Thirty answers,
math-first.

Around 0.41% on a standard 6-deck S17 DAS late-surrender 3:2 table — among the lowest in any casino. Without basic strategy, the same game runs at ~2.5%. The chart cuts the edge by roughly 5x.

A chart of the mathematically optimal play for every (player hand, dealer up card) combination. Computed from billions of simulated hands, no judgment calls. Memorising it is the single highest-EV move in any casino game. See /games/blackjack/strategy for the full chart.

No. Insurance is a separate side bet with a 7.4% house edge at random card counts — worse than every cell on the basic strategy chart. The math: the chance of a 10 in the hole is ~32.7%, but insurance pays 2:1 (requires 33.3% to break even). Skip it unconditionally as a non-counter.

A natural blackjack pays $1.50 per $1 at 3:2 but only $1.20 at 6:5. That 30-cent difference, multiplied by the 4.83% rate of naturals, adds 1.39% to the house edge — turning a 0.41% game into a 1.80% game. Always sit at 3:2 tables.

Depends on the table. S17 (stand) is the canonical reference rule in most regulated live-dealer blackjack online; H17 (hit) is the Vegas Strip standard and some Evolution VIP rooms. H17 costs the player ~0.20% extra house edge.

Yes — both, regardless of the dealer up card. Splitting aces converts a weak 12 into two hands each starting with 11. Splitting 8s converts a 16 (the worst hand in the game) into two playable starts at 8. Never split 5s (you give up a strong 10) or 10s (you give up a near-certain 20).

Yes, 100% legal everywhere. Casinos are private property and can refuse service to advantage players, but mental card-counting cannot get you arrested or prosecuted. Counting only crosses into illegality when assisted by a device (concealed computer, marked cards, etc).

No. In-house RNG variants reshuffle the virtual deck every hand (no shoe state to track). Live-dealer studios use continuous-shuffle machines or shoe-shuffle machines that re-randomise far before the deep penetration counters need. The math is sound; the application is dead online.

Standard 6-deck S17 DAS late-surrender 3:2 from a major live-dealer studio (Evolution, Pragmatic Live) at a top-trust-score operator. Verify the table's rules and payout before sitting. See /games/blackjack/where-to-play for our operator picks.

Most live-dealer tables use 6-deck or 8-deck shoes. RNG variants vary: Stake Originals Blackjack is single-deck reshuffled every hand; many other in-house RNG variants are 6-deck. Single-deck has the lowest house edge, but is almost always paired with 6:5 payouts which more than wipe out the deck-count advantage.

Soft hand: contains an ace counted as 11 (so the ace can shift to 1 if needed). Cannot bust on the next card. Hard hand: no usable ace, or the ace must count as 1. The chart treats soft and hard hands separately because the ace flexibility changes the optimal play.

Always double on hard 11 (vs anything but ace), hard 10 (vs 2-9), hard 9 (vs 3-6), and on several soft hands against weak dealer cards. Doubling is the highest-EV action in the chart when conditions are right; recreational players miss 30-60% of their +EV doubles out of risk aversion.

Late surrender (after dealer checks for blackjack) on hard 16 vs 9 / 10 / A and hard 15 vs 10 — the chart's worst hands. Better to lose half the bet than play out. Some H17 charts add surrender for hard 15 / 17 vs A and pair 8s vs A.

Double after split. A rule allowing the player to double down on a hand created by splitting. Worth ~0.14% in player EV. Standard at most 3:2 tables. Without DAS, some pair splits stop being +EV and revert to hit.

When the player has blackjack and the dealer shows an ace, the dealer offers 1:1 immediately rather than waiting to see if she also has blackjack (in which case you push). Mathematically identical to insurance — same house edge — and a sucker bet at any non-counted true count. Skip.

Roughly 4-6 hours of trainer practice spread over two weeks gets most players to 95%+ accuracy. Use /games/blackjack/trainer — random hand, random dealer card, instant feedback. Decision-zone hands (hard 12-16, soft 17-18, pairs) get the most reps.

