Splendor Strategy: When to Build and When to Score

Start with a turn budget

At the beginning of your turn, choose one visible scoring target and count the actions still needed to acquire it. Include collecting tokens, reserving if necessary, and the purchase itself. Then choose a second target that uses some of the same resources. A plan with a fallback is more robust than a beautiful sequence that depends on one card surviving several opponents.

This count is a working estimate, not a prediction. Another player can buy your card, a colour can become scarce, and a new market card can create a shorter route. Recalculate after each market change. The useful question is not how powerful your collection looks, but which realistic purchase it brings closer on your next turn.

A discount has value only where you can use it

A red bonus removes one red requirement from every later purchase that still needs red. It does not give a red token, reduce a different colour, or accumulate unused savings. If your next three likely targets each require at least one red after current bonuses, a new red bonus could save three tokens across those purchases. If none requires red, its immediate saving is zero.

Buying a free card still consumes a purchase action. Suppose you can already take a scoring card this turn. A free bonus card followed by that scoring card takes two turns unless the bonus changes another step. The bonus might still help with a noble or a later purchase, but it is not automatically a faster play simply because its token cost is zero.

Worked choice: one discount versus an immediate purchase

Imagine a target with a remaining cost of two blue and one white after applying your existing bonuses. You hold one blue and one white. Taking a blue and buying on the following turn uses two actions. If you can instead buy a free blue-bonus card now, the target becomes affordable next turn: that route also uses two actions, provided both cards remain available.

In this position the free bonus is attractive because it develops your collection without delaying the target. Now change the position: you already hold two blue and one white. The target can be bought immediately, so taking the free bonus first delays the score by one turn. The same bonus card changes value because the missing-resource calculation changed.

Finally check the opponent. If they can buy the target next, neither two-turn route is secure. Reserving can protect the target, but it adds its own action and may be impossible with three occupied reserve slots. These comparisons assume ordinary Classic cards; abilities from expansion modules can change the accounting.

Price a reservation in actions, not excitement

A reservation can secure a contested card and usually provides a gold token if one remains. Separate those benefits. If the card is not contested and ordinary gem collection gives the resources you need, reserving may be slower. If a single inaccessible colour blocks a purchase, gold can solve a genuine supply problem.

Before reserving, name the next action that moves the card toward purchase. A hand full of distant targets removes future flexibility: reserved cards cannot simply be discarded, and three is the limit. Reserving blindly from a deck hides the card from opponents but also trades a known opportunity for an uncertain one. Use that uncertainty deliberately rather than calling it a guaranteed improvement.

Treat a noble as a shared destination

For each face-up noble, count the missing bonuses by colour. Tokens do not satisfy those requirements. Two nobles that share colours can give your development purchases several possible payoffs, but the requirements must be completed for an actual available noble. Do not add hypothetical noble points to your score before the visit happens.

A noble is worth considering when its missing bonuses also improve purchases you want anyway. If the last required bonus costs several preparation turns, compare that complete route with a scoring card. Opponents see the same noble and can finish first. At most one noble visits you per turn, so satisfying two simultaneously does not immediately award both.

Change the calculation near the finish

In Classic, reaching fifteen prestige triggers the end of the round; it does not instantly end every player’s opportunity to act. Track the first player and who still has a turn in that round. A purchase that reaches fifteen can expose you to a higher final score from someone who acts later. A blocking reservation may matter more than a small engine improvement.

Once you estimate that you have only one or two actions left, evaluate bonuses against those remaining actions. A discount that would repay itself across five future purchases has little practical value if those purchases will never occur. Prefer a concrete final score calculation to the vague hope that a larger collection will eventually become stronger.

A short review after each game

Write down one turning point: the market target, your bonuses and tokens, the alternative action, and what the next opponent could do. Count the actions for both plans using only information visible at the time. Do not judge a decision solely by a favourable replacement card that nobody could have predicted.

Practise one question at a time against bots: did a free purchase delay scoring, did a reservation solve scarcity, or did a noble route have a realistic finish? Bots make a convenient practice table, but their programmed preferences are not evidence of what all human opponents will do. The goal is a repeatable decision method, not a universal opening sequence.