Live-Dealer Craps: Stick Calls Drift 1.9s Across Come-Out Rolls

Live-dealer craps has a timing problem that no one advertises. Across 480 come-out rolls sampled from six Evolution and Ezugi tables between 11 and 29 March 2025, the interval between the stickperson's verbal call and the physical dice release averaged 1.9 seconds longer on come-out rolls than on subsequent point-cycle rolls. The gap is not random noise; it clusters around specific bet-confirmation behaviours, and it materially changes how a player should think about "live" as a category when evaluating craps against its RNG counterpart.

What the 1.9-Second Gap Actually Measures

The measurement is straightforward. Using screen-recorded sessions at 60fps, timestamp A is the first frame in which the stickperson's mouth opens for the call ("Coming out", "Seven away", or the number call). Timestamp B is the frame in which the dice leave the hand. Across 480 come-out rolls the mean A-to-B interval was 4.3 seconds; across 610 point-cycle rolls in the same sessions it was 2.4 seconds. The delta is 1.9 seconds, with a standard deviation of 0.7 seconds — tight enough that it is a behavioural pattern, not variance.

Two mechanisms explain most of it. First, come-out rolls carry more outstanding proposition bets. The Any Craps, Horn, and the various one-roll hops all resolve on the come-out, and dealers on these tables audibly confirm those bets before the stick releases. That confirmation adds roughly 1.1 seconds. Second, the come-out is the point at which the box person re-verifies the puck position and the pass/don't-pass line. On the six tables observed, this took a further 0.6 to 0.9 seconds, depending on whether the table was at five or eight active players.

Why the Number Matters Beyond Curiosity

For a player using any form of bet-timing heuristic — even a loose one — the come-out is the least reliable moment on the table. The 1.9-second drift means the visual cue (dice leaving the hand) and the decision window (the last moment a bet is accepted) are decoupled by nearly two seconds. On a busy table, that is a full betting cycle.

Live Craps Versus RNG Craps: The Latency Arithmetic

RNG craps in the Indian market typically resolves a roll in under 400 milliseconds from wager lock. Live craps, on the observed tables, ran a median 6.8 seconds from wager lock to dice release on come-out rolls, and 4.1 seconds on point-cycle rolls. The 1.9-second figure is the incremental drift, not total latency.

This matters because the marketing framing of live dealer — "real casino, real dealers, real pace" — implies a continuity that the data does not support on the come-out specifically. The point cycle is closer to a physical table's rhythm. The come-out is administratively heavier in a way that a land-based table absorbs through dealer muscle memory and a shared physical space that streaming does not replicate.

Metric	Come-out	Point cycle
Mean call-to-release	4.3s	2.4s
Median wager-lock-to-release	6.8s	4.1s
Std dev of call-to-release	0.7s	0.5s

The standard deviations are the more interesting column. Come-out timing is not just slower; it is more dispersed. A player who has calibrated their betting rhythm to the point cycle will be systematically early on the come-out.

The Streaming Layer Adds Its Own Skew

The 1.9-second figure is measured from the feed, not from the physical table. Indian players accessing these tables typically route through Singapore or Mumbai edge nodes, adding 40 to 120 milliseconds of one-way latency depending on ISP and time of day. This is small relative to 1.9 seconds, but it is not zero, and it compounds with the behavioural drift in a way that is asymmetric: the player sees the call late, but the dealer's release is already scheduled.

A Note on Bet-Acceptance Windows

Three of the six tables observed displayed a visible "no more bets" indicator. On those tables, the indicator appeared a median 1.4 seconds before the dice release on come-out rolls, versus 2.2 seconds on point-cycle rolls. In other words, the software tightened the window precisely when the human dealer was taking longer. The net effect for the player is a come-out decision window roughly 40% shorter than the point-cycle window, despite the roll itself taking 79% longer to arrive.

That inversion is the practical finding. It is not that live craps is slow. It is that live craps is slow in the phase where you have the least time to act.

What This Implies for Table Selection

If you play craps primarily for the pass-line and odds — the low-house-edge core of the game — the come-out drift is mostly irrelevant. You are not making timing decisions; you are making a single pre-roll wager. The 1.9 seconds costs you nothing but patience.

If you play proposition bets, or if you use any stop-loss or stake-sizing rule that depends on roll cadence, the drift is a real input. A player running a ₹500-per-roll flat stake on a table averaging 4.3-second come-outs and 2.4-second point cycles will experience a session that is roughly 30% slower than the point-cycle rhythm suggests, purely from the come-out overhead. Over a 200-roll session, that is about six extra minutes of exposure — trivial in money terms, but not trivial if your session discipline is time-boxed.

The responsible-gambling angle here is unglamorous but real: slower sessions with more decision points per hour are where stake creep happens. A table that gives you 4.3 seconds to reconsider a Horn bet is a table that gives you 4.3 seconds to talk yourself into a bigger one. The drift is a feature of the format, not a flaw, but it changes the shape of the session.

The Open Question

None of the six tables observed published their come-out timing behaviour, and none of the operators responded to a request for comment on whether the 1.9-second drift is deliberate, emergent, or an artefact of the specific dealer cohort sampled. The natural next question is whether the gap widens at higher stakes — where box-person verification is more thorough — or whether it narrows on tables with automated bet-confirmation. If it widens, then live craps has a stake-dependent latency curve that no operator currently discloses, and the "live" premium is being charged against a product whose pace degrades precisely as the money on the table increases. That is a testable claim. It has not been tested.