How Dealer Performance Analytics Works in a Live Casino in 2026

On July 27, 2026, Regulation (EU) 2026/1744 of the European Union’s Digital Omnibus on AI entered into force and introduced several obligations, including the requirement for artificial intelligence (AI) systems used to monitor staff performance (such as software studios use to grade croupiers/live dealers) to be under human oversight by December 2, 2027. This has resulted in dealer performance analytics having two sides. The first is what a studio thinks about its staff, and the second is what a platform says about live tables. Online operators generally get the most value from analyzing tables and shifts from their own round data and, more often than not, leave individual scoring to the studios employing the dealers.

Every live round (i.e. game cycle) has a unique round ID. This ID is shared by the studio and the platform and allows both parties to analyze tables. On the NuxGame casino platform, game performance and revenue reports are in the same back office as player and payment data. This document reviews metrics, data flow, regional restrictions, AI, and review processes.

Key Takeaways

1. 01

   Shift codes let brands compare tables without dealers’ personal data.

2. 02

   Pace, voids, and occupancy are fairer signals than table revenue.

3. 03

   Studios own dealer records, while the platform owns every stake and void.

4. 04

   The EU bans emotion scoring and treats staff scoring as high-risk from 2027.

5. 05

   Put the round fields you need into the provider contract before launch.

What Dealer Performance Analytics Measures and Who Retains the Data

In a conventional casino setting, a pit manager keeps an eye on each dealer’s speed and errors and can see every player at the table. In iGaming, that single view is split between two companies.

The studio employs dealers and keeps the videos. The brand’s platform records every single stake through game integration. This is how the studio’s casino game becomes part of the brand’s lobby.

• The studio keeps records: employee rosters and shift schedules, procedure errors, and videos. Only contract summaries are shared with the brand.
• The platform keeps records: every single stake, win, and void (a round of play that is cancelled, with stakes returned to players), all of which are tagged with a round ID, a table ID and a time stamp.
• Shared tables: Evolution, a live game supplier, reported they served around 870 customers and had roughly 2,000 tables in operation at the end of 2025. So one table serves multiple brands.

Dealer Performance Metrics and KPIs at Table Level

At table level, many KPIs are recorded by the platform and taken as is from round records.

Metrics	Plain meaning	Owner	A change usually signals
Rounds per hour	Game cycles a table completes each hour	Platform	Slower dealing or a new countdown length
Void rate	Share of rounds cancelled due to an error, e.g. a misdealt blackjack hand	Platform counts it; studio explains it	Procedure problems on one shift
Seat occupancy	Brand's players at the table each hour	Platform	Demand by hour, language, or lobby position
Table GGR and hold	GGR (gross gaming revenue: stakes – winnings) and hold (share of stakes the house keeps)	Platform	Revenue level, noisy over short periods
Return rate and complaints	Players back within 7 days, plus ticketing rate per 1,000 rounds (a proxy for customer satisfaction)	Platform and support desk	Player response to the table

A public benchmark doesn’t exist because each provider sets its own rules and limits. That’s why a brand compares each table to its own history hour by hour.

Pace is the one thing a dealer controls most directly during live play. In Great Britain, the UK Gambling Commission caps online slots at one game cycle every 2.5 seconds in requirement RTS 14D of its remote technical standards. However, roulette tables are at the croupier’s discretion, so pace and complaints often go hand in hand.

How Round Data Reaches the Dashboard

Business Intelligence (BI) is all about turning data into information. A dashboard is the screen that displays it. Live table data analytics works the same way as a retailer who looks at sales data by store and by shift. The real work is in joining records that already exist.

1. Studio event: The game control unit (GCU), which runs the table, opens and closes a countdown and time-stamps the result.
2. Wallet calls: The provider asks the platform to charge players their stakes and pay out their winnings. The round ID is quoted, so the brand can get its data on every round.
3. Stake record: The platform stores stakes, results, players and round IDs as the brand’s data.
4. Shift join: A studio shift file maps tables and times to a shift code (a label for one dealing shift), and then performance across shifts can be compared.
5. Data quality checks: Duplicates, late results, and voids are matched up against the provider’s round report.
6. Alerts: A real-time feed drives alerts on pace and voids, while monthly reviews run on slower, less expensive batch jobs at night.

The most interesting information comes from comparing each table against the others across shifts, since the game and limits remain the same, and it’s the dealers that change. So a shift with a higher void rate is a question for the studio. Live studios are connected via the NuxGame game aggregation layer, which gives a brand a single connection to numerous game providers, including Evolution, Ezugi, Pragmatic Play and LuckyStreak. Reports are refreshed every 5 minutes, depending on how quickly the data arrives. The back office then verifies the round records against those of the providers. There are also scheduled exports, which send player data to a brand’s Business Intelligence tools.

Performance Tracking Rules – Different Regulatory Regimes

There are various gambling regulation frameworks, e.g. those in the EU, Brazil and Great Britain, that govern which dealer records studios need to retain, and separate rules cover how staff can be monitored. The following table explains all of this.

