docs + audit sweep 2026-07-23: fill-model lesson, five-test day, archive pass

FINDINGS: 'The fill model is the next scorer' section (A2 chain grade
-$7.54/fill x1344 confirms the surge kill; oracle harness chain grade
vetoes the ledger-positive tiers; virtual-book +26% variance footnote;
the five tandem tests and the makers-on-the-wall through-line).
HANDOFF: snapshot -> 07-23 (7-wallet rev 5, dark flags, Friday agenda
incl #20/#21). READMEs: /test consolidation row, measurement-harness
research row, study statuses + new script inventory.
Archive: value/ (closed 07-19) + its test, ETHERSCAN_MIGRATION.md,
order_probe v1 -> archive/; replay_out/ gitignored; links repaired.
Tests: all 7 active scripts pass post-move.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

.gitignore

@@ -53,3 +53,4 @@ live/tape_sync.log
 research/meta/
 research/.surge*.pull.*
 research/.oracle*.pull.*
+research/replay_out/

FINDINGS.md

@@ -613,6 +613,48 @@ trigger, and the label-gap artifact class dies at the source. Tests are
 additive, never subtractive: raw streams are append-only, analysis reads
 only, and a study that needs different data records a NEW stream.
 
+## The fill model is the next scorer (2026-07-23)
+
+The first full chain grades of both measurement harnesses landed together
+and told one story from two directions.
+
+**Study A, confirmed dead twice.** The A2 arm's 1,344 chain-graded settles
+read **−$7.54/fill** — within $2 of the corrected ledger, at real
+execution, on the full trigger stream. Meanwhile the offline replay of the
+v1 $100/5% bankroll spec over the same attempts stream sat at **+26%** —
+its cash gate picked 70 of 2,692 attempts and got lucky (29W/17L). A book
+trajectory at n=46 carries no information a measurement at n=1,344
+doesn't overrule; that asymmetry is why bankroll specs lost verdict power
+in the A2 redesign, filed here as the standing example.
+
+**Study B, the divergence that matters.** The oracle harness's first chain
+grade: **E≥0.07 −$8.51/fill (n=235) · E≥0.10 −$5.90 (n=127) · ~40% hit**
+— while the forward ledger holds those tiers slightly positive. Both now
+score through payouts_for, so this is not round three again — it is the
+same LESSON one layer down: after the resolution model, the **fill model**
+is the next place a sim quietly manufactures edge. The ledger's
+6.7s-lag/3s-hold worst-print fills include entries no real order could
+have gotten (the book repriced) and price the rest against prints instead
+of the adversely-selected residue a live taker actually receives. The
+capturability instrument existed precisely to veto this; it has. Taker
+arm: evidence-dead (formal tier bars keep accruing, per pre-registration).
+
+**Five tandem tests, one day, all on already-flowing data** (scripts in
+research/, each rerunnable): the copy edge is REAL but concentrated
+(esports ~all of it; the live book's drag sits exactly there — the #14
+tension quantified); resting-bid copy entries beat taker FAKs **+$17.45
+vs +$12.86 per $100 signal** at 93% 60s trade-through (→ #20,
+pre-registered, dark-deployed); mirrored sharp exits are bankroll ops,
+not signal — no wallet's exits added value, 0xbadaf319's were 99% wrong
+(→ #21, same); the orders_matched stream holds a **673-wallet
+maker-sharp species** (+$5.9M pooled, ~33 expected by chance, 86%
+invisible to every taker screen) — not copyable by taking, the
+inventory-lean signal is the follow-on; and the sibling-sum scan showed
+print-substrate "arbs" are resolution-dust artifacts — standing-book data
+or nothing. The through-line of the week, now written as one sentence:
+**every taker edge died at the requote wall, and every instrument that
+measured the wall found the makers standing on top of it.**
+
 ## Repo layout
 
 - `insider.py` — the detector: z-score/p-value, timing/freshness/sizing signals,

HANDOFF.md

@@ -42,14 +42,18 @@ min_order_usd $1 = venue reality). If something looks genuinely dangerous,
 DISARM (`flyctl secrets unset LIVE_CONFIRM -a wwf-copybot-live`) rather than
 push through.
 
