A player sits down at a Free Bet Blackjack table and gets dealt a pair of 7s against the dealer's 6. He splits. The dealer doesn't take any chips — the casino is paying for the second hand. He gets a 4 on the first hand (hard 11) and the dealer marks the double automatically; no chips moved. The pit boss watches him play through what should have been one $25 hand with $25 of his money still on the felt and three $25 'free' hands stacked beside it: one split-of-7s, one double-on-7-of-spades-into-hard-11, the other half of the split sitting at hard 12. Three free chances at a winning hand, dealer showing a 6 — looks like a giveaway. Then the dealer turns over a 5, draws an Ace and a 10, lands on 22, and pushes every single one of those hands instead of busting. The 'free' bets returned the original wagers; the casino kept the player's $25. He walks away wondering what just happened.
Free Bet Blackjack is the highest-marketing-to-math-ratio table game on the modern casino floor. The casino-pays-your-splits-and-doubles framing is real — Geoff Hall's variant genuinely refunds those bets — but the dealer-pushes-on-22 rule that funds the giveaway is the most expensive single rule shift in mainstream blackjack. This post is the honest EV breakdown: where the free splits and doubles actually pay back, where the push-22 rule eats them, what the published house edge runs at, and the strategy deltas that separate a tourist losing 1.5% to the house from a thinking player losing 0.6%.
The three rule changes that define the game
Free Bet Blackjack starts from a standard 6-deck H17 base game with late surrender allowed. Three changes are what make it Free Bet:
- The casino pays for every player split on pairs of 2s through 9s. (Pairs of 10s and Aces still cost the player.) The split bet is matched by a 'Free Bet' lammer; if the resulting hand wins, the player gets paid as if the second bet were real. If it loses, the player loses nothing extra. If it pushes, the lammer is removed and no money changes hands.
- The casino pays for every player double on hard 9, 10, or 11. Same lammer mechanism. Doubles on soft hands and on hard totals outside 9-11 are not 'free' — the player puts up real chips.
- The dealer pushes on 22 against any non-busted player hand. If the dealer makes 22, every player who hasn't busted gets the original bet back. Player blackjacks (which are paid before the dealer plays out) are the exception — those still win 3:2.
The first two rules are the marketing. The third is the math. The dealer's hand draws to 22 in roughly 7% of rounds — about one in fourteen hands — and that 7% would otherwise be dealer-bust wins for the player. Converting those wins to pushes is what funds the free splits and doubles.
What the free bets are actually worth to the player
A free split on pair-of-7s vs dealer 6 in classic 6D H17 is worth approximately +0.18 units to the player (the split-7s hand has positive EV against the dealer's bust card, and the second hand is real money). In Free Bet, the casino is putting up the second bet for you — the EV of taking the split becomes the EV of the two resulting hands evaluated independently, minus zero risk on the second hand. A back-of-the-envelope: the freed bet roughly doubles the split's EV contribution, from +0.18 units to +0.36 units.
Across all the split-eligible pairs (2-2 through 9-9), summed across the dealer up-cards where the split is correct in Free Bet strategy (which is much wider than classic — split on almost every dealer up-card for free-pair-eligible pairs), the free-split feature adds roughly +0.85% of EV back to the player.
Free doubles on hard 9, 10, and 11 add another ~0.55%. Total: roughly +1.40% of EV returned to the player from the free-bet features.
What the push-22 rule actually costs
The dealer ends with a final total of 22 on about 7.4% of rounds in a 6D H17 game (this includes all dealer draws that finalize at exactly 22, not just busts to that number). In classic blackjack, every one of those rounds is a dealer bust — and every non-busted player at the table wins. In Free Bet, every one of those rounds is a push for every non-busted player who didn't make blackjack.
The EV cost of converting wins to pushes on 7.4% of rounds is approximately -2.10% to -2.30% depending on how the player's bet mix shakes out (bigger bets at high counts make this expensive faster, but Free Bet isn't typically counted). The mainstream published figure is -2.05%.
Net: +1.40% (free bets) minus 2.05% (push 22) equals roughly -0.65% house edge against an optimal Free Bet strategy player. Published numbers from Wizard of Odds and other math sources cluster between 0.60% and 1.00% depending on the exact rules variant — late surrender allowed adds maybe 0.07%, S17 (dealer stands on soft 17) shifts the math slightly, and a few casinos run a 'no late surrender' Free Bet that drifts higher.
