Infinity Pool — Economic Model

Simulation of the emission, staking and secondary-market loop. Generated from model/simulate.py — every assumption is a named constant in that file.

The headline. Across every adoption scenario, aggregate staker profit converges on seeded liquidity minus protocol fees. That is not a tuning problem — it is an accounting identity. The only ETH that can leave this system is ETH that players put into it, so the more successful the launch, the more negative the aggregate player result becomes.

What the model does

Each simulated day: new depositors arrive on a launch-spike-then-decay curve, stake ETH (refundable principal, minus a one-off entry fee), and buy some INF on the open market. Emission mints to stakers pro-rata. Stakers sell a fraction of what they earn back into the same pool the buyers funded. Price is set by a constant-product AMM.

Principal is not counted as a loss — it is withdrawable at any time, by design. The P&L line below is strictly open-market buys versus sale proceeds, which is the part players can actually lose.

INF price

Log scale. Higher adoption buys a higher peak and a steeper fall, not a different ending.

Cumulative aggregate staker P&L

All stakers combined, in ETH. Open-market buys versus sale proceeds.

Scenario outcomes

Scenario	Depositors	Peak price	Drawdown	90% down by	Staker P&L	Protocol

Read the last two columns together. They are close to mirror images. Protocol revenue is not created alongside staker returns — it is drawn from the same pot, which is why scaling adoption moves both numbers in opposite directions by roughly the same magnitude.

Why the shape is fixed

Emission is constant, demand is not. 25,000 INF/day arrives whether or not anyone is buying. Buy-side flow tracks new arrivals, which decay after launch. The two curves diverge within weeks in every run.
Halvings arrive after everything has already happened. The first is at day 730 — beyond the 730-day horizon modelled here entirely. Collapse lands between day 97 and day 402, so no supply cut arrives in time to affect any of it.
Adoption scales both sides. Ten times the users means ten times the buying — and ten times the stakers splitting the same fixed emission, so ten times the sell pressure a few days later.
The cap never binds here. Supply reaches 18.25M of a 50M lifetime cap by day 730. Emission is not the thing that runs out — buyers are. The cap is a bound on the far future, not a constraint on this chart.

What would actually change the outcome

These are the levers that alter the identity rather than the timing. None of them are in the current contracts; all of them are additive.

External revenue. Route real fees from something that earns outside the system into the buy side. This is the only structural fix — everything else redistributes.
Emission tied to inflow rather than the clock. Mint against realised protocol revenue instead of a fixed schedule, so supply cannot outrun demand.
Deep, protocol-owned liquidity. Does not fix the identity, but changes who absorbs the volatility and stops the price gapping on the first large exit.
Lower emission, longer schedule. Pure delay. Worth doing, but it buys months, not a different destination.

Model assumptions:
Change any of them in model/simulate.py and re-run

python3 model/simulate.py &&
  python3 model/build_report.py

. The conclusion is not sensitive to these values — it follows from the absence of external revenue, which no parameter choice repairs.