Plinko at Stake

What Stake publishes about Plinko

Plinko is one of 31 games in Stake's Originals group, built and priced by Stake itself, so every number that matters comes from Stake's own pages. The game page carries an RTP badge of 99.00% and describes the game as having "a 10,000x max win, a 1% house edge and a 99% RTP" (Stake Plinko game page, 1 Nov 2026). Two settings decide which payout table a drop uses: the number of rows and the risk level. Stake's fairness documentation puts the row range in one sentence: "Players can choose between 8 and 16 pins of play".

The ball does not decide anything on screen. The slot is computed from your seeds the moment the bet is placed, and the bouncing animation is drawn to end in that slot. Risk level changes no probability at all: it swaps the multipliers printed on the same 9 to 17 slots.

The 16-row board, slot by slot

At 16 rows the ball makes 16 left-or-right steps, which gives 65,536 equally likely paths and 17 slots. The number of paths into each slot is a binomial coefficient, so the centre slot collects 12,870 paths and each outermost slot exactly one. The multipliers below are Stake's published 16-row tables; the chances are arithmetic from the 50/50 step in Stake's formula.

Chance is for the slot and its mirror image together, except the centre. Multipliers: Stake payout table as published on 1 Nov 2026.

The 10,000x slot is reached by one path in 65,536 on each side, so the chance of hitting either of them is 1 in 32,768 per drop. At an assumed pace of one drop every two seconds that is, on average, a little over 18 hours of continuous play between hits, and an average says nothing about any particular run: each drop is independent of the last.

Top multiplier for every row count

Fewer rows make the outermost slot far easier to reach and pay far less for it. The table lists the top multiplier of all 36 tables, with the chance of landing in either outermost slot (2 paths out of 2 to the power of the row count).

Arithmetic from the published tables. The 16-row High table tops out at 1,000x; only Expert goes higher.

Return of each table: arithmetic from the published multipliers

Stake states one figure, 99%, for the whole game. Each of the 36 tables has its own exact return, which is the sum of every slot's multiplier times its binomial probability. Because the multipliers are rounded to one or two significant figures, no table lands on 99% exactly:

The spread runs from 98.906% (8 rows Medium and 10 rows Medium) to 99.160% (11 rows High). Eighteen tables come out a fraction under 99.000% and eighteen a fraction over. The largest shortfall against the stated figure is 0.094 percentage points, which is about 9 cents per $100 wagered, and the largest surplus is 0.160 points. In practice the choice of rows and risk moves the expected cost of a drop by less than a quarter of a percentage point; what it changes a great deal is how that cost arrives.

What the risk level changes: variance, at the same cost

All four 16-row tables return within 0.03 percentage points of each other, yet they behave nothing alike. The table shows how often a drop pays back less than the stake and how often it pays a multiple.

On Low risk four drops in five return the stake or more, and the worst slot still pays half. On High and Expert about 79% of drops pay 0.2x or 0.1x, and the return is carried by slots that are hit a few times in a thousand drops or less. Expert moves value from the middle ring (2x to 9x on High) out to the two edge slots, which is why its 2x-or-better share is a third of High's. None of this alters the roughly 1% long-run cost. It decides whether a balance drains slowly or in steps with rare large refunds.

The profit cap and the 10,000x slot

Stake's help centre sets a maximum bet and a maximum profit per bet for all Stake Originals, in units of each coin, plus a minimum bet of $0.0001. On a game with a 10,000x slot the profit cap binds long before the bet cap. A 10,000x hit is a profit of 9,999 times the stake, so the largest stake whose top prize fits under the cap is the cap divided by 9,999:

Limits: Stake help centre, "Crypto: Casino Betting Limits", dated 24 Aug 2026 and unchanged on 12 Nov 2026. Last column is our arithmetic.

In USDT that is about $250 per drop. The article lists the caps and does not say whether a larger bet is refused or paid only up to the cap, so anyone staking more than these amounts on Expert at 16 rows should check the bet panel first. Local-currency wallets have their own table in a second help-centre article (for example EUR 1,000,000 bet and EUR 100,000 profit).