Licensed RNG and live-dealer blackjack are audited by independent test labs (eCOGRA, GLI) and large operators publish provably-fair seeds for individual rounds. Sustained statistical deviation from theoretical RTP is detectable and ends an operator's licence. Individual unlucky sessions feel rigged; the math doesn't care.

A scheme where the operator commits to the round outcome via a hash before play, then reveals the seed afterwards so you can verify the cards were not altered mid-round. Combines client seed + server seed + nonce → HMAC-SHA-512 → card draws. Doesn't replace operator-level audit, but verifies individual rounds.

RNG: software-driven, fresh virtual deck per hand, 60-150+ hands/hour, lower minimums. Live: real human dealer at a real table broadcast over webcam, 40-80 hands/hour, multi-player social atmosphere, higher minimums. Same math, different pace and feel. See /games/blackjack/online-vs-live.

Conservative bankroll management: bet size ≤ 1% of session bankroll. So a $1,000 buy-in supports $5-$10 bets. Variance per hand is ~1.15× bet size; over a session of 100 hands the one-sigma swing is ~$290 at $25/hand. Bet sizing matters more for variance than for edge.

Only with card counting at a brick-and-mortar table with hand-shuffled deep-penetration shoes — and counting flips the 0.41% house edge into a ~0.5% player edge. Online: no. In-house RNG and live-dealer studios both eliminate the persistent shoe state counting depends on.

Sitting at a 6:5 table when a 3:2 table is available. The payout difference adds 1.39% to the house edge — bigger impact than every other rule combined. Second-worst: not using the chart at all (~2.1% extra edge). Both compound.

No. Perfect Pairs (6%), 21+3 (3-8%), Bust It (6.9%), Lucky Ladies (17%), Insurance (7.4%) — every common side bet runs at 5-15× the edge of the main game. Side bets exist because they have a much higher house edge, packaged as variance entertainment. Skip every one.

In Vegas / Atlantic City pits, yes — counters can still find tables with deep enough penetration to count profitably. Online: no. Crypto-casino blackjack reshuffles too aggressively for any counting system to work. Card counting is a brick-and-mortar advantage technique in 2026.

Recall bias plus the actual math: when the dealer shows 7-A, she makes 17+ about 75-83% of the time. Hard 16 vs 10 is the worst hand in the game — you lose ~77% standing and ~60% hitting. Surrendering when allowed saves half the bet. The dealer feels unbeatable on 16 because the math says she usually is.

1-12 is the standard solo-counter spread (1 unit at TC ≤ +1, scaling up to 12 units at TC ≥ +5). Higher spreads (1-16, 1-20) bring more EV but more attention from pit bosses. Unit size is typically 0.5%-1% of bankroll. See /games/blackjack/card-counting for the full betting-spread breakdown.

In a vacuum, yes — single-deck S17 DAS 3:2 has ~0.18% house edge vs ~0.41% for 6-deck. But almost every single-deck table on a casino floor pays 6:5 instead of 3:2, which adds 1.39% and more than wipes out the deck-count advantage. Always check the payout before celebrating "found single-deck."

The 18 highest-EV deviations from basic strategy that a Hi-Lo counter should make at specific true-count thresholds. Adding them to a Hi-Lo count adds ~0.07% to the player's edge. The most valuable single one is taking insurance at TC ≥ +3. Full table on /games/blackjack/card-counting.

Yes. Most crypto operators (Stake, BC.Game, Cloudbet, Bitcasino, Roobet, Shuffle, Rainbet) ship mobile-responsive blackjack via in-house RNG and live-dealer studios. Live-dealer tables in particular use 16:9 video that resizes well to portrait phones. Verify table rules in the in-game info panel before sitting.

Both. Skill caps the house edge at ~0.41% with perfect basic strategy; below that requires card counting and stops working online. Variance from there is luck — within any single session, swings of ±$300 at $25/hand are normal even with perfect play. The skill is making the chart automatic so the variance you ride is the lowest the game offers.