Designing Bingo: 38 balls, 12 lines, and why the centre star is worth a whole ball

The card: five columns, one free star

A 75-ball bingo card is not a random 5×5 grid. Each column draws from its own range of fifteen numbers, B from 1 to 15, I from 16 to 30, N from 31 to 45, G from 46 to 60 and O from 61 to 75, and the centre cell of the N column is a free star that counts as already marked. That is the traditional American card, and the engine builds it the traditional way.

The free star is not decoration. Four of the twelve lines pass through it: the middle row, the middle column and both diagonals. Those lines need only four drawn numbers instead of five, and on a 38-ball draw that difference is large.

The draw: 38 of 75

Thirty-eight balls are drawn without replacement from 75, so any particular number on your card is drawn with probability 38/75, a little over half. From that one figure the chance of each line follows directly.

Drawing without replacement: each factor is the chance the next needed number is among the balls still to come. The expected number of completed lines per card is 4 × 6.07% + 8 × 2.91% ≈ 0.48.

So the star roughly doubles the chance of every line it touches, which is why the four centre lines carry most of the game: they complete about 24 times per hundred cards between them, against about 23 for the other eight combined. A card whose middle row and middle column both light up has done the likely thing, not the lucky one.

The paytable: pay by count, not by pattern

Hall bingo pays the first card to finish a pattern. With one card and no race, that design has nothing to pay for, so solo bingo pays by how many lines the card completes. The engine keeps seven tiers.

Multiply each share by its multiplier and add: the sum lands within rounding of the 96.5% return the header states. The 37.1% hit rate is simply one minus the 62.9% of cards with no line.

Two design choices are visible in that table. The first tier pays ×1.5 rather than ×2, which is where most of the house edge is collected: a single line is the common win, and paying it below double keeps the game near its target return while the rare counts can be generous. The second is the steepness above three lines. Four lines at ×30 and five at ×100 happen a few times per thousand cards, and they are what the paytable is built to make memorable; the ×500 tier for seven or more lines is reachable only when the whole card is nearly full, which on 38 balls is a once-in-many-thousands event the header cannot even resolve beyond “about 0.005%”.

Why 38 balls

The ball count is the single dial that sets the whole shape of the game. Draw fewer balls and lines almost never complete; draw more and every card fills. Thirty-eight is a little over half the pool, which puts the single-line chance for a centre line near 6% and the no-win share near 63%, a ratio that makes a card feel live without making wins routine. Move the count to 45 and the expected line count roughly doubles; move it to 30 and it halves. The paytable would have to move with it.

It is the same trade the keno risk dial makes with its spot count and Hanabi makes with its symbol weights: the operator fixes a return, then chooses where along the frequency-versus-size curve the game should sit. Bingo’s designer chose frequent small wins, one card in three, and a long thin tail.

Replaying a card

Because the card and the balls both come from the round’s seed stream, a bingo round is one of the easiest on the site to verify: 63 cursors, drawn in a fixed order, from HMAC-SHA256 of the committed server seed over your client seed, the nonce and the cursor. The first 25 rebuild the card, the next 38 rebuild the balls, and counting the completed lines against the paytable reproduces the payout. The verification walkthrough covers the mechanics; the bingo derivation is the same function the demo uses to deal.