2ef3f2e39f
P3 complete. 8 docs/freeholders/ pages (signals, standing, stands-guilds, councils-voice, bonds, partner-spectrum, anchor-preview, index) + 2 docs/reference/ pages (architecture, components). Docs firewall green across all 26 docs pages. Full docs deliverable (README + mkdocs.yml + 26-page site) complete — REQ-027 complete. ---ci--- project: oy phase: 3 milestone: v0.3 status: complete tag_base: v0.2.x phase_role: execution requirements: covered: [REQ-027] partial: [] ---/ci---
49 lines
2.5 KiB
Markdown
49 lines
2.5 KiB
Markdown
# Bayesian Standing
|
|
|
|
**Standing** (REQ-006) is a Holder's measured history on the mesh — the
|
|
anti-gaming metric that gates [Freeholder](index.md) participation and weighs
|
|
[Voice](councils-voice.md) in the [Councils](councils-voice.md). Standing is
|
|
not a count of transactions and not a reputation score you can farm. It is a
|
|
Bayesian score that resists the obvious attacks: volume spam, self-dealing,
|
|
fake vouches.
|
|
|
|
## The formula, at conceptual depth
|
|
|
|
Standing combines four signals and a penalty:
|
|
|
|
- **Bayesian prior + updates.** The mesh starts with a prior for each Holder
|
|
and updates it from each observed action. A burst of activity cannot
|
|
inflate Standing because the prior anchors it.
|
|
- **Time-decay.** Old evidence decays, so a Holder cannot rest on a burst
|
|
from years ago. Standing reflects *recent, sustained* real production.
|
|
- **Diversity weighting.** A Holder who acts across many services, many
|
|
[Stands](stands-guilds.md), and many [bearers](../nomads/bearers.md) accrues
|
|
more Standing than one who repeats the same action with the same
|
|
counterparty. Diversity is the anti-collusion lever.
|
|
- **Voucher-weighting.** A vouch from a high-Standing Freeholder counts for
|
|
more than a vouch from a low-Standing one. This makes fake vouches expensive:
|
|
the voucher must themselves have Standing to lose.
|
|
- **Minus slashes.** Bad behavior (failed attestations, broken Pacts) removes
|
|
Standing. Slashes are the penalty that bounds the upside of gaming.
|
|
|
|
> The full sub-tables (priors, decay rates, diversity categories, slash
|
|
> conditions) are deferred per PROJECT.md Q2. This page gives the conceptual
|
|
> depth; the [nomads Standing page](../nomads/standing.md) gives the plain-
|
|
> language version.
|
|
|
|
## Why it cannot be gamed
|
|
|
|
There is no single input a Holder can pump. Volume is bounded by the Bayesian
|
|
prior; recency is bounded by time-decay; breadth is bounded by diversity;
|
|
social trust is bounded by voucher-weighting; and any attempt that misfires
|
|
costs Standing via slashes. The four signals (the [90-day Stash](signals.md),
|
|
3-category threshold, Capital, Vouch) sit on top of this metric, so the
|
|
Freeholder gate inherits the same anti-gaming property.
|
|
|
|
## What Standing is not
|
|
|
|
Standing is not a custodial position, a tier you buy, or legacy history. It
|
|
is a measured, decayed, diversified Bayesian score. See
|
|
[Storage Pools](../shared/storage-pools.md) for why the mesh says "Stash"
|
|
rather than the legacy custodial words, and [Councils & Voice](councils-voice.md)
|
|
for how Standing weights a Freeholder's vote. |