Rule	Requirement in plain words	Effect on dealer data	Source
UKGC RTS 17, requirement 17A	Fair, auditable live dealing, including training records, video and game stats by staff	British-facing studios already hold dealer stats	UKGC RTS, February 2021
SPA/MF Ordinance No. 1,207 (Brazil)	Studio certification by a lab accredited by the Secretariat of Prizes and Bets (SPA), renewed annually	Data stays available to the SPA for at least 5 years	Articles 5, 8, 10, July 29, 2024
EU AI Act, Article 5(1)(f)	Inferring emotions at work is prohibited, except for medical and/or safety reasons	Mood scoring of croupiers is not permitted in EU studios	Regulation (EU) 2024/1689, from February 2, 2025
EU AI Act, Annex III (high-risk list), point 4	Tools for assessing the work of employees must be logged and supervised by natural persons	Dealer scoring software requires these controls	Regulation (EU) 2026/1744, from December 2, 2027

Compliance with the Regulatory Technical Standards (RTS) is a significant element in obtaining a license in Great Britain. The guidance around RTS 17 is that studios need to collect all these game events to build statistics which show trends for staff members. In this way, a brand that offers those games can ask for table-level stats in its content contract.

Both EU rules apply to the studio and to the vendor supplying the software. A brand that starts rating its dealers using its own tool may find it is also taking on some of the responsibilities of the person who deployed that tool in the first place – which is another good reason to stick to table level.

What’s Changing in Latin America

In Brazil, the SPA regime applies to every licensed betting operator, and Evolution opened up a studio there in 2025. In the rest of the region, tables with Spanish tend to be busy at different hours than English ones.

That can give you a bit more competitive edge with regards to timing. To find out if it does, take a look at your round data and see whether table occupancy is higher.

Where Do We See AI with Dealer Analytics?

AI has a few different roles to play. Before anything happens, make sure your data is properly joined up.

You Can’t Do AI Without Good Data

Gartner has predicted that by 2026, about 60% of all AI projects will fail due to a lack of good data. Additionally, in July 2024, one of their surveys revealed that 63% of all organizations do not have the right data management practices, or are not sure if they have them. A model trained on a suboptimal data join will learn the errors in the join, so that is the first thing that needs to be checked.

Anomaly Detection: A Low-Risk AI Use Case

Anomaly detection software identifies atypical behavior in a series of events. Because it does not name any individual person, it poses a lower risk to a brand. One use case for this technology in live studios is to flag a dealer who has a quiet shift.

Computer Vision on the Studio Floor

A subcategory of AI that interprets images and videos is called computer vision. This software is already being implemented in several studios to read out card and wheel results. However, if it is instructed to check individual shuffles, that falls into a much higher risk category under the EU AI Act. Every flag would then have to be reviewed by a human operator under the standard AI governance processes.

Scoring Croupiers’ Friendliness

You won’t be able to score how friendly a croupier looks using an AI tool because, according to the EU rules, it would mean inferring emotions from a person’s facial expression. Simple metrics like the return rate and complaints per table, however, would be okay to use.

Best Practices for Early Warning and Performance Reviews

Managing dealers’ performance is typically studios’ job, since they are the ones who have access to croupiers’ performance data. The brand just routes each table signal back to its owner.

Signal	Who acts	First action
Pace down on one shift	Studio account manager	Procedure log review; the studio decides on additional training
Void rate spike	Studio	Immediate intervention: video review, with the table paused if necessary
Return rate falling	CRM (customer relationship management) team	Check whether those players stay active elsewhere

Monthly performance reviews determine which tables keep their lobby placement. Also, to make the right decisions on peak-time positioning, occupancy by hour over a 3-month period is valuable. In the background, shift comparisons can show where the studio might need to do some follow-up work, such as unexplained voids that happen only at night.

A Table That Fills Up During Launch Campaigns

A table filling up during a launch campaign only works as a growth engine if occupancy holds after the welcome offer. Of course, you then need to consider the profitability of that table after the provider takes their revenue share.

Pro tip: When it comes to live dealer analytics, be sure to draft the round fields you’ll need into the game provider contract before going live. That way, you should not be put in the position to have to recreate a missing round field using wallet records.

Some of the Problems Affecting Live Dealer Analytics

• Judging dealers by table revenue: This is probably one of the worst things to do, because a couple of high rollers can make the results very misleading. Pace and void rate are fairer first signs.
• Reading a seasonal spike as a dealer effect: Just because a table is filling up during a large event like a football tournament doesn’t mean the dealer is causing it. Comparing the same hour will tell you if the table is really getting more traffic.
• Treating the provider’s dashboard as complete: The studio is reporting the entire table, but that doesn’t mean the brand is going to see those numbers. The two only match up after reconciliation.

Tech Snapshot

Conclusion: Where Dealer Performance Analytics Really Pays

We hope you’ve found that useful. For a brand looking to choose a studio in 2026, the main thing that’ll decide for them is whether the round feed has a shift code. If it does, then dealer performance analytics will be running off their own data from day one. If it doesn’t, the first thing will likely be a contract amendment.

According to the 2026 UK Gambling Commission data (published 17th September 2026), the online casino sector in Great Britain generated £5.7 billion in gross gambling yield between April 2025 and March 2026, and £4.8 billion of that came from slots. That means the remaining £900 million must have been generated by all other gaming formats, including table games. That’s why table-level evidence matters when deciding which studio tables get a place in that smaller cash pot.

While their live lobbies are in the works, brands can access round reports and data exports through the NuxGame platform demo before signing a studio contract. They can also bring a sample round report from a studio of their choice to see what fields are available.