-## Snapshot (2026-07-21)
+## Snapshot (2026-07-23)
 - **wwf-copybot-live** (REAL, ARMED): ~$62 equity ($66.42 contributed,
   realized −$12.17 lifetime — day-one incident + honest recognitions),
-  6-wallet Set E rev 4, 4% of working equity/bet, alarm-free after the
+  **7-wallet Set E rev 5** (JuiceFarm promoted 2026-07-23 off the #13
-  07-20 reconcile (CASH≠CHAIN −$1.75: empty-cond Odyssey bet the venue
+  bench method run early: +17%/+34% both windows, insider z 2.5, 1%
-  auto-redeemed → hand-settled won +$0.68, $0.05 rounding folded; root
+  refunds; auto p80 floor $753), 4% of working equity/bet.
-  cause FIXED same day, #18: RTDS metadata-less rows now enrich at seed,
+  **#20/#21 dark flags deployed 2026-07-23** (`follow.entry_mode` /
-  repair_market_meta un-sticks cond-less books each settle pass). **wwf-copybot** (paper $1k): same set, FAK-parity fills.
+  `exit_mode`, defaults taker/mirror = byte-identical behavior; flip at
+  Friday's read opens the pre-registered forward windows — maker entries
+  rest at the sharp's price w/ 60s registry TTL; hold-through logs every
+  ignored sell to copybot_ignored_exits[.live].jsonl as the mirror
+  counterfactual). **wwf-copybot** (paper $1k): same set, FAK-parity fills.
   Both on the audit-hardened build (locks, chain-gated sweep, boot-id
   single-writer guard, TLS'd user-ws — HANDOFF_ARCHIVE rev 16). 2026-07-20:
   FAK no-match OPENs get one re-quote retry on both bots, PER-NICHE waits
@@ -66,7 +70,7 @@ push through.
   nightly → ~15 min; the box no longer needs the Mac to stay healthy.
   `live/parquet/` = complete durable layer (Stage-1 MotherDuck feedstock).
 - **VALUE experiment: CLOSED 2026-07-19** — sub-2¢ hypothesis refuted
-  (1W/993L, 0.075x); post-mortem in value/PLAN.md; app destroyed.
+  (1W/993L, 0.075x); post-mortem in archive/value/PLAN.md; app destroyed.
 - **research/ (SILO — never touches the bots)**: tape-era edge factory.
   Sharp screen (`live/tape_sharps.py`, proxy chain-validated 742/742, 25
   copyable candidates); execution sim fitted on 29 labeled live attempts.
@@ -89,9 +93,13 @@ push through.
 - **wwf-oraclebot (PAPER, ~$3/mo, live 2026-07-22)**: Study B real-time
   harness — fair value tick-by-tick on the venue's own settlement feed,
   all E-tiers tracked, $100 FAK walks, three settle layers ending in
-  nightly chain truth (grade_oracle.py → oracle_paper_ledger.jsonl). Early
+  nightly chain truth (grade_oracle.py → oracle_paper_ledger.jsonl).
-  live cells lean positive at E≥0.07 — consistent with the corrected tape
+  **First chain grade (2026-07-23): E≥0.07 −$8.51/fill (n=235), E≥0.10
-  read. Same attempts/markouts/settles streams as A2.
+  −$5.90 (n=127), ~40% hit — real execution loses at every tier while the
+  sim ledger reads positive. Both instruments share the scorer now, so the
+  divergence is the sim's 6.7s-lag FILL model flattering the tiers (round
+  3's lesson recurring in the fill model). Taker arm is evidence-dead;
+  formal tier bars keep accruing; maker pivot (T1 sim) is the successor.**
 - **Data moat (2026-07-22, DATA LAW in research/README)**: all raw streams
   append-only and Mac-independent (Fly volumes + daily snapshots; recorder
   has ~3+ weeks offline headroom); forward.py backfills ledger-missing
@@ -100,8 +108,16 @@ push through.
   study (chain-true): NO scalp inside the dead surge signal — hidden-loss
   cohort bleeds from minute one; taker case closed at every horizon.
 - **Verdicts pending**: edge/size-up (#14, ~end of July, pre-registered) ·
-  Friday's combined read (#13 bench + #16 formal close + #17 oracle-tier
+  Friday's combined read (#13 bench + #16 formal close [A2 chain grade
-  decision + #19 sprint-plan disposition — every number now chain-true).
+  −$7.54/fill × 1,344 independently confirms the kill; virtual book's
+  +26% on n=46 is the variance footnote, not a signal] + #17 taker-arm
+  decision + #19 disposition + **#20/#21 flip decisions** — every number
+  chain-true). Five tandem tests 2026-07-23 (research/: copy_edge_slices,
+  copy_maker_entry, maker_sharps, sibling_sum_scan, sell_mirror_study):
+  T3 maker entries +$17.45 vs +$12.86 taker → #20; sells anti-signal →
+  #21; 673-wallet maker-sharp species (+$5.9M, 86% invisible to taker
+  screens) → inventory-lean follow-on; sibling-sum = print artifact,
+  parked; T1 crypto maker-quote sim needs its re-run.
 - Dashboards: jaxperro.com/{trading,live,test,value} — /test = both paper
   studies on one page (old /surge + /oracle URLs redirect) · daily pipeline
   on the Mac at 08:00 (launchd, lockfile) — floors, bench forward table,
@@ -126,4 +142,4 @@ push through.
 - The bots commit their own state — always `git pull --rebase --autostash`
   before pushing from a session.
 - Read next: README.md (architecture + gotchas 1-18) · FINDINGS.md (research
-  story) · value/PLAN.md (the refutation) · HANDOFF_ARCHIVE.md (history).
+  story) · archive/value/PLAN.md (the refutation) · HANDOFF_ARCHIVE.md (history).

