Live dealer latency: 300ms on camera, 4s on the chat ban

Live dealer latency is usually discussed as a video problem: the stream lags 300–500ms behind the physical table, and everyone accepts it. But there's a second latency nobody prices in, and it runs on a different clock. A player can be looking at a hand dealt 300ms ago and, in the same session, be sitting in a chat timeout that won't lift for another four seconds—or forty minutes, depending on the room. The camera and the ban are two separate systems, and they rarely agree on what "now" means.

That mismatch matters because live dealer is the one vertical where the player's sense of time is anchored to something physical. You can see the wheel. You can see the dealer's hand move. When the software's response to your behaviour doesn't map onto that visible reality, the trust gap opens fast.

The two clocks in a live dealer session

The video clock

Video latency in live dealer sits in a fairly narrow band. Most studios target 300–600ms glass-to-glass, and the better ones push toward the low end using WebRTC rather than the older HLS stack, which can add 2–6 seconds on its own. That's a solved engineering problem in 2024 terms. Players tolerate it because a 400ms delay on a spinning roulette wheel is invisible unless you're counting.

The video clock is also roughly constant. It doesn't spike because you typed something the dealer didn't like.

The moderation clock

Chat moderation runs on an entirely different architecture. A message hits a filter—keyword, pattern, sometimes a language model—and either passes, gets held for human review, or triggers an automatic mute. The mute itself is applied by a back-office tool that may be the studio's, the operator's, or a third-party moderation vendor's. These systems are not synchronised with the video pipeline, and they were never designed to be.

The result is the thing the title describes. A player sees a dealer react in real time, types a response, and then waits. If the mute lands, the confirmation can take seconds to surface in the UI—sometimes the chat simply stops accepting input with no explanation at all. Four seconds is not an exaggeration; it's roughly the round-trip you get when a moderation call has to cross from the game client to an operator's API, get evaluated, and come back.

Why the numbers don't line up

The asymmetry comes from where each system lives.

Video is pushed from the studio to a CDN edge close to the player. It's optimised for throughput and jitter, and the whole chain is measured because buffering is a commercial emergency.

Moderation is pulled. The client sends a message to a backend, the backend consults rules, and the response travels back. Nobody measures this end-to-end because a slow mute isn't a revenue event—it's a compliance one. The metric that gets tracked is "messages actioned," not "milliseconds to action."

There's also a jurisdiction problem. A single live table can serve players under three or four different regulators, each with its own chat rules and reporting requirements. The moderation layer often has to check against all of them, which means the slowest applicable rule sets the response time. A player in one market gets a near-instant mute; a player in another waits while the system resolves which rulebook applies.

The dealer is not the moderator

Worth stating plainly, because players frequently don't know: the person on camera almost never controls the chat. Dealers are trained to keep the table moving and may flag abuse verbally, but the mute button sits with a remote team. So when a player sees the dealer smile and then finds themselves silenced, the two events are unrelated—and the four-second gap makes it feel like a betrayal rather than an automated action.

What the latency actually costs

The concrete cost is dispute volume. Operators I've spoken with describe chat-mute complaints as a small but stubborn share of live casino support tickets—often cited around 3–5% of live-specific contacts, though few publish the figure. The complaints aren't usually about being muted. They're about not understanding when or why it happened, because the UI feedback arrived late or not at all.

There's a second cost that's harder to measure: the erosion of the "live" promise. Live dealer's entire pitch is presence—a real person, a real table, real time. Every desynchronised interaction chips at that. A player who gets a mute confirmation four seconds after the offending message has already scrolled past learns something about the product: the human on screen is not actually connected to the system judging them.

The 300ms figure is a red herring

Studios love to quote video latency because it's a number they control and it looks good. But the player's experienced latency is the maximum of every clock in the session, not the minimum. If the video is 300ms and the moderation response is 4 seconds, the session's felt latency is 4 seconds. Marketing the first number while ignoring the second is a bit like advertising a car's top speed and leaving out the braking distance.

Where the fix is likely to come from

The obvious technical answer is to move moderation decisions client-side where possible—instant local mutes for clear-cut violations, with the backend reconciliation happening asynchronously. Some operators already do this for profanity filters. It's cheap and it collapses the visible delay to near zero.

The harder problem is transparency. A mute that says "you're timed out for 10 minutes because of X" in the same frame the message was sent reads as fair. The same mute delivered four seconds later, with no reason, reads as arbitrary. The latency isn't really the issue—the silence during the latency is.

Regulators are starting to notice. Several European licensing regimes now require operators to inform players of chat restrictions and their duration, which forces the moderation layer to be at least as responsive as the video. That's a compliance requirement doing what UX advocacy couldn't.

The open question is whether live dealer studios will own this problem or leave it to operators. The video pipeline is the studio's; the moderation stack is usually the operator's; the player experiences them as one product. Until someone owns the seam between them, the camera will keep running 300ms ahead of the system that decides whether you're allowed to speak—and players will keep noticing.