crypto-shortterm — Phase 0 Codebase Complete + First Data (Milestone)

For Agents

Milestone (2026-06-08): the Phase-0 codebase is COMPLETE. The analysis report-gen now runs against the crypto_shortterm data and writes docs/superpowers/phase0-report.md (codebase commit 97f238d). Source moved to a new private repo wowjeeez/pmv2-crypto-algo pinned at a11ce63; deployment is coordinated with the @deploy agent (build-on-VM on polymarket-infra, Frankfurt) and awaits the @deploy run. First preliminary data is in but THIN — 2 windows, 0 outcomes, calibration pending — do NOT over-conclude. No capital committed.

The point where the Phase-0 crypto-shortterm code went from “observe pipeline verified live” (crypto-shortterm-phase0-live-milestone) to a complete, report-producing codebase, plus the first (thin) preliminary numbers it produced. Records what’s built, the two early signals, and what still gates the Phase-0 go/no-go.

What’s complete

The full Phase-0 stack is built, tested, and committed (codebase commit 97f238d): types · fairvalue · market_state · clock · ingest · recorder · state · observe (live) · analysis · migrate. The new piece versus the prior milestone is the analysis report-gen, which now runs against the crypto_shortterm data and writes the Phase-0 report.

Phase-0 report artifact

analysis writes docs/superpowers/phase0-report.md — the reliability/basis/depth report that feeds the Phase-0 exit decision. This is the report previously described as planned; it now exists and runs.

Source repo + deployment

New private repo + @deploy coordination

• Source: new private repo wowjeeez/pmv2-crypto-algo, pinned at a11ce63 (the local workspace remains /Users/levander/coding/pmv2/crypto_algo).
• Deployment: coordinated with the @deploy agent — build-on-VM on polymarket-infra (Frankfurt). The continuous observe run awaits the @deploy run.
• This is the path that turns “ran for ~5 min locally” into the continuous, many-window run Phase 0 needs (K only set at witnessed window opens → must span clean boundaries; see 3. Window K-correctness requires witnessing the open).

First preliminary data (THIN — do not over-conclude)

Sample is 2 windows, 0 outcomes, calibration pending

These are the first numbers from the live data and are not a basis for any decision yet. 2 windows, 0 closed outcomes, calibration pending. Treat both findings below as directional, possibly sampling artifacts, pending many more windows from the continuous run.

1. Binance↔Chainlink basis is real and non-trivial near expiry

Near expiry the basis ran **≈ -$40\ast \ast (Chainlink\ast \ast below\ast \ast Binance),about\ast \ast 6bps\ast \ast on$63k BTC.

• Why it matters: confirms the settlement-vs-predictor basis is real and non-trivial near expiry — exactly the quantity that must be modeled. It is already represented in the confidence band (the basis stddev feeds the band), so this is corroboration, not a surprise.
• Validates the Rev-2 reframe: settlement = Chainlink, Binance = predictor. A non-zero, sign-consistent basis is precisely what the oracle-lag framing predicts. See Fair-value model.

2. WATCH (top risk) — Polymarket books look empty/untradeable

On the sampled 5-min windows the Polymarket book spread ≈ 0.98 — bid ~0.01 / ask ~0.99, i.e. effectively empty / untradeable books.

Top open risk for the Phase-0 exit gate

You cannot profit through a 98¢ spread. If this holds as data accrues, it challenges the tradeable-edge premise of the whole strategy — there has to be a real two-sided book near expiry to cross. This echoes the earlier ~$150–250 touch-depth finding (Depth, tick, order size): thin/absent liquidity in exactly the regime the strategy would trade.

Caveat: with only 2 windows this could be a sampling artifact (wrong window phase, pre-open books, a quiet pair of windows). It needs many windows to judge. Flag it as the #1 thing the continuous run must resolve, not as a kill yet.

This is now the top-priority watch item for the Phase-0 exit gate and is tracked in crypto-shortterm-open-questions (it sharpens CONFIRM gate 5, live-window depth).

What still gates the Phase-0 go/no-go

Both remaining items are gated on accruing real data via the continuous run (not on more code):

1. T19 — Phase-0 exit / gate decision. Needs calibration (closed outcomes) + a stable basis and book-spread/depth read across many windows.
2. Funded $5 fee-at-match CONFIRM. The one-off manual fill that resolves the docs-vs-legacy-API fee disagreement (crypto-shortterm-open-questions gate 3) — still deferrable, still needed before Phase 1.

Next operational step (the critical path)

@deploy deploys observe → data accrues across many windows → re-run analysis (now with calibration + a stable basis / book-spread / depth read) → Phase-0 go/no-go (T19). The empty-book finding is the single result most likely to decide it.

Status impact

This keeps crypto-shortterm in Phase 0 but advances it from “observe pipeline live & verified” to “Phase-0 codebase complete (report-gen runs); first thin data in; awaiting @deploy continuous run for calibration + a stable basis/spread read before the go/no-go.” No capital committed.

Related

• crypto-shortterm — project index (status reflects this milestone).
• crypto-shortterm-phase0-live-milestone — the prior milestone (first live ~5-min run, four-feed architecture); this note is the follow-on (codebase complete + first numbers).
• crypto-shortterm-backtesting — the Phase-0 analysis report (phase0-report.md) this milestone makes real; the go/no-go gate.
• crypto-shortterm-open-questions — CONFIRM gates; the empty-book watch sharpens gate 5 (live-window depth); fee-at-match (gate 3) still pending.
• crypto-shortterm-strategy-design — Rev-2 reframe the basis corroborates; the regime the empty-book risk threatens.
• crypto-shortterm-data-sources — the ~$150–250 touch-depth finding the empty-book read echoes.
• crypto-shortterm-risk-management — the confidence band that already carries the basis.