What the table measures
| Level | Cards | Cost range | Mean cost | Mean prestige | Zero-point cards |
|---|---|---|---|---|---|
| 1 | 40 | 3–5 | 4.13 | 0.13 | 35 |
| 2 | 30 | 5–8 | 6.83 | 1.83 | 0 |
| 3 | 20 | 7–14 | 10.75 | 4.00 | 0 |
For each card, printed cost is the sum of its five coloured requirements. Mean cost is the sum of those totals divided by the number of cards in that level. Mean prestige is calculated separately. A zero-point card still supplies its printed permanent bonus; the last column does not count it as useless.
These are deck-level summaries, not the distribution of cards in your current market. Four face-up cards from a level are a small sample, and purchases continuously change that sample. The data describes the current implementation; different editions, expansion modules, and physical printings should be checked against their own components.
How the figures can be checked
The calculation reads each card’s level, prestige, and colour-by-colour cost from the same card array used by the Classic engine. It groups by level, adds each cost vector, and reports the minimum, maximum, and arithmetic mean. No simulated matches, user records, or hidden analytics feed into the table. The build recalculates it when the card data changes.
To reproduce one row manually, add the printed costs of every card in that level and divide by its card count. To check a single card, add only its required colours; omitted colours mean zero. The two decimal places are presentation rounding, not a claim that a player can collect a fractional token. The dataset’s documented origin is linked below.
Replace printed cost with remaining cost
A useful purchase comparison starts with your bonuses. For colour c, remaining cost is max(0, printed cost[c] minus bonus[c]). Add that result across the five colours. Then subtract the coloured tokens you can actually pay and allocate any gold to the remaining gaps. Extra bonuses in one colour cannot pay a shortage in another.
In a constructed example, a card costs three blue and two white. With two blue bonuses and one white bonus, its remaining cost is one blue and one white. Holding one blue and one gold makes it affordable. The printed cost of five is still correct, but it is a poor description of what this player must spend now. Another player with different bonuses faces a different purchase.
Token cost and action cost answer different questions
Suppose two targets each have a remaining cost of three tokens. One needs one each of blue, red, and green; the other needs three blue. From an empty hand, the first can potentially be funded by one ordinary collection when those colours are available. The second normally needs more than one collection because the same-colour action takes only two and requires at least four in the supply.
Both targets therefore have the same remaining token cost but can require different numbers of actions. Scarcity can reverse the comparison again. Existing tokens, available gold, reservations, and opponents’ purchases all affect the action count. This is why dividing prestige by printed cost produces an interesting description of a card, not a complete strategic ranking.
Give the new bonus a realistic horizon
A low-prestige card may accelerate several later purchases. Estimate that benefit using two or three actual visible candidates rather than every theoretical future card in the deck. For each candidate, ask whether the new bonus reduces a requirement that is still positive. Summing those reductions gives a scenario-based saving, not an assured return.
Then count the purchase action used to obtain that bonus. A discount that saves two tokens can be excellent if it removes an entire collection turn. The same saving is less exciting if both plans still require the same collection and purchase actions. Near the end of the game, a bonus with no remaining opportunity to be used has little development value, although it may still complete a noble requirement.
Read the market alongside the summary
Higher-level averages can help you understand broad differences in the deck, but you buy individual cards. Start from the market and your reserved cards. Compare two realistic targets with their remaining costs, missing colours, estimated actions, prestige, bonus colour, and whether an opponent can remove the opportunity. Write these values in a small personal table if the choices feel close.
For example, a two-point target requiring one collection and one purchase may be faster than a four-point target requiring three collections and a purchase. That does not make the two-point card universally better: the larger card could complete your finish, or the smaller card could strengthen a rival’s market after replacement. Keep the assumptions attached to the comparison.
What these numbers do not prove
This page contains no opening win rates, best-player claims, or predictions about shuffled order. It does not infer what is in an opponent’s hidden reserve or assume that the next reveal will help you. The deck totals are deterministic; strategy conclusions are conditional on a visible position and the remaining time in the game.
Use the table to learn the range of investments and the worked examples to avoid confusing price with tempo. After a game, revisit one purchase using the bonuses and supply that existed at that moment. If a calculation seems wrong, report the card, level, printed colours, and article section so that the implementation data and the explanation can be checked separately.
Rules reference
Publisher’s official rulebook (PDF) ↗Analysis and worked examples by Gem Merchant. Constructed positions are not recorded matches or win-rate claims. The publisher reference supports the rules context.