RAI Dollar · Control Meets Game Theory
RAI Dollar's peg controller is public bytecode: anyone can run
nextParWithPrice() and see where par is
headed. Below, a share α of the market does exactly that — projecting the
controller forward and trading against its own future. Drag the slider and watch anticipation do
the controller's work for it.
The controller is the same 1:1 port of the RateParControl contract
used by the Baton Pass simulator —
deadbands, sign-asymmetric gains (rate ceiling 50% APY), the 24-hour bound-entry dwell plus
seven-day sustained-error fallback, slew limits ($0.001/hr on par),
integrator leak, and output clamps — reading a 24h TWAP. The market is the same agent model
(redemption arbitrageurs with throughput limits, walk-away mint arbitrage above par × MCR at the lowest deployed MCR, 110%,
carry traders whose demand collapses with impaired backing, panic/demand flows, noise), except the
momentum-chasing front-runner is replaced by model-based anticipators: every 4 hours a share
α of capital re-runs the controller's own preview over the lookahead window (10 days by default) at the current price — on-chain
this is literally calling nextParWithPrice() in a loop — then bids up to the projected
redemption value (discounted at 20%/yr over the projected re-peg time, capped at $1) and refuses to hold
above the projected walk-away ceiling. Beliefs are dispersed by ±0.3¢ so flows turn on smoothly.
Re-projecting against a price that its own trading just moved is the fixed-point loop the article
describes — watch the projected ceiling hunt when α is high. Numbers are illustrative; the control
behavior is the contract's.