How the 0–100 Board Score Is Calculated

The score in one formula

Every board starts at 100 and loses points for four measurable defects. The result is clamped to 0–100 and rounded to a whole number. The function is analyze() in src/app.js:

score = round( clamp( 100 − 18×red − 4×same − 150×dev − 0.6×spread, 0, 100 ) )

One touching 6/8 pair costs as much as 4.5 same-resource pairs, so adjacent red numbers are the defect the formula punishes hardest. The label next to the score comes from the same script: 85–100 Excellent, 70–84 Good, 55–69 Fair, under 55 Unbalanced.

Measured over 4,000 generated 3-4 player boards: median 75, 90th percentile 82, and 85+ about one board in twenty. The label is a band, not a percentile — most rule-compliant boards land in the 70s, and a board only reaches Excellent when every rule holds and the six start spots are close together.

Worked example: the best of 30 candidates for seed method-demo

Rules code 1101 = no adjacent 6 & 8 on, no same numbers touching on, no 2 & 12 touching off, balanced resources on. The link in the caption reloads this exact board.

Under the score, the tool prints the inputs it used — this is that line on this board:

Red 6/8 touching: 0 · same-resource pairs: 2 · max pip skew: 2.3% · best start spots: 13, 12, 11, 11, 10, 9 pips

The six marked intersections, in the order the tool numbers them. No two of them share more than one hex — that is how the list is chosen.

How the 30 candidates are generated and picked

1. A click makes 30 boards. The page takes the base seed (typed, or six random characters) and generates candidate 0 from it, candidate 1 from seed#1, and so on up to seed#29.
2. Each candidate must pass your rules. Terrain is shuffled and rejected when a hex touches more than one hex of its own terrain (with Balanced resources on). Number tokens are shuffled onto the 18 or 28 number hexes and rejected when two 6/8 touch, when equal numbers touch, when two 2/12 touch (each only if that rule is on), or when any resource’s pip share is more than 7 percentage points from its hex share (with Balanced resources on). The loops give up after 5,000 terrain and 20,000 number attempts and keep the last layout, so a seed that cannot satisfy a rule can slip through.
3. Every surviving candidate is scored with the four-term formula.
4. The highest score wins. The comparison is strict, so a tie keeps the earlier candidate: on this seed #23 and #28 both score 85, and neither is the winner because #4 scores 86.

The rule switches are part of the random seed (seed + "|" + player count + "|" + rule settings), so changing one rule reshuffles all 30 candidates instead of filtering the same set. The candidate count is conditional too: with Balanced resources off, the page generates a single board and this loop never runs, which is why unchecking that rule can drop the score sharply.

The share URL stores only the base seed, the player count and the four switches. Opening it regenerates the board: with Balanced resources on, all 30 candidates are rerun and the same winner comes back; the 29 other boards are not kept.

All 30 scores for this seed, in generation order. The highlighted cell is the winner; #23 and #28 tie at 85 behind it.

Lowest 67 · median 76 · highest 86. Candidate #0 — the board a single-shot generator would have shown — scored 67. On this seed the selection step is worth 19 points, and the candidate you see is not the first one that passed the rules.

What the score does not measure

• Harbors. Ports are shuffled and drawn, but no scoring term reads them. A board with a 2:1 harbor on your resource scores the same as one without.
• Access. Pips are counted per resource across the whole board. The score never asks whether a legal intersection reaches the high-pip hex, or who can claim it.
• Number diversity inside a start. Diversity is a tie-breaker when the six starts are ranked (a three-resource spot beats a two-resource spot with the same pips); it does not enter the score.
• The other skews. dev is the single worst resource, not the mean or the sum: a board with one 7-point outlier and a board with a 7-point plus a 6-point outlier both lose 10.5 points.
• Anything after setup. The robber’s later moves, the harbors you take, cities, development cards and scenario rules are outside the formula.

The other half of reading a board is in how the per-resource pip totals work; the odds behind the pips are on the dice probability chart, and the rule list is on balanced board rules.

Questions

More tools & guides