The strategy deltas that actually matter
Five differences from classic 6D H17 basic strategy account for most of the EV loss when players walk in cold:
- Hard 12 vs dealer 4: hit, not stand. Classic strategy stands because the dealer is likely to bust to a low card; in Free Bet the dealer's 22-pushes erode too much of that bust value to justify standing on a stiff total.
- Hard 13 vs dealer 2: hit, not stand. Same logic — push-22 reduces the value of dealer-busts on the bust-prone up-cards.
- Soft 17 (A,6) vs dealer 3-6: hit, not double. The push-22 rule reduces the value of doubling soft 17 because too many dealer hands that 'should' bust now push instead. The free-double rule only applies to hard 9/10/11, so the soft-17 double is real money — and the math says don't bother.
- Soft 18 (A,7) vs dealer 2: stand, not double. Classic H17 doubles vs 2 with soft 18; Free Bet flattens this to a stand.
- Pair of 2s, 3s, 6s, 7s vs almost any dealer up-card: SPLIT. Because the split is free, the threshold for splitting drops to 'is the EV of the resulting hand greater than the EV of the unsplit pair?' for the original bet, with the second bet contributing free upside. The published Free Bet chart splits pair-7s vs 2-7 (classic splits 2-7 too, but the EV math is much stronger in Free Bet) and splits pair-6s on 2-6 (wider than classic's 2-6 DAS).
The pair-of-5s and pair-of-10s rules are unchanged: never split pair-5s (it's a hard 10, double it instead), never split pair-10s (a 20 standing is way better than two unknown hands). Pair-of-Aces is unchanged: always split, but the split is not free because Aces are excluded from the free-pair feature.
Where Free Bet looks like fun but isn't
Two failure modes are common at Free Bet tables:
First, the swing illusion. Splitting and doubling for free feels like the casino is giving away money — a player who splits a pair of 8s into two hard 18s, one of which wins, feels like the win was a gift. The math says the win was paid for in dealer-22-pushes that the same player took on five rounds where his hand should have won outright. The visible giveaway is funded by the invisible tax.
Second, the 'might as well take it' trap on bad splits. Free Bet's free-pair feature does NOT mean every pair should be split. Pair of 4s vs dealer 2-3 is still a hit (splitting into two weak hands is worse than playing the hard 8 even when the second hand is free, because the resulting hands are systematically underwater). Pair of 9s vs dealer 7 is still a stand (a hard 18 wins more often than two new hands starting from 9 vs dealer's expected 17). The chart at lib/strategy/free-bet.ts encodes these exceptions; eyeballing 'it's free, I'll split' is the way most cold players give back the variant's already-thin edge.
What about counting Free Bet?
Hi-Lo counting works on Free Bet — the rules don't break any of the count assumptions — but the push-22 rule changes the high-count payoff. A 10-rich shoe makes dealer hands MORE likely to draw to 22, which makes pushes more frequent at high count and reduces the ramp benefit. Published estimates put the edge gain at high TC at roughly half what it would be at a classic 6D H17 table for the same spread. Net: counting Free Bet is possible but inefficient compared to counting classic; advantage players who run Free Bet usually do it as cover or as a comp-generation move, not as a primary EV game.
When Free Bet is the right table choice
Free Bet is a reasonable table for a recreational player who: (a) actually learns the chart at lib/strategy/free-bet.ts, (b) enjoys the dramatic-feeling free-split rounds (the variance is higher than classic because the split-and-double feature increases the size of typical wins and losses), and (c) is sitting at a property where the alternative is a 6:5 table or a continuous-shuffle table that runs at higher EV cost. The 0.6-1.0% edge is acceptable for casual play — about half a step worse than a clean 6D H17 3:2 game (which runs ~0.50%) and dramatically better than 6:5 blackjack (~1.85%).
It's the wrong table for someone trying to grind serious hours of bankroll-EV play. The push-22 variance plus the inefficient counting makes it a recreational-only game in practice. Walk past it if there's a 3:2 H17 DAS game open on the same floor; sit down if it's that or the 6:5 trap next to it.