Context
GoldCup is the consumer surface of WallGold, a gold trading platform. Buying gold is a thing people do once and then forget for a year, which makes it a terrible habit and a hard business. So we built the habit instead of the transaction: predict World Cup matches, play games, run missions, earn gold. I have been the product manager since the first commit, I design it, and I ship it.
The reframe
Our reward economy was minting keys far faster than players spent them. The obvious read was inflation, and the obvious fix was to close the tap. I pulled the distribution before touching anything, and the median player balance was in single digits. We were printing at scale and the median player had nothing. The mint was not reaching players; it was pooling with the few who had already learned the loop, while most people bounced long before they held enough to buy anything at all. We had not built an inflation problem. We had built a product that paid experts and confused beginners, and inflation was only the shadow it cast. Capping the faucet would have made the median worse. So I stopped balancing at the top and built a floor at the bottom. The daily cap stayed, but hitting it stopped being a wall: once a player caps out, the games go free, with no entry and no reward, XP only. The cap now protects the economy from the few big earners instead of ejecting everyone else at the exact moment they were enjoying themselves. And the luck engine underneath is tuned a few points player-positive, because a game that is fair on paper feels rigged in the hand.
What I built
Six surfaces, shipped over a season. A games hub that replaced the prediction-only tab, holding an arcade of skill games and a territory game with real-time matchmaking against live players and a bot fallback when nobody is waiting. A mission map built as a dependency tree across six chapters, timed to the real World Cup stages. A per-game XP system that quietly raises difficulty as you get better, decoupled from the reward economy so it keeps giving you something after the daily cap fills. A weekly competition with a live leaderboard, public profiles, and a prize-collection flow. A server-authoritative key economy, since the balance is real money's cousin and the client is never trusted with it. And a luck engine underneath all of it: per-user, anti-losing-streak, with a newbie hook, a whale shield, and daily budgets, tuned to sit a few points player-positive rather than house-positive. Behind that: an admin panel so operators can retune every game's economy, art, and odds without me, and instrumentation feeding funnel and engagement dashboards I built to answer questions I was tired of guessing at.
What I killed
I built a Home tab and then deleted it. It was the first screen in the app and the one I had argued hardest for: a place to orient, see your streak, see what was next. It tested as a toll booth. Every session started with a screen people paid attention to and then left, on the way to the thing they had actually opened the app for. I had designed a lobby for a product whose users already knew where they were going. I removed it and made the games the front door.
Outcome
The specific numbers belong to WallGold, so I can only give the direction, not the figures. After the floor shipped, the median balance stopped living in single digits and started to climb. More new players reached a first reward instead of leaving before they held enough to spend. The drop-off at the daily cap softened too, once hitting the ceiling stopped ejecting people and started handing them a free mode to stay in. The absolute figures are real and confidential; the direction is what I can stand behind in public.
Where my role stopped
I own product, design, and the economy model, and I ship the front end myself. I do not write the backend that enforces the economy; that is another engineer's code, and the anti-cheat design in it is theirs, not mine. The national TV campaign that drove our biggest week was not my work either. I was handed the traffic and asked to keep it.
What I know now
I used to treat a reward economy as a balancing problem: if the coins minted and the coins spent netted out, the design was sound. That is arithmetic, not product. What GoldCup taught me is that an economy is a legibility problem first. What a new player believes is possible in their first few minutes decides almost everything downstream, and no amount of correct math in aggregate fixes a median of four. I design the economy from the newest, most confused player now, not from the spreadsheet.
A note on numbers
Some figures here are given as direction rather than absolutes, and some are omitted entirely, to respect what is confidential to GoldCup. Everything stated is my own and does not necessarily reflect the views of GoldCup.