Equalize co-op gift magnitudes to kill 'always leave' dominant strategy

Old values (open ±7, leave ±4) had matching EV but asymmetric variance;
risk-averse parties had a strict preference to always leave gifts, which
collapsed the dilemma. New values (all ±6) keep EV=0 for both strategies
AND equalize variance. Choice now depends entirely on inferring sender
intent.

Adds TestCoopGiftVarianceSymmetric — fails loud if anyone reintroduces
asymmetric magnitudes. EV test was already in place but only checked the
mean; missed this on first pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
prosolis
2026-04-26 19:12:22 -07:00
parent 1ca9e77fa0
commit a1ede53fa9
3 changed files with 40 additions and 15 deletions

View File

@@ -626,12 +626,12 @@ Multi-day party runs separate from solo. 24 players, 27 days depending on
**Spectator betting.** Parimutuel pool, 10% rake. Bets can be placed or increased any time during the run; positions are locked in (success or failure can't be switched). Party members can bet on their own run — sandbagging while heavily bet against is exploitative and visible. Odds line shows estimated success% based on party levels, pets, neutral funding assumption, and a hidden TwinBee Helpfulness Rating (rolling last 30 floor events: how often his recommendation was right). Helpfulness shifts the line ±20% but is never shown directly — players notice the line moves after events resolve.
**Gifts.** Anyone *not* in the run can spend one harvest action to send a Care Basket or a Mimic. The party never sees which type. Vote `open` or `leave`. Modifiers are EV-symmetric at a 50/50 sender mix:
**Gifts.** Anyone *not* in the run can spend one harvest action to send a Care Basket or a Mimic. The party never sees which type. Vote `open` or `leave`. Magnitudes are equalized across all four outcomes — both "always open" and "always leave" yield EV=0 with σ=6, so neither is risk-averse-dominant. The choice is purely about reading sender intent.
| | Open | Leave |
|-------------|------|-------|
| **Care Basket** | +7% (boost) | 4% (basket explodes) |
| **Mimic** | 7% (mimic does what mimics do) | +4% (sad mimic helps anyway) |
| **Care Basket** | +6% (boost) | 6% (basket explodes) |
| **Mimic** | 6% (mimic does what mimics do) | +6% (sad mimic helps anyway) |
Multiple gifts stack additively — coordinated senders can swing odds significantly. The full gift log is public at end of run.