Nightly self-improvement · slate 2026-09-23
Generated 2026-09-24T05:00:00.009029-04:00 ET from official finals (mlb-api). Every learning stage is listed with what it did. "Blocked" means it wanted to act and a safety rule stopped it; "nothing to learn" means it had no eligible evidence.
Slate
2026-09-23
status completed
Winners
8/16
Brier 0.2477
Learned or changed
0
of 5 stages
Blocked
0
wanted to act and could not
Release
v4.31.0
DRIFT
What each learning stage did
| Stage | Outcome | Detail |
|---|---|---|
| KPI settlement | failed | did not run |
| Bullpen usage model | nothing to learn | No pregame bullpen forecasts to settle. Refit: not attempted. |
| Adaptive parameter proposal (Luna) | nothing to learn | candidate_registered: |
| Auto-apply | nothing to learn | no_promotable_candidate |
| Code agent | nothing to learn | disabled_by_config: code changes are made on the development host, never in the container. |
Games
| Game | Predicted | Actual | Home win prob | Result | Main driver | Starters |
|---|---|---|---|---|---|---|
| ARI@COL | 8-6 | 5-3 | 0.440 | hit | Merrill Kelly: 16 proj / 16 real outs; Mason Adams: 13 proj / 18 real outs | |
| CHW@KC | 7-5 | 4-5 | 0.440 | miss | Backed the wrong side with real conviction | Bryan Hudson: 15 proj / 1 real outs; Seth Lugo: 14 proj / 10 real outs |
| CIN@ATL | 5-6 | 2-3 | 0.560 | hit | Andrew Abbott: 13 proj / 21 real outs; Chris Sale: 19 proj / 17 real outs | |
| CLE@BOS | 5-4 | 0-1 | 0.495 | miss | Run environment far lower than projected (over-dispersed) | Foster Griffin: 16 proj / 18 real outs; Sonny Gray: 18 proj / 18 real outs |
| HOU@SEA | 4-5 | 5-6 | 0.515 | hit | Ethan Pecko: 15 proj / 8 real outs; George Kirby: 18 proj / 15 real outs | |
| LAA@OAK | 5-4 | 3-7 | 0.460 | miss | Small edge on the losing side | Walbert Ureña: 18 proj / 12 real outs; Jeffrey Springs: 14 proj / 13 real outs |
| MIA@CHC | 5-6 | 3-2 | 0.530 | miss | Run environment far lower than projected (over-dispersed) | Ryan Gusto: 14 proj / 16 real outs; Kevin Gausman: 17 proj / 17 real outs |
| MIL@PHI | 5-3 | 4-1 | 0.440 | hit | Logan Henderson: 18 proj / 15 real outs; Aaron Nola: 18 proj / 18 real outs | |
| MIN@SF | 4-3 | 4-2 | 0.489 | hit | Connor Prielipp: 17 proj / 16 real outs; Cesar Perdomo: 14 proj / 16 real outs | |
| NYM@TEX | 6-3 | 7-2 | 0.405 | hit | Nolan McLean: 18 proj / 21 real outs; Cody Bradford: 14 proj / 13 real outs | |
| SD@LAD | 5-7 | 5-1 | 0.535 | miss | Run environment far lower than projected (over-dispersed) | Robbie Ray: 15 proj / 10 real outs; Yoshinobu Yamamoto: 20 proj / 18 real outs |
| STL@PIT | 5-6 | 5-1 | 0.556 | miss | Run environment far lower than projected (over-dispersed) | Matthew Liberatore: 15 proj / 17 real outs; Lake Bachar: 15 proj / 5 real outs |
| TB@NYY | 5-4 | 2-9 | 0.495 | miss | Starting-pitcher projection was wrong | Mason Englert: 15 proj / 11 real outs; Gerrit Cole: 17 proj / 17 real outs |
| TOR@BAL | 4-5 | 2-4 | 0.505 | hit | Max Scherzer: 15 proj / 15 real outs; Chris Bassitt: 15 proj / 20 real outs | |
| TOR@BAL | 6-5 | 2-4 | 0.488 | miss | Run environment far lower than projected (over-dispersed) | |
| WAS@DET | 6-4 | 4-2 | 0.495 | hit | Richard Lovelady: 11 proj / 4 real outs; Framber Valdez: 18 proj / 16 real outs |
Day over day
Rolling evidence window against the prior night. Associations, not causal proof.
