STAKE HILO STRATEGY & ODDS

Updated Oct 202611 min read

Hilo is the Stake Originals card game where you guess whether the next card is higher or lower than the one on screen, building a multiplier streak you can cash out at any time. It runs at a sourced 99% RTP, a 1% house edge, lets you skip up to 52 cards, and — uniquely — shows you the real probability on each guess. This guide ties those on-screen percentages to the actual draw math, verifies a real card from its seeds, and answers the only question that matters: when to cash out. For the pure step-by-step proof, see our verify provably fair Hi-Lo walkthrough.

How Stake Hilo Works

A base card is dealt face-up. You bet whether the next card will be higher (or the same) or lower (or the same). Guess right and the drawn card becomes the new base, your multiplier grows, and you choose again. Guess wrong and the streak ends. At any point you can cash out the accumulated multiplier, and you can skip a card whenever the odds look unappealing. Hilo draws each card independently from a full 52-card deck — like an infinite shoe — which is why the same card can legitimately repeat.

The defining feature is honesty about odds: Stake displays your exact win probability and the corresponding payout for both "higher" and "lower" on every card. Those numbers are not decoration — they are computed straight from the deck, and the next section ties them to the math.

Stake Hilo RTP & House Edge

Hilo returns 99% to players, a 1% house edge. Each guess's payout is calculated as the fair odds shaved by that 1%: multiplier = 0.99 / win probability. So if the deck gives you a 75% chance of being right, the offered payout for that guess is 0.99 / 0.75 = 1.32x, not the fair 1.333x — the gap is the house edge on that single step. String several guesses together and the edge compounds, which is the entire reason long streaks are so rare.

Tying the On-Screen % to Real Draw Math

Here is the part competitors never show: where those percentages come from. Stake ranks cards 2 through Ace (Ace high) and draws from a full 52-card deck. For a base card of rank R, the probability the next card is higher or equal is the count of deck cards ranked ≥ R divided by 52; the probability it is lower or equal is the count ranked ≤ R divided by 52. (Because every draw is from a fresh full deck, the four cards of the base rank itself count toward both "or equal" sides, which is how ties are handled.)

Base card	Cards ≥ this rank	P(higher or equal)	Cards ≤ this rank	P(lower or equal)
2 (lowest)	52	100.0%	4	7.7%
5	40	76.9%	16	30.8%
8	28	53.8%	28	53.8%
10	20	38.5%	36	69.2%
K	12	23.1%	48	92.3%
A (highest)	4	7.7%	52	100.0%

This table is the on-screen display. Land a base 2 and "higher or equal" reads ~100% (every card is ≥ 2) for a tiny payout, while "lower or equal" reads ~7.7% for a roughly 12.8x payout. Land a King and the smart guess flips: "lower or equal" is ~92.3% for a small payout, "higher or equal" is ~23.1% for a far bigger one. The game is honest about all of this — the percentage you see equals the count of qualifying cards over 52, every time.

The Skip Button: Up to 52 Cards

When the base card sits in the awkward middle — a 7, 8 or 9, where both directions hover near 50% and neither payout is attractive — the skip function deals a new base card without ending your streak. You can skip as many times as you like, in principle cycling through all 52 ranks until a favourable base appears (an extreme low or high card, where one direction is near-certain and the other pays well).

Skipping does not change the house edge — the new card is still drawn from a fresh full deck at the same 1% edge — but it is a genuine variance tool. Used patiently, it lets you only ever commit on cards where the probability/payout trade-off suits your plan, rather than being forced to guess on a coin-flip 8.

Provably Fair Hi-Lo Game: How It Works

What makes hi-lo a provably fair game rather than a trust-me game is that every card is fixed cryptographically before you guess. The casino commits to a hashed server seed up front, you control a client seed, and each card in your hand is drawn from an HMAC-SHA256 stream over clientSeed:nonce:cursor — so the operator cannot swap the next card after seeing your bet, and you cannot predict it before. After you rotate your seed pair, the unhashed server seed is revealed and you can re-run the exact derivation to reproduce every card of the round. That after-the-round verifiability is the entire difference between a provably fair hi-lo game and the higher/lower card games at conventional RNG casinos, where you simply trust an auditor's certificate.