HANDOFF_ARCHIVE.md

@@ -365,7 +365,7 @@ reviewing the fix. Recommended: re-run a probe on an IN-PLAY market first.**
 
 Self-contained pickup for a fresh session (human or AI). Read
 [README.md](README.md) gotchas 1–16 (16 is the new-stack one), 
-[FINDINGS.md](FINDINGS.md), [ETHERSCAN_MIGRATION.md](ETHERSCAN_MIGRATION.md),
+[FINDINGS.md](FINDINGS.md), [archive/ETHERSCAN_MIGRATION.md](archive/ETHERSCAN_MIGRATION.md),
 [LIVE_ROLLOUT.md](LIVE_ROLLOUT.md). The repo is authoritative.
 
 ## THE CRITICAL PATH IS DONE (2026-07-10 05:08Z)
@@ -474,7 +474,7 @@ that; the 5-min backstop poll covers detection meanwhile).
 
 ## Queued work (unchanged priorities)
 
-1. ETHERSCAN_MIGRATION.md phases 0→5.
+1. archive/ETHERSCAN_MIGRATION.md phases 0→5.
 2. Empirical fill model from `book` snapshots; depth gate before sizing up.
 3. Funding-cluster tracer port to Etherscan logs.
 4. Live-app watchdog (poll mode has no /health).

