Compete as a team, never lose alone
15 ways to rank and play, from MathTug-style tug-of-war to boss battles. Tap a card to play it on a phone. Each one says how the loser feels, because that decides whether a shy adult comes back.
What the research says
Three findings decide the design. Sources are in research/findings-vong7c-ranks-and-games.md.
One big leaderboard
A 16-week class study found badges plus a leaderboard lowered motivation, satisfaction and exam scores (Hanus & Fox 2015). Leaderboards mostly motivate whoever is already near the top.
Teams compete, members help
Cooperation beat competition across 148 studies (Roseth 2008). Groups competing while members cooperate gave the most enjoyment in 3 experiments (Tauer & Harackiewicz).
Drop the "31.3%" number
The "31.3% felt worse from leaderboards" figure used in earlier decks only appears on one consultant's blog. The likely source paper doesn't contain it. The rule stays; the number goes.
Recommended combos
Every item here is on the menu below. Tap to open.
All 15 options
Grouped by who you play against. Filters stack.
Your peer review idea, checked
Your flow: the learner does the task, a peer guesses the mistake, the AI reveals, the peer gives praise. Research backs the order, with five guard rails.
Giving feedback teaches the giver
L2 writers trained to give feedback improved their own writing more than those trained to receive it (Lundstrom & Baker 2009). Tested on writing only, not speaking yet.
Guess first, then see the answer
Guessing before the answer improves later memory, even when the guess is wrong (Kornell, Hays & Bjork 2009). Never tried inside peer review of speaking.
The AI mishears Vietnamese accents
Speech recognition does worst on Vietnamese-accented English (Graham & Roll 2024). If the AI mishears, it "proves" a correct reviewer wrong. So: show the AI's transcript, allow "AI misheard", never take points on low-confidence items.
Teach-the-AI, checked
Learning by teaching works mostly because it forces recall. Four rules keep that benefit.