From 0ea6f642d1f26280dce35799bb70e9dce4c54d0d Mon Sep 17 00:00:00 2001
From: jaxperro <jaxtorres@gmail.com>
Date: Thu, 23 Jul 2026 15:11:33 -0400
Subject: [PATCH] =?UTF-8?q?research:=20copy=5Fage=5Fslices=20(T10)=20?=
 =?UTF-8?q?=E2=80=94=20inconclusive=20at=20n<=3D15/cell;=206-24h=20bucket?=
 =?UTF-8?q?=20positive=20both=20books=20(age~market-type=20confound=20note?=
 =?UTF-8?q?d);=20reruns=20as=20fills=20accrue?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 research/copy_age_slices.py | 76 +++++++++++++++++++++++++++++++++++++
 1 file changed, 76 insertions(+)
 create mode 100644 research/copy_age_slices.py

diff --git a/research/copy_age_slices.py b/research/copy_age_slices.py
new file mode 100644
index 00000000..b4c2e354
--- /dev/null
+++ b/research/copy_age_slices.py
@@ -0,0 +1,76 @@
+#!/usr/bin/env python3
+"""T10 EXPLORATORY (2026-07-23) — market-age copy weighting: do sharp
+entries on freshly-listed markets (books young, thin, unopinionated) carry
+more edge than steady-state copies? Age proxy = signal_ts − first print on
+the cond in the tape (complete coverage; markets older than the tape era
+bucket as 'pre-tape'). Chain-true grading; parity-era BUY fills both
+books. Output: EV by age bucket → a follow-filter weight if a gradient
+exists (kill: no gradient at n>=30/bucket)."""
+import json
+import os
+import sys
+
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+import tape                                    # noqa: E402
+import forward as fwd                          # noqa: E402
+
+ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+PARITY_T0 = 1784260140
+BUCKETS = [(0, 3600, "<1h"), (3600, 6 * 3600, "1-6h"),
+           (6 * 3600, 86400, "6-24h"), (86400, 3 * 86400, "1-3d"),
+           (3 * 86400, 1e12, ">3d")]
+
+
+def main():
+    db = tape.connect()
+    tape.build_resolved(db)
+    t_lo, = db.execute("SELECT min(ts) FROM trades").fetchone()
+    fills = []
+    for path, book in ((os.path.join(ROOT, "copybot_fills.jsonl"), "paper"),
+                       (os.path.join(ROOT, "copybot_fills.live.jsonl"), "live")):
+        for ln in open(path):
+            r = json.loads(ln)
+            if (r.get("side") == "SELL" or r.get("untracked")
+                    or r.get("ts", 0) < PARITY_T0 or not r.get("my_price")):
+                continue
+            r["_book"] = book
+            fills.append(r)
+    pays = fwd.payouts_for(db, [str(f["token"]) for f in fills])
+    rows = []
+    for f in fills:
+        pay = pays.get(str(f["token"]))
+        if pay is None or pay == 0.5:
+            continue
+        first, = db.execute(
+            "SELECT min(ts) FROM trades WHERE asset = ?",
+            [str(f["token"])]).fetchone()
+        if first is None:
+            continue
+        age = f["ts"] - first
+        pre_tape = first <= t_lo + 3600      # market predates tape coverage
+        rows.append({"book": f["_book"], "age": age, "pre": pre_tape,
+                     "px": f["my_price"],
+                     "ev": 100.0 / f["my_price"] * (pay - f["my_price"]),
+                     "win": pay == 1.0})
+    print(f"chain-graded parity fills: {len(rows)}")
+
+    def line(tag, rs):
+        if not rs:
+            return
+        n = len(rs)
+        print(f"  {tag:<10} n={n:<4} EV/$100 "
+              f"{sum(r['ev'] for r in rs)/n:+7.2f} · hit "
+              f"{sum(r['win'] for r in rs)/n:.2f} · avg px "
+              f"{sum(r['px'] for r in rs)/n:.2f}")
+
+    for book in ("paper", "live"):
+        rs = [r for r in rows if r["book"] == book]
+        print(f"\n== {book.upper()} ==")
+        line("pre-tape", [r for r in rs if r["pre"]])
+        fresh = [r for r in rs if not r["pre"]]
+        for lo, hi, tag in BUCKETS:
+            line(tag, [r for r in fresh if lo <= r["age"] < hi])
+
+
+if __name__ == "__main__":
+    main()