Plinko odds: rows, risk levels and the real percentages

Treat each peg as an independent fair coin. After n rows the ball has taken n left-or-right steps. The number of paths that finish with k steps toward one edge is the binomial coefficient C(n, k). If every bounce is 50/50:

P(k) = C(n, k) / 2^n

The board is usually symmetric, so the leftmost and rightmost buckets share the same path count. The centre collects the most paths. That is why edge multipliers can look huge: they pay for rare paths, not for a generous game.

This page is the maths twin of plinko gambling. Use that guide for the product loop, fairness claims and why a pretty bounce is not a proof. Use this page to price a board.

If a studio weights pegs, or maps a seed straight to a bucket instead of a path, the binomial table is the wrong model. Then you need the info-panel probabilities. No panel, no price.

Adults 18+ only. Plinko odds do not become plus-EV because you watched the last ten balls. This cluster sits in games and odds.

Row count is n. More rows means more buckets and a thinner tail. Raising n does not raise the chance of “hitting something good.” It moves more of the return into rarer landings if the operator puts the jackpot-style multipliers on the edges.

8 rows (256 paths, 9 buckets)

12 rows (4,096 paths, 13 buckets)

C(12, 0) = 1 → 0.0244%. C(12, 1) = 12 → 0.293%. C(12, 2) = 66 → 1.611%. C(12, 3) = 220 → 5.371%. C(12, 4) = 495 → 12.085%. C(12, 5) = 792 → 19.336%. C(12, 6) = 924 → 22.559% in the centre.

The two outermost buckets are about 16 times rarer than on 8 rows. A 1,000x sticker on a 12-row edge is still a short pay versus the 4,096x fair price.

Fair 12-row payouts if the house paid true odds: 4096x, 341.3x, 62.1x, 18.6x, 8.3x, 5.2x, 4.4x in the centre. No live board pays that column. A “1,000x high risk” sticker on k=0 is a 76% haircut versus fair. Always compare m_k to 1/P(k), not to last night’s screenshot.

16 rows (65,536 paths, 17 buckets)

C(16, 0) = 1 → 0.001526%. Fair edge payout is 65,536x. C(16, 8) = 12,870 → 19.638% in the dead centre. Operators love 16-row high-risk tables because a 1,000x or 2,000x label looks generous next to a path that almost never happens.

A 1,000x cell on k=0 pays about 1.5% of fair. A 100x cell on k=2 pays about 18% of fair. Those two sentences are the whole “jackpot row” story.

Count buckets on the screenshot: rows + 1 if the layout is the usual one-bucket-per-k. If the count does not match, stop using the binomial row and read the panel.

Low, medium and high risk are three multiplier vectors on the same P(k). They are not three different bounce physics.

Write the plinko payout table as a vector m_k. Expected return per $1 is Σ P(k) m_k. House edge is 1 minus that sum. Risk is how the operator spreads the same (or a nearby) sum across k.

• Low risk. Centre m_k sits near 0.5x to 1.1x. Edges sit in single digits. Session standard deviation is small.
• Medium risk. Centre drops further below 1x. Mid-side buckets rise. Variance rises.
• High risk plinko. Several centre buckets can sit at 0.2x. The two outermost buckets take three-digit multipliers. Most balls lose. A few balls print a screenshot. That is a variance skin, not a better plinko RTP.

If two tables are built to 99% RTP, every risk level returns about 99 cents per dollar dropped. High risk does not “pay better.” It concentrates the same return into fewer balls. That is a variance choice, the same family of idea as a 14x colour versus a 2x colour.

Do not compare risk labels across brands. “High” on site A can be milder than “medium” on site B. Compare m_k, not the adjective.

A useful invariant: if you scale every multiplier by the same constant c, RTP scales by c and the edge becomes 1 − cE. Houses do not do that. They lower the centre and raise the tails so that Σ P m stays near a target. When a streamer says “high risk has better odds,” ask for E. If they cannot produce it, they said “higher screenshot” and called it odds.

The house edge of a plinko board is not a vibe. It is a finite sum.