README.md

@@ -26,10 +26,10 @@ Three deployed pieces + one static dashboard:
 | **daily pipeline** (`live/daily.sh`) | this Mac, launchd **08:00** (a `pmset repeat wakeorpoweron` RTC wake at **07:58** makes this run on time even from sleep — set 2026-07-08, undo with `sudo pmset repeat cancel`; job wrapped in `caffeinate -i` so mid-run idle-sleep can't kill it. Heads-up: the wake fires even lid-closed-in-a-bag — cancel it if the Mac travels) | refresh the bet cache → 5-gate skill scan → fee-aware sharp selection → conviction floors → backtest book → publish JSON feeds to GitHub |
 | **copybot worker** (`copybot.py` via `host/start.sh`) | **Fly.io app `wwf-copybot`, region `arn` (Stockholm), 24/7** — migrated off Railway 2026-07-06: Railway ran it in a US region, which Polymarket's IP geoblock would 403 the moment orders got real; Stockholm is unrestricted AND ~25ms from the CLOB's eu-west-2 primaries. Every boot self-checks the geo-gate (`host/geocheck.py`) and verifies the clone against `git ls-remote` (stale-replica guard) | **T0 = the RTDS trade stream** (gotcha 18): every platform trade, wallet-attributed, **~1s detection** (measured p50 0.8s over 22k msgs), silent-stale guard force-reconnects a quiet socket. Backstops: the Alchemy address-activity webhook (~3s, `POST /alchemy`, signature-verified — **T0b chain seed** 2026-07-14: its tx hashes decode straight from the receipt's OrderFilled logs when the data-api indexer lags, so RTDS-silent markets still copy at push speed), a 5-min backstop poll, and the reconcile janitor. Per-wallet **trade cursor** paginates the activity feed so clip bursts can't scroll past a fetch (H3). 60s heartbeat settles/publishes; paper-copies with real fees/lag/slippage AND live-FAK thin-book reality (**paper FAK parity** 2026-07-16: a quote with no ask inside the protected band books a MISS, not a fill — keeps the live-vs-paper per-signal ratio honest; `live/edge.py` turns it into the daily bankroll-decision verdict in `history/edge.csv` + the Discord digest footer); commits its book back to the repo |
 | **REAL-MONEY worker** (same code, `COPYBOT_ROLE=live`) | **Fly.io app `wwf-copybot-live`, `arn`, 24/7, ARMED 2026-07-10** — separate app so the two books never share a process, state file, or failure mode. Trades the signer's **Deposit Wallet** `0x455e…45a1` (pUSD collateral) via the unified SDK (`polymarket-client`) | Same T0 RTDS detection (`RTDS_DETECT=1`) + Alchemy/poll backstops; **own-fill push** via the CLOB user channel (`userws` in the heartbeat) makes in-play pending holds adopt the moment they match — ws events only *trigger* the resolver, `get_order` stays the arbiter (gotcha 17). Paper-parity sizing (4% of equity, $1 venue floor, no price floor, hard caps retired — the **depth gate** is the scaling rail: skip dust books (ask5c<$50), stake ≤10% of 5c depth — the spread>0.08 SKIP was retired 2026-07-19 as a lag-era relic (guard + protected band bound overpay)). Every placement/exit/settle pings Discord. Disarm: `flyctl secrets unset LIVE_CONFIRM -a wwf-copybot-live` |
-| **VALUE bot** (CLOSED 2026-07-19 — [`value/README.md`](value/README.md)) | ~~Fly app `wwf-valuebot`~~ destroyed after the paper verdict: 994 tickets, 1W/993L, 0.075x — today's sub-2¢ asks are fully informed; the 2025-era calibration edge is gone — a hard SILO: own code (`value/valuebot.py`, zero copybot imports), own state/feed, own image, no shared wallet ever | Systematically buys sub-2¢ contracts (the calibration study's one underpriced bucket — [`value/PLAN.md`](value/PLAN.md)); honest FAK fill model, chain-truth settles, $1 tickets, event cap 1. Verdict at ~2k resolved tickets vs a ~1.05% break-even hit rate |
+| **VALUE bot** (CLOSED 2026-07-19 — [`archive/value/README.md`](archive/value/README.md)) | ~~Fly app `wwf-valuebot`~~ destroyed after the paper verdict: 994 tickets, 1W/993L, 0.075x — today's sub-2¢ asks are fully informed; the 2025-era calibration edge is gone — a hard SILO: own code (`value/valuebot.py`, zero copybot imports), own state/feed, own image, no shared wallet ever | Systematically buys sub-2¢ contracts (the calibration study's one underpriced bucket — [`archive/value/PLAN.md`](archive/value/PLAN.md)); honest FAK fill model, chain-truth settles, $1 tickets, event cap 1. Verdict at ~2k resolved tickets vs a ~1.05% break-even hit rate |
 | **RTDS tape recorder** ([`recorder/README.md`](recorder/README.md)) | **Fly.io app `wwf-recorder`, `arn`, 24/7** — third silo: no repo clone (code in image), no keys, 25GB volume | Records the FULL firehose (~8M events/day): every trade, order MATCH (maker side), comment, crypto tick — dual-socket capture (one socket alone measured 92.9% coverage; the twin covers per-conn silences). **Stage-0 warehouse (2026-07-21): the box folds its own segments into row-verified Parquet on the volume; the Mac mirrors + appends into `live/rtds.duckdb` every 15 min** (`com.jaxperro.tape-sync`). Nothing deletes without a verified second copy (gz needs its parquet; parquet needs the Mac's ack) |
 | **Discord digest** (`live/discord_daily.py`) | end of the daily pipeline | one message/day: the sharp list with profile links + 30-day conviction stats (per-trade pings retired 2026-07-04 for PAPER; the LIVE book pings every real placement/exit/settle) |
-| **dashboard** | [jaxperro.com/trading](https://jaxperro.com/trading) + [jaxperro.com/live](https://jaxperro.com/live) + [jaxperro.com/surge](https://jaxperro.com/surge) (static, in the `jaxperro` repo) | `/trading` renders the paper book, backtest book and sharp table; `/live` is the REAL MONEY page (reads `live/copybot_live_real.json` — live since 2026-07-10); `/surge` is the Study-A paper harness (reads wwf-surgebot's /feed + the nightly chain-graded ledger). Wallet cards everywhere show LIFETIME numbers server-side (window sums lied twice — 2026-07-21/22) |
+| **dashboard** | [jaxperro.com/trading](https://jaxperro.com/trading) + [jaxperro.com/live](https://jaxperro.com/live) + [jaxperro.com/test](https://jaxperro.com/test) (static, in the `jaxperro` repo) | `/trading` renders the paper book, backtest book and sharp table; `/live` is the REAL MONEY page (reads `live/copybot_live_real.json` — live since 2026-07-10); `/test` is BOTH paper studies on one page (A2 surge measurement arm + oracle harness, each reading its bot's /feed + nightly chain-graded ledger; old /surge and /oracle URLs redirect). Wallet cards everywhere show LIFETIME numbers server-side (window sums lied twice — 2026-07-21/22) |
 
