Algorithm's Choice

At first glance, being a recommender system looks glamorous. Millions of items at your fingertips, users eagerly waiting for suggestions, conversions rolling in like clockwork. Easy job, right? Wrong.

Most of the time, you barely know your users. A click here, a half-watched video there, a shopping cart abandoned at midnight. That’s all the data you get to work with—and somehow you’re expected to read minds from it.

Meanwhile, users keep changing their tastes, new constraints appear out of nowhere—“Only recommend the freshest items! But also from this narrow, obscure category!”—and to top it all off, rival recommender systems are challenging you in relentless A/B tests. One small improvement and they immediately steal the spotlight.

Now imagine stepping into those shoes yourself.

Components

• 48 Playing Cards (per set)
• 12 Item Cards (per set)

Playing Cards have two sides:

• Action side – actions (changes and items)
• Feature side – user/editorial requirements

Item Cards have two sides:

• User side – active requirement (goes in the middle of the table)
• Item side – item for recommendation

💡 Playing Cards act as both actions (transformations) and features (rules to be satisfied). Two sets can be combined for up to 6 players.

Goal of the Game

Be the first player to recommend all your items to active users by building valid recommendation chains.

Setup

Next to each User, place 1 Playing Card feature side up.

Course of the Game

The player most recently on social media or in an online shop starts. Play proceeds clockwise.

Each turn has three steps:

1. Play Card(s)
   • Place 1–2 Playing Cards at the end of a User’s recommendation line.
   
   • Chains must follow these rules:
     • No two Items can be adjacent.
     
     • No more than two consecutive Actions.
     
2. Pitch an Item
   • Choose one of your Item Cards and attempt to recommend it by placing it at the end of the chain.
   • Draw 2 new Feature Cards and place them next to the User.
3. If successful:
   • Chain satisfies all active features.
   • Flip your Item Card.
   • Replace the User with this Item Card (User side up).
   • Discard the entire chain.
   • Choose 1 of the 2 new features to remain; discard the other.

4. If unsuccessful:
   • Take your Item back.
   • Choose 1 of the 2 new features to replace the old one.
   • Chain remains in play.

5. Redraw
   • Draw until you have 5 Playing Cards.
   • If the draw pile runs out, shuffle the discard pile.

End of the Game

The game ends when a player has recommended all their items. That player is the Master Recommender.

Items

Each Item has three properties:

Actions

Each Action corresponds to one property:

User Features

There are 12 possible features:

• Colour Features
  • Chain must include a given colour (including pitched item).
  
  • Chain must not include a given colour (including pitched item).
    
  
• Length Features
  • Chain must contain at least 8 cards (including pitched item).
  
  • Chain must contain at most 7 cards (including pitched item).
    
  
• Serendipity Features
  • Chain must include at least one pair of consecutive Changes.
  
  • Chain cannot include consecutive Changes.
    
  

Sets

• One Set: Up to 3 players (or 4 in short variant)
• Two Sets: Up to 6 players
• Sets are marked with stripes in the corners for easy distinction.

Variants

• Team Play – players form teams; all recommendations count for the team.
• Short Variant – each player starts with 2 Items instead of 3.
• Cooperative Mode – User features are not discarded after failed pitches; all items must be recommended before the deck runs out. If the deck empties twice, everyone loses.