How one drop is derived from your seeds

Stake's "Implementation" page defines four inputs for every Original: a server seed, a client seed, a nonce and a cursor. The server seed is "a random 64-character hex string" whose SHA-256 hash you are shown before betting; the client seed is yours to edit; the nonce "increments as every new bet is made". For Plinko the steps are:

1. Stake computes HMAC-SHA256 with the server seed as the key and the text clientSeed:nonce:0 as the message. That yields 32 bytes.
2. Bytes are taken four at a time and turned into a number between 0 and 1: byte 1 divided by 256, plus byte 2 divided by 256 squared, and so on. Thirty-two bytes give eight numbers.
3. A board of more than eight rows needs more than eight numbers, so the message is hashed again with the cursor raised to 1 (clientSeed:nonce:1). The documentation lists Plinko among the games with "2 increments per game to cover possible 16 decisions".
4. Each number becomes one step: multiply by 2 and round down. 0 is left, 1 is right.
5. The slot is the count of right steps. Zero rights is the far-left slot, 16 rights on a 16-row board the far-right slot. The multiplier is read from the table for your rows and risk.

float = b1/256 + b2/256² + b3/256³ + b4/256⁴
direction = floor(float × 2)  →  0 = left, 1 = right
slot = number of right steps  →  multiplier from the rows-by-risk table

Two details follow from the arithmetic. First, only the first byte of each four-byte group can change a step: a first byte of 128 to 255 is always right and 0 to 127 always left, so each step is an exact 50/50 and the published binomial odds hold without approximation. Second, the code sample on Stake's "Game events" page declares an index called DIRECTIONS and then reads from CARDS[Math.floor(float * 2)], a leftover from the card games on the same page. The description around it is unambiguous, and Stake's own calculator follows the left/right reading.

To check a drop after the fact you need the unhashed server seed, which Stake releases only when you rotate the seed pair: "To reveal the server seed from its hashed version, the seed must be rotated by the player". Then:

• Commitment: the SHA-256 of the revealed seed must equal the hash shown before your bets. Stake has an "unhash" page for this and any SHA-256 tool gives the same answer.
• Outcome on Stake: the calculation page under Provably Fair takes the seeds, the nonce, the risk level and the row count and shows the result of that drop.
• Outcome outside Stake: our Stake seed checker runs the same HMAC-SHA256 steps in your browser for 8 to 16 rows and prints the left/right path and the landing slot. It does not look up the multiplier; read that from the table for your risk level. The Plinko verifier page uses the same code.

What this proves is narrow and useful: the seed was fixed before the bet and the slot follows from it. It does not prove anything about the multipliers, which are a pricing decision, and it does not change the expected cost. Background: what provably fair gambling proves and the general verifier.

Stake Plinko against Shuffle, Roobet and Gamdom

The same game is priced differently from one casino to the next, and the documents show where.

Shuffle: game page and help centre, 12 Nov 2026. Roobet and Gamdom: game pages and fairness pages, 1 Nov 2026.

Shuffle's nine Low, nine Medium and nine High tables are the same numbers as Stake's tables of those names (both as carried in the games on 1 Nov 2026), and its documentation describes the same byte-to-float generator, so a Shuffle drop and a Stake drop with identical seeds take the same path. Stake's difference is the fourth risk level and the per-coin limits. Roobet and Gamdom use a different hash construction, which is why our checker reproduces Stake and Shuffle paths but not theirs.

Sources: Stake fairness documentation, implementation, conversions and game events (1 Nov 2026); Stake Plinko game page and payout table (1 Nov 2026); Stake help centre, crypto betting limits, local-currency betting limits and provable fairness (12 Nov 2026). Probabilities, table returns and stake ceilings are our arithmetic from those figures.

Licence, KYC, withdrawals and bonus terms for Stake are in the Stake review; the Plinko calculator turns any table into expected loss per session.

Frequently asked questions

Related pages