protocols/STEP0-protection-gap.json · v1

Step 0 · Launch-guard protection gap on public launches

Between the slot at which external buy flow on a new launch first reaches a threshold and the slot of the creator's first exit, a material amount of external flow still enters. If the gap is ~0 slots everywhere, a faster guard has nothing to win and a guard-speed claim is not worth making.

draft

5 values must be agreed before this can run

$.parameters.thresholds_lamports · $.parameters.observation_window_slots · $.parameters.max_signature_pages_per_mint · $.parameters.mint_selection · $.parameters.launchpad

Check with python3 -m rbench.prereg check protocols/STEP0-protection-gap.json.

Design

Why it matters: Sizes the window any faster launch guard could close, whoever runs it. Useful to any launch platform whether or not it uses Rout.
Unit: One launch (mint) from a published mint list.
Primary metric: gap_slots: exit_slot - cross_slot for launches where the creator exits after the threshold is crossed (rbench.step0).
Denominator: Every mint in the published input list, including excluded (truncated/no_history) ones.
Randomization: Mint list fixed and hashed before extraction (none)
Stopping rule: Single analysis after extraction of the whole pre-registered list.

Arms

Arm	What it does
observed	Observation only; no transactions sent.

Secondary metrics

Metric	Definition
`entered_after_lamports`	External buy lamports in slots cross_slot < s <= exit_slot.
`share_crossed`	Share of all input launches whose external flow reached T within the window (Wilson CI).
`outcome_mix`	Counts of exit_after_cross, crossed_no_exit_in_window, exit_before_cross, never_crossed, excluded.

Exclusions, decided before the run

Mints whose history exceeds max_signature_pages_per_mint are labelled truncated and counted, not dropped.

Acceptance

Criterion	Agreed value

Descriptive study: no pass/fail.

What each tier hides

Signatures, slots and ids are never redacted: they are how anyone re-checks a result. A run can't be published wider than the tier it was registered at.

Partner-confidential removes: nothing
Public also removes: nothing

Waits on

APPROVE-STEP0

The file

protocols/STEP0-protection-gap.json

{
  "protocol_id": "STEP0-protection-gap",
  "version": 1,
  "experiment": "STEP0",
  "status": "draft",
  "title": "Step 0 · Launch-guard protection gap on public launches",
  "hypothesis": "Between the slot at which external buy flow on a new launch first reaches a threshold and the slot of the creator's first exit, a material amount of external flow still enters. If the gap is ~0 slots everywhere, a faster guard has nothing to win and a guard-speed claim is not worth making.",
  "customer_value": "Sizes the window any faster launch guard could close, whoever runs it. Useful to any launch platform whether or not it uses Rout.",
  "unit": "One launch (mint) from a published mint list.",
  "arms": [
    {
      "arm_id": "observed",
      "description": "Observation only; no transactions sent."
    }
  ],
  "primary_metric": {
    "id": "gap_slots",
    "definition": "exit_slot - cross_slot for launches where the creator exits after the threshold is crossed (rbench.step0)."
  },
  "secondary_metrics": [
    {
      "id": "entered_after_lamports",
      "definition": "External buy lamports in slots cross_slot < s <= exit_slot."
    },
    {
      "id": "share_crossed",
      "definition": "Share of all input launches whose external flow reached T within the window (Wilson CI)."
    },
    {
      "id": "outcome_mix",
      "definition": "Counts of exit_after_cross, crossed_no_exit_in_window, exit_before_cross, never_crossed, excluded."
    }
  ],
  "parameters": {
    "confidence": 0.95,
    "quantiles": [
      0.5,
      0.9,
      0.95,
      0.99
    ],
    "bootstrap_resamples": 0,
    "thresholds_lamports": null,
    "observation_window_slots": null,
    "max_signature_pages_per_mint": null,
    "flow_mode": "gross_buys",
    "mint_selection": null,
    "launchpad": null,
    "block0_policy": "report_both"
  },
  "randomization": {
    "unit": "none (observational)",
    "method_id": "none",
    "method": "Mint list fixed and hashed before extraction",
    "strata": []
  },
  "denominator": "Every mint in the published input list, including excluded (truncated/no_history) ones.",
  "exclusions": [
    "Mints whose history exceeds max_signature_pages_per_mint are labelled truncated and counted, not dropped."
  ],
  "stopping_rule": "Single analysis after extraction of the whole pre-registered list.",
  "acceptance": {},
  "redact_by_tier": {
    "partner_confidential": [],
    "public": []
  },
  "requires": [
    "APPROVE-STEP0"
  ]
}

Search