Andar Bahar’s random ×5 boosts don’t change the house edge — here’s the proof

The game in 20 seconds

A joker card is drawn from a 52-card deck. The remaining cards race onto two piles — ANDAR (inside) first, then BAHAR (outside), alternating — until a card matches the joker’s rank. The side that catches the match wins. You can back either side, or bet on how long the race runs: buckets of five cards, from 1–5 out to 46–49.

Then the wrinkle this article exists for: every deal, exactly two buckets get stamped with a random boost — ×2, ×3 or ×5 — revealed with a flourish just before the joker flips. A ×750 bucket under a ×5 boost pays ×3,750. It feels like the game’s odds just moved. Let’s check whether they did.

Where the numbers come from

After the joker leaves the deck, 51 cards remain and exactly 3 of them share the joker’s rank. The race ends the moment one of those three appears. The probability that the match lands on exactly the k-th raced card is a classic first-match distribution:

P(count = k) = C(51−k, 2) / C(51, 3), for k = 1 … 49, with C(51,3) = 20,825

In words: the match sits at position k exactly when the other two rank-mates hide among the 51−k cards behind it. Because only 48 non-matching cards exist, the race can never reach card 50 — every round ends by card 49. Summing odd and even positions gives the two sides their exact win rates.

Main bets: the exact odds

ANDAR receives the first card, so it catches the match slightly more often — 51.50% exactly — and pays slightly less. This is the standard shape of the game everywhere it is dealt; what varies between houses is only the two payout numbers.

The boost system, mechanically

• Every round, exactly two distinct buckets are chosen for boosts by the seed pair (cursors 0–1 of the round’s derivation).
• Each boost’s magnitude comes from its own cursor: ×2 with 65%, ×3 with 25%, ×5 with 10% probability.
• Boosts are revealed only after the deal starts — betting is locked first. You can never place a bet on a bucket you already know is boosted.
• Boosts multiply side-bet (bucket) wins only. ANDAR/BAHAR main bets are untouched.

Why boosts cannot move the return

Three structural facts, all checkable in the derivation code, pin the long-run return in place:

1. You bet blind. Bets lock in the idle phase; the boost stamp happens after the deal begins. There is no way to select into boosted buckets.
2. The draws are independent. Boost placement uses cursors 0–3 of the seed derivation; the joker and the race use cursors 4 onward. Which bucket wins tells you nothing about which bucket was boosted, and vice versa.
3. The expectation is a constant. Any given bucket carries a boost with probability 2/10. The average boost, when present, is 2×0.65 + 3×0.25 + 5×0.10 = ×2.55. So the multiplier applied to any bucket bet averages 0.8×1 + 0.2×2.55 = ×1.31 — every bucket, every round, forever.

That constant is the whole trick, and it is an honest one: the base bucket payouts are deliberately set low — around 72–74% return on their own — precisely so that the ×1.31 boost expectation lifts them back to a normal 94–97%. The boost is not a gift on top of fair odds. It is part of the odds, paid out in lumps instead of a trickle.

The full bucket table

Headline hit: 46–49 at ×750 under a ×5 boost = ×3,750 total return. Probability of the bucket itself: 20 in 20,825 races.

What the boosts actually change: variance

If the expected return is pinned, what did the boosts buy? Spread. A bucket that would pay you a steady ×2.7 now pays ×2.7 most rounds and occasionally ×13.5. Long-run value is identical; the shape of the ride is not. That is a deliberate design choice — lumpier payouts are more fun to watch — and it is the same trade a lottery makes against a savings account, in miniature.

The practical takeaway: seeing a ×5 land on a bucket you didn’t bet is not a missed edge, and betting “because a big boost is due” is exactly the fallacy we wrote about separately. The race does not remember, and the boosts were priced before you sat down.