mirror of
https://github.com/jaxperro/winning-wallet-finder.git
synced 2026-08-02 18:57:47 +00:00
jaxperro
cf9465f38a
leanbot: orders_matched inventory tracking on the nightly-published screened set (box never screens itself), frozen $150-500 one-sided crossing, paper FAK $100 capped at print+3c, complement routing for net-short leans, MAX_PER_EVENT=2, premium logged on every attempt. guards.py: ex-best-day / ex-top5 / top-event share — the cuts that caught both studies today, now machinery every grader reports nightly. grade_lag: v2 band arm (15-40c, forward-only from V2_FREEZE_TS, bar +$8) + three report-only control bands. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
85 lines
3.3 KiB
Python
85 lines
3.3 KiB
Python
#!/usr/bin/env python3
|
||
"""Concentration guards, shared by every harness grader (#26/#27).
|
||
|
||
Two studies passed their headline bars this week on results that one
|
||
session or five episodes were carrying (Study C ex-best-day +$1.47 vs a
|
||
+$2 bar; Study D ex-best-day −$5.86 on a +$4.49 headline). These cuts
|
||
turn that lesson into machinery: every grader reports them nightly, and
|
||
a PASS whose ex-best-day EV is <= 0 is recorded CONCENTRATED and does
|
||
not graduate.
|
||
|
||
rows: dicts with chain_pnl, cost, ts, and (optionally) event.
|
||
"""
|
||
import collections
|
||
import datetime as dt
|
||
|
||
|
||
def cuts(rows, pnl_key="chain_pnl"):
|
||
"""-> dict of the guard readings (empty dict when there is nothing)."""
|
||
rows = [r for r in rows if r.get(pnl_key) is not None]
|
||
n = len(rows)
|
||
if not n:
|
||
return {}
|
||
tot = sum(r[pnl_key] for r in rows)
|
||
staked = sum(r.get("cost") or 0 for r in rows)
|
||
out = {"n": n, "pnl": round(tot, 2), "ev": round(tot / n, 2),
|
||
"pct_staked": round(100 * tot / staked, 1) if staked else None,
|
||
"wins": sum(1 for r in rows if r.get("chain_payout") == 1.0)}
|
||
out["hit"] = round(out["wins"] / n, 3)
|
||
# episode concentration
|
||
srt = sorted(rows, key=lambda r: -abs(r[pnl_key]))
|
||
for k in (1, 5):
|
||
if n > k:
|
||
ex = sum(r[pnl_key] for r in srt[k:])
|
||
out[f"ex_top{k}_ev"] = round(ex / (n - k), 2)
|
||
out["top5_share"] = (round(100 * sum(r[pnl_key] for r in srt[:5]) / tot, 0)
|
||
if tot else None)
|
||
# TIME concentration — the guard both studies needed
|
||
byday = collections.defaultdict(lambda: [0, 0.0])
|
||
for r in rows:
|
||
d = dt.datetime.utcfromtimestamp(r["ts"]).strftime("%m-%d")
|
||
byday[d][0] += 1
|
||
byday[d][1] += r[pnl_key]
|
||
out["days"] = len(byday)
|
||
out["by_day"] = {d: [v[0], round(v[1], 0)] for d, v in sorted(byday.items())}
|
||
if len(byday) > 1:
|
||
best = max(byday.values(), key=lambda v: v[1])
|
||
exn = n - best[0]
|
||
if exn > 0:
|
||
out["ex_best_day_ev"] = round((tot - best[1]) / exn, 2)
|
||
out["ex_best_day_n"] = exn
|
||
# EVENT concentration (the gate Study C's fade arm failed at 32%)
|
||
ev = collections.defaultdict(float)
|
||
for r in rows:
|
||
if r.get("event"):
|
||
ev[r["event"]] += r[pnl_key]
|
||
if ev and tot:
|
||
out["events"] = len(ev)
|
||
out["top_event_share"] = round(
|
||
100 * max(ev.values(), key=abs) / tot, 0)
|
||
return out
|
||
|
||
|
||
def line(tag, c, pass_ev=None):
|
||
"""One printable verdict line + the guard verdict."""
|
||
if not c:
|
||
return f"{tag}: no graded rows"
|
||
s = (f"{tag}: n={c['n']} {c['wins']}W · EV/fill {c['ev']:+.2f} · "
|
||
f"hit {c['hit']:.3f} · {c['pct_staked']:+.0f}% staked · "
|
||
f"{c['days']}d")
|
||
if "ex_best_day_ev" in c:
|
||
s += f" · EX-BEST-DAY {c['ex_best_day_ev']:+.2f}"
|
||
if "ex_top5_ev" in c:
|
||
s += f" · ex-top5 {c['ex_top5_ev']:+.2f}"
|
||
if "top_event_share" in c:
|
||
s += f" · top-event {c['top_event_share']:+.0f}%"
|
||
if pass_ev is not None:
|
||
bars = [c["ev"] >= pass_ev,
|
||
c.get("ex_best_day_ev", -1) > 0,
|
||
c.get("ex_top5_ev", -1) > 0,
|
||
abs(c.get("top_event_share") or 0) < 30]
|
||
s += (" -> " + ("ALL GUARDS CLEAR" if all(bars) else
|
||
"CONCENTRATED (headline only)" if bars[0] else
|
||
"below bar"))
|
||
return s
|