1. List every bucket multiplier m_i.
2. Attach P_i from the binomial table for that row count, or from the info panel if pegs are biased.
3. Compute E = Σ P_i m_i. That is expected return per $1, also called RTP as a decimal.
4. House edge = 1 − E.

A 98% RTP board has a 2% edge on every ball. RTP explained is the definition page. This page only needs the identity so you can finish a sum.

Speed is a multiplier on the dollar leak

The edge applies to total wagered, not to the deposit. Fifty $1 balls are $50 of action. Recycled wins can turn the same $50 into $200 of action. At 2%, expected cost is $4, not $1. Expected value gambling is the dollar form of the same line: EV = (E − 1) × stake per ball.

If the site hides P_i, treat the advertised RTP as a claim. Some operators ship several RTP skins of the same art. Read the instance in front of you.

Same 256 paths. Two multiplier rows built to land near 99% RTP. These are teaching numbers, not a promise that any live brand uses them.

Low-risk multipliers

5.6x, 2.1x, 1.1x, 1.0x, 0.5x, 1.0x, 1.1x, 2.1x, 5.6x

Contribution to E (paths × multiplier):

E = 253.4 / 256 = 0.9898. RTP ≈ 98.98%. House edge ≈ 1.02%.

High-risk multipliers

29x, 4x, 1.5x, 0.3x, 0.2x, 0.3x, 1.5x, 4x, 29x

E = 253.6 / 256 = 0.9906. RTP ≈ 99.06%. House edge ≈ 0.94%.

Same order of leak. Opposite session shape. On low risk, a $1 ball often returns $0.50 to $2.10. On high risk, the centre 27% of balls return $0.20, and the 0.39% edges return $29.

A 100-ball, $1 session has about $1 of expected cost on either table. The high-risk path has a much wider distribution around that $1.

Once E is fixed, risk only changes the second moment. A rough check: variance per $1 ball is Σ P_i (m_i − E)^2.

On the low-risk row, most mass sits within about 0.5 of E. On the high-risk row, the 29x cells dominate the variance even though they are 2/256 of the mass: each contributes (29 − 0.99)^2 / 256 ≈ 3.07 to the variance from one side, about 6.1 from both edges, before the 0.2x centre adds a little more. You do not need a perfect variance number to use the board. You need to know that a 29x cell at 0.39% will not “show up tonight” just because the last twenty balls died in the middle.

Independence: P(k) does not update after a centre landing. The next ball is still C(n, k)/2^n. Streaks are samples, not a debt the pegs owe you.

Expected balls to first edge

On 8 rows, P(k=0 or 8) = 2/256 = 1/128. Expected balls to the first outer hit is 128 if you stop at the first one. That is a mean, not a promise the 129th ball pays. On 16 rows, P(k=0 or 16) = 2/65,536. Expected balls to that pair is 32,768. A 200-ball “session” on 16-row high risk is almost all centre and near-centre. Plan the screenshot rate with 1/P, not with hope.

If you came here from a stream overlay that shows “hot edges,” the overlay is decorating a memoryless process. The percentages on this page do not move.

Open the info panel. Write:

1. n (rows) and bucket count.
2. Whether pegs are 50/50. If yes, copy P(k) from the tables above.
3. The multiplier row for the risk you will actually click.
4. E = Σ P m. Edge = 1 − E.
5. Expected cost = stake × balls × edge.

If step 2 or 3 is missing, you cannot complete the sum. Skip the board. The general plinko gambling guide is the product page when you need rules, auto-drop cost, and why PVPspinArena does not run a drop game.

PVPspinArena publishes a 15-slot wheel instead. You can count the slots on Roulette and compute a 6.67% edge without a binomial table. Jackpot and Coinflip use pot share, not a bucket row.

Do not import a “row strategy.” Changing n or risk changes variance. It does not create an edge. The only plus-EV move on a 1–3% board is not dropping the ball.

Paper check you can do in sixty seconds

Pick the board on screen. Write n, risk label, and the nine or more multipliers. If n is 8, copy P from the 256-path table. Multiply down the row. If E lands between 0.97 and 0.99 you have a typical crypto board. If E lands at 0.94, the art is the same and the leak is fatter. If you cannot finish the multiply, you do not have plinko odds. You have a cartoon.