The Half-life of a Game
“I’ve got 10 forks right here baby!”
From fanatic worship of (3,3) to (1,1), from a peak market cap of 4 billion to now a market cap of just 500mm, from 150 forks to the death of the entire fork population, the birth of Olympus DAO has sparked the imagination and talk of DeFi 2.0, and its death (?) has left everyone wondering if this was it for DeFi.
We do not yet know how Olympus will evolve in the months to come. We do not know if it will survive for the coming years. But in the wake of the explosion of interest and innovation it has sparked, we might be able to draw some lessons.
One of the first things I noticed about DeFi was how similar it was to gaming. Never an avid gamer myself, I still couldn’t help but notice the myriad similarities: the isomorphism of activity in how both were about maximising reward, the nature of competition between players, the overlapping communities, the same analysis skillset. Interviews at both crypto VCs and projects inevitably involve gaming questions - LOL, Warcraft, CIV 6, Left 4 Dead, Fortnite, chess, go, poker, bridge. Gaming ineptitude is a strong predictor of DeFi ineptitude.
And indeed, one of the things that new DeFi players fail to grasp is that these protocols that you yield farm on, are not really investments. The swaps you make are not trades. If an investment is to be understood as one based on fundamental analysis, then the yield farming and the aping that we see, cannot be catagorised as investment, but as speculation, and gambling.
This therefore makes every protocol into a gambling table, where there must always be losers. It is zero sum. Trades are not necessarily zero sum. Investments are not necessarily zero sum. A new protocol popping up is the opening of a new casino. An innovation is a game invented. A fork, is another round of game started.
Games can be solved. A solved game is a game whose outcome (win, lose or draw) can be correctly predicted from any position, assuming that both players play perfectly. A game is solved if and only if optimal strategies for any player are all discovered. Clearly, on a reductive level, solved games are no fun. Tic-tac-toe is a solved game. A game of tic-tac-toe played by two perfectly rational players must result in a draw. If they were betting money on it, then neither player wins - and in fact the whole endeavour would have in fact cost them time - opportunity cost - so net loss. The rational thing to do when someone offers you play a game of tic-tac-toe is therefore to get lost.
Connect Four is a solved game. And it has been mathematically proven to be first-player-win. If the players were betting on it, then the second player would always lose money. Therefore, if you are offered to play a game of Tic-tac-toe as the second player, you should refuse.
While solved games have a stringent mathematical definition in Game Theory land, we intuitively feel that it is the case, with more plays of the same game, we get increasingly better at a game, and we therefore get closer to knowledge of optimal strategies. In simple words, we learn. Evolutionary game theory captures this intuition. Interestingly, the invention and formulation of evolutionary game theory was motivated by the need to explain altruistic behaviours in Darwinian evolution - that was why Olympus tried to memefy the (3,3) strategy, which is the optimal strategy if all players were rational. Alas, humans are not rational.
But given humans are not rational, it would not be exactly accurate to use the mathematical definition of optimal strategies to describe the end state in which players simply stop playing because they know how the gameplay is going to play out. Olympus forks die not because the game has been solved as players have figured out the strategy optimal according to the mathematical structure of the game, but because players have figured out the strategy optimal according to how players actually play the game. That’s fine. Let us call a game “resolved” in such a case.
If we view DeFi protocols as games, then it stands to reason that as more forks appear, people play the game better and better, and the speed in which a game is resolved increases. In other words the time in which a game becomes resolved gets shorter and shorter.
We can see this in the graphs below.
We also this kind of decay behaviour when we look at the major primitive DeFi protocols
Whether a game is resolveable should be considered as a different phenomenon to whether a game has exhibited mania. Mania looks like this:
We can probably call Olympus a Resolved Game, and all its instances definitely exhibited mania, but mania is not necessary to a game being resolvable. A game that exhibited mania is not necessarily a resolveable game. If people just lose interest, and the price crashes, whether the game is resolvable is irrelevant. If I don’t care about the game, I don’t care whether everyone knows how to play and win the game.
However, if a game is resolved, it will definitely exhibit mania.
A game that’s yet not resolved, is likely to exhibit the run-up section of mania - because people don’t know how the game is played and they all feel they might win. Once the game appears resolved, people will want to get out, either because they figured their strategy has not been the optimal one, or that they are losing already. But if a game is unresolvable, players are more like to stay around, simply because they feel they might win. Therefore, an unresolvable game is a game in which the fear and capitulation section of the mania curve is extended outwards. An unresolvable game is a game with an infinite half-life. New iterations of the same game can still be launched, in which newer games do not necessarily have jackpots capped by the jackpots of previous iterations.
Projects have an interest in making their projects unresolvable - because if a game is resolved before their tokens unlock, then the fear and capitulation section of the mania curve will have happened before the team tokens unlock. And given the nature of a resolvable game is that no players will return to the game, what has happened is that the team has built a game with a massive jackpot at one point which they themselves have not enjoyed.
How can a protocol make itself as unresolvable as possible? I can think of three ways to approach it.
Just make it unresolvable: like Chess or Go. This of course assumes unsolvability implies unresolvability. In DeFi protocol design this probably manifests itself as massively increasing the degrees of freedom and the number of possible things you can do with a protocol.
Keep Building Protocols on top - V2 bond and inverse bonds in Olympus.
Make the Game self-evolving - like how new tokens and projects keep evolving on top of loot, Convex on top of Curve or how you have DeFi protocols built on top of Ethereum, which keeps the Ethereum game running.
In practice, one probably needs to do all three, and these tricks to increase unresolvability should be implemented before peak mania, lest you on the dev team become saddled with tokens of a project whose prime has already passed. While post-top implementation might still introduce new interest, projects of course prefer their market caps to plateau over topping and then go downhill.
One of the interesting developments with Treasure DAO is how under the hood, it’s still all DeFi, but it looks and is marketed as a game. As outlined in its 43 page whitepaper, it has a high number of operations in the protocol with many plausible strategies for players. This seems like a model in which future DeFi protocols might look like. The desire to avoid resolvability, combined with a desire to keep value capture away from TradFi (which also slows down resolvability), it is plausible that we will see more and more protocols that look like this in the future.