 **The calibration experiment (running now):** a fresh $1,000 paper book,
 **reset 2026-07-08** so it measures exactly one thing — the follow set in
@@ -137,7 +137,7 @@ does).
 | `insider.py` | the original detector: z-score, pre-resolution timing, fresh-wallet flags, funding-cluster rings |
 | `smart_money.py` | shared HTTP helper + survivorship-corrected win-rate dashboard (`:8899`) |
 | `archive/webhook_receiver.py` | retired 2026-07-04: Alchemy webhook → per-trade Discord pings (replaced by `live/discord_daily.py`'s daily digest) |
-| `research/` | tape-era edge factory, **SILO'd from the bots** ([research/README](research/README.md)): read-only RTDS loaders, calibrated execution sim, pre-registered studies scored through a MANDATORY chain-truth overlay (the 2026-07-22 scorer law — tape-resolution timing is win-biased and killed Study A #16 at −$6/fill once corrected; oracle #17 continues at E≥0.07), the surge paper harness (wwf-surgebot), nightly forward ledger — the only source of study verdicts |
+| `research/` | tape-era edge factory, **SILO'd from the bots** ([research/README](research/README.md)): read-only RTDS loaders, calibrated execution sim, pre-registered studies scored through a MANDATORY chain-truth overlay (the 2026-07-22 scorer law — tape-resolution timing is win-biased and killed Study A #16 at −$6/fill once corrected, independently confirmed by the A2 measurement arm at −$7.54/fill chain-true; oracle #17: E0.04 killed, higher tiers ledger-positive but vetoed by the harness's real-latency chain grade — the 2026-07-23 fill-model lesson), TWO measurement harnesses (wwf-surgebot A2 + wwf-oraclebot: every-trigger $100 paper FAKs, attempts/markouts/settles append-only, bankroll specs replayed offline), nightly forward ledger — the only source of study verdicts |
 | `wide/` | frozen-subgraph bulk scanner (1.76M wallets, historical only — subgraph froze Jan 2026) |
 | `archive/` | everything retired, kept honest ([archive/README](archive/README.md)): the six failed strategies, earlier research sweeps (`hunt/huntwide/oos/copyback`), the superseded live selection layer (`live-research/`), the scrapped Polymarket-US venue probe (`us-venue/`), and retired infra (`retired-infra/`: Railway config, Mac launchd runner, GH-Actions cron) |
 

