Game On: DeHub Launches Prediction and Raffle Games
By DeHub Team — 2021-10-01
In Q4 2021 we released our first prediction and raffle games. They were the first DeHub products where the honesty of the result was the product.
Two different problems
Predictions and raffles look similar from the outside and are not.
A prediction market needs a settlement source everyone agrees on in advance, and a rule for what happens when that source is ambiguous, late or wrong. Most disputes in prediction products are not about the outcome — they are about whether the question was well-formed.
A raffle needs randomness nobody can influence, including us. The entire value of a raffle collapses the moment the operator could plausibly have chosen the winner. Not "did choose" — could have.
Why they came first
They were small enough to build properly and severe enough to force the standard. A raffle where the draw is not verifiable is worthless whatever else it does well, so there was no version of shipping this that let us defer the hard part.
They also fit the app we actually had. The beta had proven daily habit, the community was past 8,000 members by the end of that quarter, and games are the shortest path from "I opened the app" to "something happened".
How we handled the draw
The commitment we made was that mechanics would be provably fair and verifiable on-chain rather than asserted. In practice that means the source of randomness and the rules of settlement are fixed and published before entries open, and the result can be checked afterwards by anyone, against data they did not get from us.
It is the same principle as the snapshot published for the v1 airdrop earlier that year: state the rule first, apply it mechanically, publish enough that being trusted is optional.
Where it led
This work scaled up considerably. The arcade launched in Q1 2022 with five games and predictions alongside, on-chain leaderboards and rewards followed in the same quarter, and by Q3 2022 the raffle mechanics were carrying a $1,000,000 house with more than 15,000 participants.
That escalation is only possible because of what was decided here. A raffle for a games prize and a raffle for a million-pound property are the same code and the same trust problem. Getting it right while the stakes were small is the reason the large one was possible at all.
The unglamorous part
Prediction settlement gave us more trouble than the draws did, and it is the part nobody writes about.
A raffle has one moment of contention and a clean rule. A prediction market has a question, and questions are ambiguous in ways their authors never anticipate — an event that half-happens, a source that revises its own figure after publication, a resolution time that lands in a different timezone than someone assumed.
We learned to spend more time writing the question than building the market. Almost every dispute we have had in a prediction product traces back to wording, not to code, and no amount of on-chain verification rescues a badly specified question.