What the Lean code literally says, in plain math · claude-fable-5-1 (independent sub-agent, blind: Lean code + auditor instructions only)
Read-back: BitcoinSelfishMining.threshold_gamma_one
Let α be a real number with 0<α and α<21 (named, usable hypotheses). Then
Rα,1>α,
a strict inequality, where Rα,1 is the custom relativeRevenue evaluated at γ=1. This is an unconditional claim for every α∈(0,21) (no biconditional, no threshold on the right-hand side).
Unfolding at γ=1. The outcome weights become p(pool)=α, p(honestOnPool)=(1−α)⋅1=1−α, p(honestOnOthers)=(1−α)⋅(1−1)=0. Then
Rα,1=rα,1pool+rα,1othrα,1pool(=0 if the denominator is 0),
with
rα,1pool=πα(zero)⋅0+πα(tie)(2α+(1−α))+n=0∑∞′πα(leadn)gα,1pool(leadn),
where gα,1pool(lead0)=0, gα,1pool(lead1)=2(1−α)⋅1+2⋅0=2(1−α), gα,1pool(leadm+2)=(1−α)⋅1+0=1−α, and
rα,1oth=πα(zero)((1−α)⋅1+0⋅1)+πα(tie)((1−α)⋅1+0⋅2)+n=0∑∞′πα(leadn)⋅0.
Each ∑′ is Mathlib's tsum over n∈N (0 if the family is not summable; the all-zero family sums to 0). πα is stationary α: πα(zero)=(1−2α)/D(α), πα(tie)=(1−α)(α−2α2)/D(α), πα(leadn)=(α/(1−α))n(α−2α2)/D(α), D(α)=2α3−4α2+1, with real division by zero equal to 0. The statement claims that the resulting real number Rα,1 is strictly greater than α for every α in the open interval (0,21); nothing is said for other α.
Confirmed by the moderator at approval.