research/README.md

@@ -39,9 +39,26 @@ Layout:
               win-biased; see SCORER LAW) · title parsers · chain_overlay()
   sim.py      execution replayer calibrated on OUR live fills ledger
               (lag, FAK no-match, protected band, 3% taker fee)
-  study_flow.py    Study A — surge momentum (KILLED 2026-07-22, #16)
+  study_flow.py    Study A — surge momentum (KILLED 2026-07-22, #16; A2
-  study_oracle.py  Study B — crypto oracle fair value (#17: E0.04 killed,
+                   chain grade −$7.54/fill × 1,344 independently confirms)
-                   E≥0.07 accumulating)
+  study_oracle.py  Study B — crypto oracle fair value (#17: E0.04 killed;
+                   E≥0.07/0.1 ledger-positive but the harness chain grade
+                   2026-07-23 reads −$8.51/−$5.90 per fill at real latency
+                   — taker arm evidence-dead, tiers accrue to their formal
+                   bars, maker pivot is the successor hypothesis)
+  copy_edge_slices.py   T5 — parity-era copy edge by niche/lag/band/wallet
+                        (esports carries it; feeds #14)
+  copy_maker_entry.py   T3 — resting-bid copy entries beat taker FAKs
+                        (+$17.45 vs +$12.86/signal; basis of #20)
+  sell_mirror_study.py  sharp exits are bankroll ops, not signal
+                        (basis of #21 hold-through)
+  maker_sharps.py       T2 — 673 improbably-winning MAKERS in
+                        orders_matched (86% invisible to taker screens);
+                        follow-on: inventory-lean signal
+  sibling_sum_scan.py   T4 — print-substrate sum-arb scan (artifact-
+                        dominated; needs standing-book data; parked)
+  maker_quote_sim.py    T1 — crypto maker quoting at fair−m (stale-quote
+                        latency model; re-run pending)
   requote.py       crater→requote timing (feeds the bots' per-niche retry)
   forward.py       scores frozen studies on new tape days → forward_ledger
                    (payouts_for chain overlay mandatory; controls + sub5c

