The six characters 'AI游戏经济学家' first appeared on my timeline, and the first thought that popped into my mind was: here comes another project party creating a new term. Having spent too long in the crypto circle, I naturally have an immunity to such beautifully packaged concepts. But later, I went back to review the early interview records of the Pixels team, the statements of founder Luke in podcasts, and the evolution of Stacked from internal tools to an independent product, and I found that this thing was indeed not just a marketing gimmick thought up on a whim. It is a hard shell that has grown out of a wound after being repeatedly slapped by reality.
We need to rewind time to 2022. Back then, Pixels was still very small, and the team could count on their fingers, but they had already hit the classic dilemma of the blockchain gaming track: giving rewards brought a swarm of bots; not giving rewards made real players leave. If you give too much, witch attacks will leave your economic system riddled with holes; if you give too little, real players feel that the game is stingy and boring. The most frustrating part is that you can hardly tell if the person on the other side of the screen is a human or a script, let alone conduct targeted operations.
Luke later said in an interview something that stuck with me for a long time. He mentioned that the team initially thought the hardest part of making a blockchain game was designing the token economics model—setting the inflation rate, drafting the release curve, single currency or dual currency. These things could be debated for three days and nights in the conference room. But once the game was up and running, they found that the trickier questions lay even earlier: Who exactly should the rewards go to? How much should be given? When should they be distributed? If you can't answer these three questions, no matter how sophisticated the economic model is, it's just a castle in the sand.
Initially, Pixels used the most primitive methods—setting fixed rules, such as how many tokens to give for completing a task or providing extra rewards for consecutive logins. This system quickly broke down because players' intelligence will always outsmart you, and bots are always faster. Once rules are set in stone, they get reverse-engineered for optimal solutions, and then the army of scripts comes in, treating your reward pool like an ATM.
The team began to experiment with user segmentation. They categorized players into active users, light users, and churn-edge users, offering different incentive plans for each level. There were indeed some effects, but the issue lay in efficiency—manually adjusting parameters was too slow. From data analysis to planning and then to development and launch, an event took at least three to five days. By the time you launched, player behavior patterns may have changed, and your plan went from precise to outdated.
This stage was probably the most painful time for Pixels. The user base was growing, especially after migrating to the Ronin chain, and daily active users began to improve. However, distributing rewards remained a black hole; you never knew how many of the hundred tokens sent out ended up in real hands versus how many were fed to bots. Luke made a controversial decision at the time: to pause the development of new features and concentrate all efforts on creating an internal reward system. No new crops, no new pets, nothing visible to players. Just the tedious data tables, rule engines, and attribution models in the backend.
This decision being questioned is normal. Players want new content, and investors want to see product iterations. You tell them that for the next few months, you will be working on a backend system that players cannot see, and anyone would feel anxious. But Luke insisted because he had done the math privately: traditional game companies spend tens of billions of dollars annually on user acquisition, with half of the purchased users being fake or uninstalling after claiming new player rewards. If you could directly distribute this budget to actual players and accurately measure how much each dollar in rewards improved retention and generated real revenue, this system's value would far exceed any single game feature.
This logic doesn't seem fresh when looking at it today, but back in 2022 in the blockchain gaming circle, there were only a handful of teams that dared to think this way and invest real money into it.
The prototype of Stacked emerged from this decision. It is not a grand blueprint from some white paper but a survival tool that Pixels forged through trial and error with its own flesh and blood. The team spent nearly a year systematizing anti-bot identification, user behavior tracking, reward attribution analysis, and A/B testing frameworks. By the second half of 2023, this system had processed over a hundred million reward distributions within Pixels, enduring several large-scale bot attacks, while also accumulating behavioral data from millions of players.
At this point, Luke felt that the timing was about right. This system could stand alone and have a formal name. Thus, Stacked was born, along with the enigmatic 'AI game economist' that many found puzzling. But to clarify, this AI is not some sci-fi movie's Skynet; it is a decision model fed by vast amounts of real player data. Every day, it scrutinizes Pixels' data flow, repeatedly asking itself three questions: Which player is about to churn? What rewards can retain them? Is the cost of this reward worth it?
The first two questions are ones many game operators are asking. The real differentiator is the third question. Stacked calculates a value return ratio—how much real retention time or actual payment behavior can be returned for every dollar spent on rewards. If the expected return of a certain event plan falls below a set threshold, the AI automatically lowers the reward amount or even shuts down the event. This decision-making logic typically requires an operations manager in traditional game companies to brainstorm and hold three or four meetings to finalize, but in Stacked, it is automated.
Data does not lie. As of now, Pixels' user base has surged to a million daily active users, and Stacked has processed over 200 million reward distributions. More importantly, this system has begun generating real revenue—not from selling tokens, but from players' actual spending in the game: purchasing land, participating in events, unlocking premium content. With this precise reward engine, Pixels achieved over $25 million in agreement revenue during the coldest phase of the bear market.
Looking back at the significance of Stacked for Pixels is now very clear. Without Stacked, Pixels would be just an ordinary farming game, propped up by token expectations, collapsing the moment user confidence wavers. With Stacked, Pixels effectively installed an intelligent reward distribution valve, accurately allocating its limited operational budget to retain the most valuable users. Moreover, this system has already been tested internally within Pixels and can be made available for external studios to integrate. This means Pixels' positioning is beginning to change—it is no longer just a game project, but is transforming into a provider of reward infrastructure.
This is also why I started to reevaluate $PIXEL. In this circle, I've seen too many blockchain game projects with roadmaps that looked like artworks, community ambassadors shouting loudly, and various partnership announcements flooding in. But if you ask them a few straightforward questions: How many times has your reward system withstood bot attacks? What is the ROI of each dollar in rewards? Most project teams cannot answer. However, Pixels can not only provide answers but also present evidence of over two hundred million reward distributions.
Stacked is not a concept written in the white paper; it is the armor that Pixels developed over four years of struggle and countless setbacks, emerging from the gaps in its bones. This cannot be conveyed merely through storytelling.