On 29 October 2026, Saudi Arabia's Capital Market Authority (CMA) circulated to all capital market institutions the final regulatory provisions on restrictions on algorithmic trading, adopted by resolution of the CMA Board. The provisions follow a draft published for public consultation on 24 September 2026 and take effect on 1 November 2026, with institutions given until 31 December 2026 to bring their systems and practices into line. The consultation closed on 24 October; the final text arrived five days later, which suggests that the draft survived substantially intact.
Algorithmic trading is the use of computer programmes to generate, submit, amend and cancel orders according to pre-set rules, without a person keying each order. It ranges from execution algorithms that slice a large institutional order into smaller pieces across the day, to market-making algorithms that keep a two-way quote in the order book and move it as prices change, to high-frequency strategies that move in and out of positions in fractions of a second. It is now a substantial part of the Saudi market: the CMA Chairman put algorithmic trading at around a quarter of traded value in September 2024, and market commentary since then suggests the share has grown. The CMA's concern is not the technology itself but one of its by-products, namely order traffic. An algorithm can enter, amend and cancel orders thousands of times without ever trading, and that traffic can mislead other participants, burden the trading system and, in a stressed market, amplify disorder. The new provisions are aimed at the traffic, not at the trading.
Why this matters
1. The ratio measures noise, not activity
The cap is often described as a limit on algorithmic trading. It is not. A registered trader may execute any number of algorithmic trades in a security; what is limited is the number of orders generated to achieve them. The formula in the consultation text counts orders entered, adds amended orders at double weight and cancelled orders in full, and divides the total by executed trades plus one, before deducting one. The double weighting of amendments is the telling feature: an algorithm that constantly re-prices an order it has no intention of filling consumes two points of the ratio each time it does so. Cancellations are treated in the same spirit. Both are the raw material of layering and quote stuffing, where an order book is filled with orders designed to be seen rather than traded. An algorithm that trades a security twenty times and generates two hundred orders along the way is comfortably inside the limit; one that generates two hundred orders and trades once is not.
The adjustments in the formula, the addition of one to executed trades and the deduction of one from the result, soften the arithmetic at low trade counts, so that a trader who places a small number of orders in a security without executing is not automatically in breach. Beyond that margin, the test is the plain one the CMA has described: no more than twenty orders for every trade.
2. The registered trader, not the firm, is the unit of measurement
The ratio is calculated for each registered trader, meaning the individual registered with the Saudi Exchange under whose trading identity orders enter the system on behalf of a member and its clients. That is a more granular basis than the member-level ratios used by most European venues, and it has a practical consequence: a capital market institution cannot net heavy order traffic under one trader identity against light traffic under another. Each identity must stay within the cap in each security on each day. Institutions that carry client algorithmic flow, including direct market access and application programming interface connections, through a small number of trader identities will need to consider how that flow is allocated, monitored and, where necessary, throttled.
3. The liquidity classification ties the cap to the market-making framework
The securities covered are defined by reference to the Saudi Exchange's classification of securities by liquidity for market-making purposes, which is set out in the Exchange's Market Making Procedures. The four categories caught, high, medium, low and very low liquidity, are the ones in which the Exchange's market-making incentives apply; the very high liquidity category, broadly the most actively traded large-cap shares, sits outside the cap. The logic is defensible on both sides. The most liquid names have deep order books and can absorb heavy algorithmic traffic without disorder; the thinner names are where a burst of orders can move the price and where a misleading order book does the most damage. The same logic, however, means that the cap bites precisely where the Exchange has been working to attract market makers, a point returned to below.
4. Article 54: the control framework behind the number
The ratio is the headline, but the second limb of the circular may require more work. The CMA has reminded institutions that Article 54 of the Capital Market Institutions Regulations already requires them to establish and maintain systems and controls so that algorithmic trading is orderly and does not harm the market or breach the Capital Market Law and its Implementing Regulations. The circular spells out what that means in practice: a documented methodology for developing and testing systems, algorithms and strategies before they are deployed, with testing sufficient and continuous enough to confirm that they behave as intended across market conditions; policies and systems that monitor orders as they are entered, generate alerts and reports, and allow the institution to shut down a system or a single algorithm when necessary; retention of the data, policies, strategies, controls and testing documentation for ten years after an algorithm is withdrawn or amended; and provision of information to the CMA and Saudi Tadawul on request. For a firm that has treated its algorithms as a trading-desk matter rather than a compliance one, this is the part of the circular that will take the most time.
It is worth distinguishing this regime from the robo-advisory framework the CMA approved in March 2026. That framework governs algorithms that decide what a client's portfolio holds, under the Managing Investments activity. The new provisions govern algorithms that place orders in the market, whoever is behind them. A firm may be subject to both, but the obligations are different.
How does this compare internationally?
Order-to-trade limits and algorithmic control frameworks are an established feature of the developed markets, and the Saudi provisions follow the same architecture.