research/maker_quote_sim.py

@@ -0,0 +1,117 @@
+#!/usr/bin/env python3
+"""T1 EXPLORATORY (2026-07-23) — crypto maker quoting on the oracle feed.
+
+Every taker study died at the requote wall (60-83% craters; makers repriced
+in <4s). This sims BEING the maker: a resting bid pinned to oracle-fair
+minus a margin, refreshed with latency R (the stale-quote window is the
+risk), filled when a tape print crosses it, graded to chain truth.
+
+Model per print at t: our active bid = fair(S(t-R), vol(t)) - m (vol drift
+over R<=4s is negligible; S(t-R) is the staleness that matters). Fill if
+print px <= bid (the book crossed our level — queue-position optimism
+stated). Fill price = our bid. Maker pays no taker fee. Per-token cooldown
+60s, max 5 lots. Sprints only inside their window (no s0 lookahead —
+harness rule, not the tape scorer's).
+
+Grid (pre-declared, not tuned after): m in {2c, 4c, 7c} x R in {1s, 4s}.
+Readouts: EV/fill (chain), hit, fills/day, fair-markout at +60s (adverse
+selection: how far fair moves against us right after we're filled).
+NOT pre-registered — Stage 1 of the maker pivot; a live paper maker arm
+only if this survives its own optimism caveats."""
+import json
+import os
+import sys
+import time
+
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+import tape                                    # noqa: E402
+import study_oracle as so                      # noqa: E402
+import forward as fwd                          # noqa: E402
+
+MARGINS = (0.02, 0.04, 0.07)
+LATENCIES = (1.0, 4.0)
+COOLDOWN_S = 60
+MAX_LOTS = 5
+STAKE = 100.0
+
+
+def main():
+    db = tape.connect()
+    series = {s: so.TickSeries(tape.load_ticks(db, s))
+              for s in ("btcusdt", "ethusdt", "solusdt", "xrpusdt",
+                        "bnbusdt", "dogeusdt")}
+    outcomes = so.outcome_map(db)
+    tape.build_resolved(db)
+    uni = so.crypto_universe(db, outcomes, series)
+    tick_lo = min(s.ts[0] for s in series.values() if s.ts)
+    tick_hi = max(s.ts[-1] for s in series.values() if s.ts)
+    span_d = (tick_hi - tick_lo) / 86400
+    print(f"universe {len(uni)} tokens · tick span {span_d:.1f}d")
+    cells = {(m, R): dict(fills=[], last=0.0, n_tok={})
+             for m in MARGINS for R in LATENCIES}
+    for u in uni:
+        mkt = u["mkt"]
+        prints = db.execute("""SELECT ts, price FROM trades
+            WHERE asset = ? AND ts >= ? ORDER BY ts""",
+            [u["asset"], tick_lo]).fetchall()
+        if not prints:
+            continue
+        s = series[mkt["sym"]]
+        last_fill = {k: 0.0 for k in cells}
+        n_tok = {k: 0 for k in cells}
+        for ts, px in prints:
+            px = float(px)
+            if mkt["kind"] == "sprint" and ts < (mkt["t0"] or 0):
+                continue                     # no pre-window quoting
+            sig = s.vol_1s(ts)
+            if sig is None:
+                continue
+            for R in LATENCIES:
+                S_stale = s.at(ts - R)
+                f = so.fair_value(mkt, u["up"], S_stale, sig, ts)
+                if f is None:
+                    continue
+                for m in MARGINS:
+                    k = (m, R)
+                    bid = f - m
+                    if not (0.02 <= bid <= 0.95):
+                        continue
+                    if px > bid:
+                        continue             # print didn't reach our level
+                    if ts - last_fill[k] < COOLDOWN_S or n_tok[k] >= MAX_LOTS:
+                        continue
+                    last_fill[k] = ts
+                    n_tok[k] += 1
+                    # adverse selection: where is fair 60s after our fill
+                    f60 = so.fair_value(mkt, u["up"], s.at(ts + 60), sig,
+                                        ts + 60)
+                    cells[k]["fills"].append(
+                        {"asset": u["asset"], "ts": ts, "bid": bid,
+                         "fair": f, "mo60": (f60 - bid) if f60 else None})
+    filled_assets = {f["asset"] for c in cells.values() for f in c["fills"]}
+    pays = fwd.payouts_for(db, list(filled_assets))
+    print(f"grading {len(filled_assets)} filled tokens (chain overlay)…")
+    for (m, R), c in sorted(cells.items()):
+        fs = c["fills"]
+        graded = [(f, pays.get(f["asset"])) for f in fs]
+        graded = [(f, p) for f, p in graded if p is not None and p != 0.5]
+        if not graded:
+            print(f"m={m:.2f} R={R:.0f}s: {len(fs)} fills, none graded")
+            continue
+        pnl = wins = 0.0
+        for f, p in graded:
+            sh = STAKE / f["bid"]
+            pnl += sh * (p - f["bid"])       # maker: no taker fee
+            wins += p == 1.0
+        mo = [f["mo60"] for f, _ in graded if f["mo60"] is not None]
+        n = len(graded)
+        print(f"m={m:.2f} R={R:.0f}s: fills {len(fs)} ({n} graded) · "
+              f"{len(fs)/span_d:.0f}/day · EV/fill {pnl/n:+7.2f} · "
+              f"hit {wins/n:.2f} · avg bid "
+              f"{sum(f['bid'] for f,_ in graded)/n:.2f} · "
+              f"fair-markout60 {sum(mo)/len(mo)*100:+.1f}c"
+              if mo else "")
+
+
+if __name__ == "__main__":
+    main()