Stake Originals Plinko generates a left-or-right decision for each of 8–16 rows, then pays the multiplier attached to the landing bucket. Stake advertises 99% RTP and a 1% house edge. On 16 rows, Hard mode has two 1,000x edge buckets: the chance of reaching either is 1 in 32,768 per drop. Expert raises the maximum to 10,000x.
For the 16-row Hard board, our 1,000x odds and waiting-time analysis separates the chance of either edge from one edge and shows why 32,768 drops offer no guarantee.
How the board, rows and risk settings work
The board resembles a Galton board, but the digital result comes from random numbers rather than a physical ball's weight or bounce. Stake's published algorithm converts each random value into one of two directions. Counting the right turns gives the final bucket index, starting at 0 on the left. Eight rows produce nine buckets; 16 rows produce 17.
The row count determines the landing probabilities. The risk setting assigns payouts to those positions. Changing risk on the same board leaves the edge probability unchanged; it changes how much that rare landing pays and how much the common middle landings return.
We use Easy, Medium, Hard and Expert, the names on Stake's current game page. Older guides call Easy “Low” and Hard “High”; these are the corresponding older labels, not two additional modes. Expert is the fourth setting, which older three-mode tables omit.
Maximum multipliers for every row count
Our payout table reference separates these Stake maxima from verified full BGaming strips and compares loss frequencies.
The maximum payout schedule published by Stake is:
| Rows | Easy | Medium | Hard | Expert |
|---|---|---|---|---|
| 8 | 5.6x | 13x | 29x | 50x |
| 9 | 5.6x | 18x | 43x | 100x |
| 10 | 8.9x | 22x | 76x | 201x |
| 11 | 8.4x | 24x | 120x | 324x |
| 12 | 10x | 33x | 170x | 619x |
| 13 | 8.1x | 43x | 260x | 1,012x |
| 14 | 7.1x | 58x | 420x | 2,369x |
| 15 | 15x | 88x | 620x | 5,000x |
| 16 | 16x | 110x | 1,000x | 10,000x |
More rows do not always mean a larger maximum: Easy falls from 10x at 12 rows to 8.1x at 13. Compare the actual payout row, including the middle buckets, before interpreting a difficulty label. A multiplier is the total return including the stake: in an illustrative A$1 calculation, 0.2x returns A$0.20 and loses A$0.80; 1x breaks even; 1,000x returns A$1,000 for A$999 net profit. AUD examples here are arithmetic, not a claim that Stake offers Australian dollar deposits.
The exact odds of an edge landing
With n independent 50/50 decisions, bucket k has probability P(k) = C(n,k) / 2^n. Here k is the number of right turns and C(n,k) = n! / (k! × (n−k)!). Many paths converge on a middle bucket; each edge has just one path, either all left or all right.
| Rows | One specified edge | Either edge |
|---|---|---|
| 8 | 1 in 256 (0.390625%) | 1 in 128 (0.78125%) |
| 12 | 1 in 4,096 (0.024414%) | 1 in 2,048 (0.048828%) |
| 16 | 1 in 65,536 (0.001526%) | 1 in 32,768 (0.003052%) |
For 16 rows, (1/2)^16 = 1/65,536 is the probability of one named edge. Counting both gives 2 × (1/2)^16 = 1/32,768. The central bucket, index 8, has 12,870 possible paths and probability 12,870/65,536 = 19.6381%. The five central buckets, indices 6–10, together receive 78.9886% of drops. Those figures describe landing positions; the selected payout schedule determines whether a landing makes a profit.
Why 32,768 drops do not guarantee a 1,000x
For independent drops with edge probability p, the chance of at least one edge in m drops is 1 − (1 − p)^m. At 16 rows it is approximately 3.01% after 1,000 drops, 26.30% after 10,000 and 63.21% after 32,768. There is still about a 36.79% chance of no edge after that last total. The average waiting time is not a deadline, and a long losing run does not make the next ball more likely to reach an edge.
What 99% RTP means for a bankroll
RTP is the probability-weighted sum of the return multipliers: RTP = Σ P(k) × multiplier(k). To audit a particular board, use every bucket in that board's payout row. A top multiplier alone cannot establish its RTP, and a breakeven-or-better hit rate is a separate calculation: add only the probabilities of buckets paying at least 1x.
Using the advertised 1% edge, A$1,000 of total stakes has an expected loss of A$10. Total stakes means turnover: reusing A$20 across fifty A$1 drops produces A$50 of turnover, not A$20. Neither example predicts a session's actual balance. High volatility can produce a much larger loss before a rare payout contributes to the average.
For a practice session, choose the rows and difficulty first, record the starting virtual balance and set a fixed number of drops. Keep the stake constant so the comparison measures the board rather than changing bet sizes. Our free Plinko demo guide explains practice options; the Plinko bankroll guide covers loss limits and the cost of increasing stakes after losses.
If a losing run makes you question the game, our Plinko rigging and predictor guide separates expected streaks from reproducible result discrepancies and payment disputes.
The Martingale strategy analysis separates a profitable ball from recovery of the whole sequence, with a fixed-stake comparison and conditional 16-row streak probabilities.
How the seed inputs produce a result
The server seed is committed through its SHA-256 hash before betting and revealed after rotation. The player supplies a client seed; a nonce distinguishes successive bets. A hash commitment and a reproduced result are two separate checks: one tests whether the revealed input matches the earlier commitment, the other tests the calculation.
The published byte generator uses the server seed's text as the HMAC-SHA256 key and a message in the form clientSeed:nonce:block. Block 0 yields 32 bytes. Four bytes make each random float: b0/256 + b1/65,536 + b2/16,777,216 + b3/4,294,967,296. Values below 0.5 go left; values at least 0.5 go right. Eight decisions fit in one hash block; rows 9–16 require block 1. The final bucket is the sum of right turns.
The server seed remains secret during active play, so a client seed phrase is not a method for choosing a winning path. Its useful role is providing a player-controlled input to the verification system. For a broader explanation of commitments, see our provably fair Plinko guide.
Set a client seed and verify a recorded drop
To choose your own client seed, open the game's Fairness controls and edit the client seed there. Save the new text and its associated server hash before recording a round. A seed change affects subsequent inputs; use the previous client seed when checking a bet placed before that change.
- Save the original commitment. Open Fairness and record the active server seed hash and client seed before the round you want to check.
- Save the bet record, its nonce, row count, difficulty and displayed result. Use the nonce attached to that bet, not the counter shown after further bets.
- Rotate the server seed through the fairness controls to reveal the previous seed. Use that revealed seed with the client seed from the old round.
- Calculate SHA-256 of the revealed server seed and compare all 64 hexadecimal characters with the saved commitment.
- Recreate the left/right path using the round's nonce and rows. Compare the resulting bucket with the bet record, then match that bucket to the payout schedule used for the round.
Seed rotation and other Stake Originals
Keep the inputs with each individual bet when moving between Originals or rotating seeds. Do not infer a Plinko nonce by counting only the balls you remember dropping: the verification needs the actual recorded number. Rotation exposes the old server seed and replaces it, so copying the new active hash into an old round's check mixes two seed pairs. A mismatch should first prompt a check of the saved inputs.
The checker here reproduces the commitment hash and landing path locally. Its sample uses invented demonstration seeds, not a recorded Stake wager. It deliberately leaves payout validation to the round's own multiplier row, so an old payout table cannot silently change your result.