Not every crypto casino offers hi-lo as a provably fair Original. Among the operators we rank, Stake is the reference implementation (its Hilo is the game this guide is built on — see our Stake review), FortuneJack includes Hilo in its provably fair originals suite (review), and Bitsler runs Hilo within one of the larger in-house PF game collections. Lucky Block lists a Hilo title through PF-capable providers, though coverage there is thinner. For the full comparison of which operators implement provable fairness genuinely — and on which games — see our ranked provably fair crypto casinos page.

Two honest caveats. First, provably fair proves the card sequence was untampered — it does not remove the 1% house edge baked into every payout, so it is a fairness guarantee, not a winning edge. Second, the verification only works if you actually use it: rotate your seed, take the revealed server seed, and reproduce a card. The next section walks through that derivation step by step.

Provably Fair Verification: Reproducing the Card

Hilo is provably fair, so the card you are dealt is fixed before you guess and revealed afterwards from the seed pair: the server seed (committed as a hash up front), your client seed, and the nonce. The derivation, as set out in our full Hi-Lo verification guide (our verifier runs the same HMAC and float steps for other Stake games, but has no Hilo mode):

• 1
  Build the HMAC. HMAC_SHA256(serverSeed, "clientSeed:nonce:cursor") produces a 32-byte hash. Hilo can need up to 52 cards in a hand, so the cursor increments to pull fresh hashes as required.
• 2
  Bytes to a float. Each card uses four bytes turned into a decimal in [0,1) via b₀/256 + b₁/256² + b₂/256³ + b₃/256⁴. A 32-byte hash yields eight cards before the cursor rolls.
• 3
  Float to a card index. index = floor(float × 52) gives a number 0–51, one slot per card. With a float of 0.8552, that is floor(44.47) = 44.
• 4
  Map to rank and suit. Rank is index ÷ 4 (integer division) over 2,3,…,10,J,Q,K,A; suit is index % 4 over Diamonds, Hearts, Spades, Clubs. Index 44 is rank 11 (Ace) suit 0 — the Ace of Diamonds. Re-running this from your seeds reproduces the same card, proving it was fixed before your guess.

Because each draw is an independent 1-in-52 pick from a full deck, the on-screen probabilities above hold exactly, and the same card can appear twice in a row as a legitimate tie. You chose the client seed and the server-seed hash was public first, so neither side can swap the card after the bet — the verification reproduces a value locked in advance.

Hilo Strategy: When to Cash Out

The honest foundation: no Hilo strategy beats the 1% house edge. The cards come from HMAC-SHA256, not from your guess history, and every guess's payout already bakes in the edge. Strategy in Hilo is entirely about which guesses you commit to and when you bank the streak.

Only Bet the Strong Side

Commit your guess only when the base card makes one direction clearly favourable — very low cards (bet higher) or very high cards (bet lower), where the win probability is 75%+ . On middling cards, skip rather than take a near coin-flip. This does not change EV, but it produces longer, lower-variance streaks.

Cash Out on Short Streaks

Because the edge compounds with every guess, the probability of extending a streak drops fast. Most experienced players bank after four or five correct guesses rather than chasing a long run that is statistically unlikely to complete. A streak of five favourable (75%) guesses already only completes about 24% of the time.

Use Skip as a Filter, Not a Crutch

Skip is powerful for waiting out bad base cards, but skipping forever does not build value — you only earn multiplier when you actually guess and win. Use it to avoid coin-flips, then commit decisively on a strong card.

Flat Bet + Session Limits (Most Practical)

The math-endorsed default, as in all our guides: flat-bet 1–2% of your bankroll, set a profit target and stop-loss before you start, and stop at either. No progression system converts Hilo's negative EV into a positive one.

Frequently Asked Questions