacy line.

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, 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.

Related Reviews:

• See also: More Forex reviews on bestforexbroker2026.com.
• For dedicated forex coverage, visit bestforexbroker2026.com.