A mid-size UK operator's own table data shows that live dealer tables between 22:00 and 23:00 pull roughly three times as many £2 side bets per active player as their RNG equivalents over the same window — 3.1x across blackjack and roulette, measured over the eight weeks to 30 September. The gap is not a rounding artefact: it holds after weighting for stake size, and it narrows to about 1.4x by 03:00.

The £2 side bet is an odd thing to build a theory around. It is small enough that most players barely check the return, and it sits on the edge of the main wager rather than in place of it. Perfect Pairs, 21+3, Hot 3, the various roulette neighbour and split bets — these are the products that fund a lot of live dealer margin while attracting very little scrutiny. That they behave so differently between live and RNG at a specific hour of the evening is worth taking apart.

What the 3.1x figure actually measures

The metric is side bets per active player per hour, not total handle. That distinction matters, because live dealer tables at 22:00 also have higher average main stakes — around £14.80 against £9.20 on RNG blackjack in the same operator's data. If you measured raw side bet handle, the multiple would look larger and mean less, because it would mostly be telling you that people who bet more also bet more on the side.

Normalising per player strips most of that out. What remains is a behavioural difference: the same kind of player, in the same hour, is more likely to put a small chip on Perfect Pairs when there is a person dealing than when a shuffling machine is.

The control that makes it credible

The operator ran the comparison against RNG tables with identical side bet paytables and identical minimums, and excluded accounts that had played both formats in the sample week. The 3.1x survived that. It also survived a check on session length — live sessions in the 22:00 hour averaged 41 minutes against 27 on RNG, so per-hour normalisation is doing real work rather than hiding a duration effect.

Why the evening peak behaves differently

Three explanations are worth separating, because they imply different things for how you build the lobby.

The social pace. A live blackjack table deals maybe 60 to 70 hands an hour; an RNG table set to fast mode can run 250. Side bets are impulse products. They get placed in the gap between the decision and the deal, and a live table has a much bigger gap. The RNG player who is three hands into the next round has already moved on.

The visible chip stack. On a live table you watch other players' side bet wins resolve in real time, with the dealer paying them. On RNG, the side bet result is a line of text. The 22:00 hour is when tables are busiest — the operator's live blackjack seats ran at 78% occupancy in that window against 34% at 04:00 — so the social proof effect is at its strongest exactly when the data says side betting peaks.

The £2 denomination itself. Live dealer interfaces make the £2 chip a single tap on a dedicated spot. Several RNG clients bury the side bet behind a secondary panel. That is a design difference, not a psychological one, and it is the easiest of the three to test.

The narrowing to 1.4x by 03:00 is consistent with all three. Late-night tables are quieter, so there is less to watch; players still awake are more likely to be grinding a system than socialising; and the £2 chip is less of a novelty at 3am.

What this means for the £2 bet's economics

Side bets are usually priced generously to the house. Perfect Pairs at typical paytables carries a house edge north of 5%, and 21+3 sits around 3.7% depending on the paytable variant. A £2 bet at those edges is worth 7p to 10p in expected margin per placement. Multiply by three and you have a product that quietly outperforms its headline RTP suggests.

That has a knock-on for how operators should treat the 22:00 window. If a live table is generating three times the side bet volume, the marginal value of a seat in that hour is much higher than seat occupancy alone implies. It also argues against the common practice of pushing side bet promotions into off-peak hours, when the data suggests the appetite is not there.

The responsible gambling angle

There is a version of this finding that should worry compliance teams. Side bets at 3.7% to 7% house edge are the most expensive thing on the table, and they are being placed most heavily at the point in the evening when players are most relaxed and least likely to be checking paytables. The £2 stake is small, but the multiplier is not — a player placing four side bets an hour on live blackjack at 22:00 is losing money at roughly four times the rate of the same player on RNG.

None of that argues for restricting the product. It does argue for the paytable being visible on the table itself rather than in a rules tab, and for session reminders that account for live table pace rather than clock time. A 41-minute live session at 22:00 involves fewer decisions than a 27-minute RNG session, but each decision carries a higher expected cost if the side bet is in play.

The open question: is this a live effect or a 10pm effect?

The data cannot yet separate the two. The 3.1x could be a property of live dealer tables generally, surfacing at 22:00 because that is when the tables are full. Or it could be a property of the hour — the post-dinner, pre-late-night window when players are more inclined to take a small punt on the side.

The test is straightforward: run the same measurement on live tables at 14:00 on a Saturday, when occupancy is comparable but the audience is different. If the multiple holds, it is a live dealer effect and the RNG side bet interface is the thing to fix. If it collapses, the £2 side bet is an evening product, and the operators pushing it hardest at midday are reading their own data backwards.

Until someone runs that comparison, the 3.1x stands as a number without a settled cause — which, for a product this small and this profitable, is an uncomfortable place to leave it.