In the European Union, MiFID II requires every trading venue to set a maximum ratio of unexecuted orders to transactions for each member and each instrument, under a methodology set out in Commission Delegated Regulation (EU) 2017/566, and to be able to throttle or reject orders beyond it. A separate instrument, Commission Delegated Regulation (EU) 2017/589, imposes on investment firms the organisational requirements that Article 54 now spells out for Saudi institutions: pre-deployment testing, real-time monitoring, a kill functionality and record-keeping. In the United States there is no single statutory ratio, but the SEC's market access rule, Rule 15c3-5, requires broker-dealers to apply pre-trade risk controls to all order flow, including algorithmic flow, and the exchanges levy charges on excessive messaging. In India, the Securities and Exchange Board and the exchanges have operated an order-to-trade ratio framework for more than a decade, with charges that rise as a member's ratio increases and additional measures where the ratio spikes.
Three features distinguish the Saudi version. First, the ratio is set by the regulator at a single level for all institutions, rather than calibrated by the venue instrument by instrument, and most European venues set materially higher thresholds for registered liquidity providers than for other members. Second, measurement is per registered trader rather than per member. Third, the ten-year retention period is twice the five years required of European firms. None of these is unreasonable for a market at the Saudi Exchange's stage of development; together they make the regime tighter at the margin than its European model, and they are the points on which international firms entering the Kingdom will want clarity.
What it means for market makers and ETF investors
Market making is, by construction, an activity with a high ratio of orders to trades. A market maker keeps a two-way quote in the book and moves it every time the reference price moves, which means a stream of amendments for every execution. On the Saudi Exchange, registered market makers are obliged under the Market Making Procedures to maintain quotes within a set spread for a set proportion of the session, and the incentives for doing so are concentrated in the same four liquidity categories to which the cap applies. The circular makes no provision for market makers. Whether the CMA or the Exchange will accommodate registered market-making activity in practice, as European venues do through differentiated thresholds, is the single most important open question in the new regime, and one that market makers in mid- and low-liquidity shares will want answered before the compliance window closes.
For exchange-traded funds, the question is sharper still. An ETF's market maker quotes around an indicative net asset value that is republished throughout the session, in practice every 15 seconds, and commits in the fund's Terms and Conditions to keep its orders within a disclosed band of that value. Each refresh of the indicative value can require a quote amendment, and most of those amendments never trade; in a thinly traded ETF the ratio of amendments to executions can be very high indeed. ETF units are listed on the Main Market, so sponsors and their market makers should establish with the Exchange whether, and in which category, ETF units sit within the liquidity classification that defines the cap. If they are within it, the amendment weighting in the formula is the number to model. For the investor, the stakes are practical: the depth of ETF market making determines whether the spread on screen is tight and whether a unit can be sold at a price close to its net asset value. A control that narrows quote maintenance in the funds would be felt as wider spreads, which is the opposite of what the CMA intends.
What capital market institutions should be thinking about
The compliance window is short. Between now and 31 December 2026, an institution that runs its own algorithms, acts as a market maker or hosts client algorithmic flow should be looking at the following.
• An inventory of algorithmic order flow by registered trader identity and by security, with the ratio calculated under the CMA's formula on recent trading days to see where the cap would have been breached.
• Pre-trade controls that track the ratio intraday for each trader identity and each security, with throttling or rejection logic as it approaches twenty, and an escalation path to the trading desk and to compliance.
• A documented development and testing methodology for each algorithm and strategy, including testing across market conditions, a record of approvals before deployment and a schedule of continuing tests.
• Confirmation that monitoring, alerting and kill-switch arrangements work at both the system level and the individual algorithm level, and that a named person is responsible for using them.
• A record-retention policy that captures algorithm versions, strategies, parameters, policies and test results for ten years after withdrawal or amendment, and a process for responding to information requests from the CMA and Saudi Tadawul.
• For market makers and ETF sponsors, a specific analysis of quoting behaviour against the cap in each security and each fund, and early engagement with the Exchange on the treatment of registered market-making activity.
None of this is exotic. It is the same discipline that firms trading in London, Frankfurt or New York apply as a matter of course, and a compliance officer arriving from any of those markets will recognise the framework immediately. The difference for Saudi institutions is the timetable.
A measured step, with questions still to be answered
The provisions are a measured step. They do not restrict algorithmic trading; they restrict the order traffic that algorithmic trading can generate without trading, in the securities least able to absorb it, and they put a recognised international control framework behind that limit. For a market that now carries a substantial share of its value through algorithms, and that wishes to attract international quantitative and market-making firms, a rulebook those firms recognise is an advantage rather than a cost.
Three points remain to be settled, and all three sit within the regulator's discretion rather than in the text. The circular does not define an algorithmic order, so the classification of semi-automated tools such as smart order routers and basket executions is left to each institution's judgment for now. It does not say how breaches between 1 November and 31 December will be treated, beyond granting the window itself. And it makes no provision for registered market makers, whose business model depends on the very amendments the formula weights most heavily. How the CMA and the Exchange answer that last question will determine whether the regime tightens liquidity in mid- and low-liquidity shares and in exchange-traded funds or, as intended, cleans the order book without thinning it.
