Plinko drops a ball through a pegboard into payout buckets — and underneath the animation is the cleanest probability object in the originals family: a binomial distribution. The ball bounces left or right at each row with equal chance; where it lands follows the bell curve, and the paytable prices each bucket against it.
The machine under the pegs
At 16 rows, the ball makes 16 left/right choices. Landing k steps from the leftmost bucket happens C(16,k)/65,536 of the time — the centre buckets are enormously more likely than the edges:
| Bucket (16 rows) | Probability |
|---|---|
| Centre | ~19.6% |
| Mid-board | a few % each |
| Edge | 1 in 65,536 |
Paytables invert the curve: centre buckets pay below 1x (often 0.2–0.5x at high risk), edges pay the headline multipliers (up to 1000x at 16-row high risk). Multiply each bucket's probability by its payout and sum: the result is the game's published RTP, commonly 97–99% (edge 1–3%) — stated on the info page, enforced by the provably fair draw.
If you want the math: the provably fair implementation draws the 16 bounce directions from the seed HMAC — the path, not the bucket, is the random object. Verification recomputes the path; the bucket follows.
What the settings change
Rows (8–16): more rows = finer distribution = fatter headline multipliers at rarer edges. Expected return barely moves; the published RTP per configuration is the check.
Risk (low/medium/high): reshapes the paytable on the same physics. Low risk compresses payouts toward 1x (long sessions, small swings); high risk drains the centre buckets to fund the edges (most drops lose money, rare drops spike). Same edge family, radically different session shape.
| Setting | Most drops | Session profile |
|---|---|---|
| Low risk, 8 rows | ~0.5–1.5x | Grind, low variance |
| High risk, 16 rows | 0.2x centre | Bleed punctuated by spikes |
Reading a plinko paytable honestly
1. Find the published RTP for your exact rows/risk combination — implementations differ across configurations, not just operators. 2. Note the centre-bucket payout: that is what most drops return. A 0.2x centre means the typical drop loses 80%. 3. The headline multiplier's bucket probability is computable: at 16 rows, 2 × C(16,0)/65,536 ≈ 1 in 32,768 drops. Expect to buy a lot of drops between spikes.
Auto-drop multiplies pace exactly as auto-bet does in dice — set stop conditions or the binomial grinds uninterrupted.