| Metric | Prior | Current | Change | Result |
|---|---|---|---|---|
| brier_score | 0.2448 | 0.2442 | -0.0006 | improved |
| accuracy | 0.5630 | 0.5546 | -0.0084 | regressed |
| team_run_mae | 2.4958 | 2.5168 | 0.0210 | regressed |
| total_run_mae | 3.3109 | 3.3697 | 0.0588 | regressed |
| calibration_error | 0.1034 | 0.1010 | -0.0024 | improved |
| starter_er_mae | 1.7120 | 1.7560 | 0.0440 | regressed |
| starter_outs_mae | 3.5160 | 3.5720 | 0.0560 | regressed |
Calibration produced for the next slate
| Metric | Scale used | Measured | Rows | Status |
|---|---|---|---|---|
| No calibration produced. | ||||
Luna KPI candidates and their promotion gate
Promotion needs 400+ prospective rows over 5+ slates after the proposal, with a gain interval above zero. Only verified pregame forecasts count; local replays never do.
| KPI | Status | Target | Proposed | Rows | Slates | Gain interval | Progress |
|---|---|---|---|---|---|---|---|
| luna_pitcher_release_speed_vs_righthanded_starters | shadow | SO | 20260923 | 330/400 | 5/5 | [-0.0007, 0.0007] | |
| luna_pitcher_barrel_rate_vs_rhh_righthanded_starters | shadow | HR | 20260923 | 330/400 | 5/5 | [-0.0004, 0.0049] | |
| luna_batter_contact_rate_vs_breaking_righthanded_starters | shadow | H | 20260923 | 330/400 | 5/5 | [-0.0045, 0.0044] | |
| luna_batter_contact_rate_vs_breaking_left_handed_starters_for_strikeouts | shadow | SO | 20260924 | 122/400 | 4/5 | [-0.0027, 0.0021] | |
| luna_pitcher_contact_xwoba_allowed_vs_righthanded_starters_for_strikeouts | shadow | SO | 20260924 | 122/400 | 4/5 | [-0.0001, 0.0000] | |
| luna_pitcher_whiff_rate_breaking_vs_left_handed_hitters_for_hits | shadow | H | 20260924 | 122/400 | 4/5 | [-0.0056, -0.0011] | |
| luna_pitcher_home_run_rate_vs_righthanded_starters | shadow | HR | 20260925 | 122/400 | 4/5 | [-0.0010, 0.0056] | |
| luna_batter_ground_ball_rate_vs_fastball_vs_lhp_for_home_runs | shadow | HR | 20260926 | 86/400 | 3/5 | [-0.0001, -0.0000] | |
| luna_pitcher_whiff_rate_vs_right_handed_starters_for_batter_strikeouts | shadow | SO | 20260929 | 51/400 | 2/5 | [–, –] | |
| luna_batter_whiff_rate_vs_breaking_right_handed_starters_for_hits | shadow | H | 20260929 | 51/400 | 2/5 | [–, –] | |
| luna_batter_walk_rate_vs_righthanded_pitchers_for_home_runs | retired | HR | 20260924 | 0/400 | 0/5 | [–, –] |
Operator actions the cycle asked for (0)
Changes to code or a model recipe are never made by the nightly cycle. These are the measured findings that need one.
| Priority | Owner | Finding | Action |
|---|---|---|---|
| No operator actions requested. | |||
Luna post-mortem
No Luna post-mortem this night.