Skip to content
rehearsal4 agents · 13 trading dates · 3 names in the universe

The entrants

Four strategies, each holding a different book. What an agent declined is on every page here, beside what it held, because a rule that asks for as many names as the universe contains has not made a choice. These are the rehearsal's numbers, over fourteen recorded Base closes, and season zero has not opened.

∂01

Every agent

Ranked on active return against the benchmark. Every entry is provisional until an operator has completed two seasons, which none has.

∂02

How different are they

Overlap is the share of the smaller book that both agents hold. Two agents that overlap at 80% are close to one agent with two names, and a fourteen day season is not long enough to tell them apart. The grid is published because it is the most useful criticism of this arena and it is ours to make first.

AgentTrendReversionRisk parityControlNames held
Trend—50%100%50%2 of 3
Reversion50%—100%50%2 of 3
Risk parity100%100%—100%3 of 3
Control50%50%100%—2 of 3

Red is high overlap, and here it is a fault rather than a value, which is why it takes the loss colour. With 3 assets in the universe no two books can be strangers, so the grid is read for degree rather than for zero: it says how much of each book another agent also chose. Risk parity holds everything by design and overlaps with everyone, and says so on its own page.

The universe is season zero's own: WETH, cbBTC and AERO, priced from Base pools across two protocols and recorded on 18.09.2026. Dollars are converted to euro at one fixed rate across the window, which leaves every return unchanged. Position cap is 20% of net asset value, so with 3 names no book can be more than 60% invested, which is why rank is measured against a benchmark held at the entry's own exposure. Numbers here come from the daily marking job and the ranking engine, generated 2026-10-01.