CASE STUDY 02 / RESOURCE & ENGINE BUILDING

Splendor.

Does a points-first approach work at every player count?

90 + 10development cards & nobles
2–4players · classic base game
337,200scheduled games
V1.0report · 4 October 2026

The question.

Splendor rewards both building a resource engine and buying points quickly. We compared nine player configurations, examined how preferences and planning affected performance, and re-tested two fixed hints.

A strategy can depend on the table.

The points-first approach had an advantage in some four-player combinations, but was not a winning approach at every player count. Against the same higher-planning accumulation opponent, the results changed between two and four players.

POINTS-FIRST VS. ACCUMULATION

Win share in the tested match-up

2 players42.2%

Equal-chance reference: 50%

4 players33.0%

Equal-chance reference: 25%

Shared wins are split between tied winners. The marker is the equal-chance reference at that player count. These results refer to this opponent combination.

What the re-tests changed.

01

Retain the base rules

The results did not support changing development-card values, noble requirements, or the 15-point ending threshold.

02

Do not adopt the two hints yet

Each hint was scheduled for 16,200 matched games. Neither met the agreed improvement target, so neither is recommended as a general winning rule.

03

Handle exceptional digital states

Some tested states left all players without a legal action. A digital version needs an agreed way to identify and handle such states.

Scope and results.

Original-version games304,800
Re-tests for two hints16,200 × 2
Total scheduled337,200
Normally completed337,035
Set aside for investigation165

141 of the 165 cases were confirmed as no-action states; 24 concerned progression in the player implementation. They were reviewed separately from ordinary game results.

Recommended next steps.

01

Keep the game’s core

Retain standard card values, resource quantities, and scoring. Use the match-up results to improve strategy examples.

02

Agree the exceptional-state handling

Confirm the intended ruling for no-action states, then add it to a digital implementation and test the resulting flow.

03

Test advice before publishing it

Compare new guidance under matching conditions and check its performance against different opponents.

Independent study · Classic base game · Report V1.0.

Request the sample report

WORK WITH ZONE IN

Tell us what you want to test.

Your game’s stage, player count, and main question are enough to start a conversation.

Discuss your project