Playtest Findings

The most surprising thing about playtesting Watts and Wealth: players who started out going all-in on coal almost always switched to renewables by the middle of the game — and they did it without anyone telling them to. The game's crisis cards gave them a cost they could feel. That single pattern is what the whole design was built around, and it held across every version of the game we tested.

Four things we kept seeing

These four patterns showed up across the paper, Unity, and browser versions of the game. Seeing the same thing in three different implementations makes us more confident they are real features of the design, not accidents of one build:

1. The mid-game flip — rounds 11–14. Players who opened the game committed to coal (cheap, high-output, no immediate penalty) almost always switched to renewables by round 14. Not because someone told them to — because the pollution-event cards started landing in their hand and producing real losses. The board created the argument the lecture can't (Kolb, 1984).
2. Coal-heavy tables hit the collective loss around rounds 22–25. Tables that leaned hard on coal and oil in the early eras (Industrial Dawn and Oil Rush) almost always triggered the Tipping Point ending — where accumulated pollution crosses 600 and everyone loses — in this window. We saw this in both the Unity and browser phases, where our event logs gave us clean numbers to check.
3. Winning the co-op path almost always needed an Activist player. The Renewable Pathway win condition (keeping the table below the pollution threshold together) rarely came together without at least one player taking the Activist archetype. Without that role, the group never quite coordinated the defense in time. This locked in a key design belief: the four archetypes aren't just flavour — they shape what a table can actually pull off.
4. The "whose turn is it?" problem took four rounds of UI iteration to fix. When we moved to the digital version (R9–R12), players kept losing track of whose turn it was, accidentally peeking at another player's hand, or skipping the build phase. We iterated on colour-coded seat indicators, a clear "pass the device" prompt, and an auto-collapsing hand panel until it felt natural (Plass, 2015). By R12 it was no longer the thing people complained about.

Iteration phases

Each card below is one round of iteration. The finding is what that phase actually taught us.

• Tabletop prototype playtests (R1–R8)
  8 sessions · v0.x physical (R1–R8)

  Players gravitated toward coal in rounds 1–3 of every session but switched to renewables by mid-game once pollution-event cards began landing in their hand. Mid-game pivots clustered between rounds 11 and 14.

  Instruments: post-session structured debrief (~15 min) · decision-log per turn (hand-written) · facilitator observation notes
  Limitations: Small N. Convenience sample. No pre/post knowledge test. Single facilitator. Hand-counted balance tweaks between sessions.
• Unity hot-seat port (R9–R16)
  8 sessions · Unity 6 hot-seat (R9–R16)

  Hot-seat handoff UX required multiple iterations to become friction-free. Players reported confusion about seat boundaries in R9–R10; iterative changes to seat indicators, turn-end confirmation, and pass-device prompts eliminated this by R12.

  Instruments: in-game analytics.ts JSONL export · facilitator observation notes · post-session debrief (~15 min) · structured bug-report intake
  Limitations: Same convenience-sample limitations as physical phase. Unity-specific UI bugs occasionally confounded gameplay observations; logged separately from playtest findings.
• Web rewrite iteration (R17–R22)
  6 sessions · Web hot-seat (R17–R22)

  Web rewrite preserved the gameplay shape of the Unity version while reducing setup friction to zero (no install). Mid-game pivot timing remained consistent with R13–R16 (rounds 11–14). Adversarial code review surfaced a class of fact-check bugs that pure self-review missed.

  Instruments: in-game analytics.ts JSONL export · deterministic action-log replay · facilitator notes · adversarial code review (production process)
  Limitations: Small N (six logged sessions in this phase). The web-rewrite phase is also the current phase; findings here are interim, not final.

One design rule that came out of all of this

Around the fifth tabletop session we tried fixing a balance problem by rewriting card text. The next session was a mess — players reasoned about the new text in ways we hadn't expected, and the fix broke something else. We tried it twice more. Same result. So the rule locked: card text is final; balance moves through hidden constants only (deck weights, build costs, pollution rates). Once that rule held, every balance fix became predictable. The card you read is the card you get — that consistency is what lets players build a mental model of the game fast enough to actually feel the strategic stakes.

What we still don't know

• Does the mid-game flip happen outside Asia-Pacific contexts? Our sample skews that way. Cross-cultural sessions are on the backlog.
• Is the Activist dependency a real design feature or a small-sample fluke? A larger group of strangers might solve it differently.
• Does playing the game actually change what people know about energy systems? We tracked behaviour, not knowledge. A before/after test is the natural next step.

Grounded in: Kolb (1984) on learning from experience; Plass, Homer & Kinzer (2015) on game UI as a learning scaffold; Squire (2011) on game-based learning in practice; Clark & Mayer (2016) on evidence for game-based instruction.