flywheel: schedule eod_replay + nightly_tune in both compose scheduler blocks

The tuner was fiction in production: the 20:00 ET scheduler ran only
premarket_eval (+ eod_reconcile), so learned_params.json sat 26 days stale and
the whole-market dataset never grew. Both scheduler loops now run, in order,
after the eval/reconcile pass:

  python eod_replay.py    || echo "[scheduler] eod_replay FAILED"
  python nightly_tune.py  || echo "[scheduler] nightly_tune FAILED"

Order matters: replay grows the dataset first, the tuner fits after. 20:00 ET is
after close, so replaying "today" is complete data — NOT look-ahead (replay is a
labeling job, not a live gate). Deletes eod_replay.sh (dead cron-era wrapper
pointing at a nonexistent venv; compose is the runner).

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

deploy/docker/truenas-compose.yml

@@ -191,6 +191,11 @@ services:
           if [ "$$(date +%H:%M)" = "20:00" ] && [ "$$(date +%u)" -le 5 ]; then
             python -m research.premarket_eval || echo "[scheduler] eval FAILED"
             python -m research.eod_reconcile || echo "[scheduler] reconcile FAILED"
+            # Flywheel: grow the whole-market dataset first, then fit the tuner on
+            # it. 20:00 ET is after close, so replay of "today" is complete data —
+            # NOT look-ahead (replay is a labeling job, not a live gate).
+            python eod_replay.py || echo "[scheduler] eod_replay FAILED"
+            python nightly_tune.py || echo "[scheduler] nightly_tune FAILED"
             sleep 120
           fi
           sleep 20

docker-compose.yml

@@ -133,6 +133,11 @@ services:
             echo "[scheduler] premarket_eval $$(date)"
             python -m research.premarket_eval || echo "[scheduler] eval FAILED"
             python -m research.eod_reconcile || echo "[scheduler] reconcile FAILED"
+            # Flywheel: grow the whole-market dataset first, then fit the tuner on
+            # it. 20:00 ET is after close, so replay of "today" is complete data —
+            # NOT look-ahead (replay is a labeling job, not a live gate).
+            python eod_replay.py || echo "[scheduler] eod_replay FAILED"
+            python nightly_tune.py || echo "[scheduler] nightly_tune FAILED"
             sleep 120
           fi
           sleep 20

eod_replay.sh

@@ -1,11 +0,0 @@
-#!/usr/bin/env bash
-# End-of-day whole-market replay: what today's logic would have done + grow the
-# stored dataset for the tuner/verification. Schedule right after the close,
-# before nightly_tune (which then tunes on the freshly-grown data).
-#   crontab -e  ->  5 16 * * 1-5  /mnt/c/code/momentum/eod_replay.sh
-ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-VENV_PY="/home/philip/.venvs/momentum/bin/python"
-cd "$ROOT" || exit 1
-mkdir -p "$ROOT/data"
-echo "=== eod_replay $(date) ===" >> "$ROOT/data/eod_replay.log"
-PYTHONPATH="$ROOT" "$VENV_PY" eod_replay.py >> "$ROOT/data/eod_replay.log" 2>&1