Daily Bet Research Report — Wednesday, September 23, 2026

Start: 11:13 ET | End: 11:18 ET

Bankroll: $60.23 | Stake (12.5%, one play): $7.53 Record: 3-1-0 (Main bets, main leg, since 9/14 reset) | Hit rate: 75.0% Slate: 16 games · 9 analysed · 23 sources · Playbooks: MLB v1 (2026-09-16) · Verified at 11:13 ET

MLB only. NFL/NHL/NBA have no bookable games today; only the MLB playbook was read.

Section 2: Game Analysis

Unanalysed: 7 of 16. ARI @ COL is a standing no-play (H1). Both legs of TOR @ BAL carry no DraftKings price on the ESPN feed. NYM @ TEX, LAA @ ATH, SD @ LAD and HOU @ SEA went unanalysed because the ESPN scoreboard payload only mapped eight event IDs to matchups and I stopped probing IDs once the eight-game floor was cleared with time in hand — I'd rather report that plainly than pad the count with games I priced by guesswork.

Section 3: Today's Plays

Negative edge on #1, #2 and #3. All three are moneylines inside a tight, efficiently-priced band; the calibration haircut eats the margin. Only #4 shows positive edge, and it is the lowest-confidence play on the card. That tension is the honest state of this slate.

Section 4: Reasoning

#1 — Detroit Tigers ML, 1.55. This is a §3 bullpen-state play, the playbook's "most under-priced factor available" [MODERATE]. Washington's listed starter is Richard Lovelady, a career reliever — 154 appearances against 5 career starts, 5.01 ERA — so this is a bullpen game, and it comes the day after their pen threw 108 pitches across three arms in Detroit (J. Kent 72 pitches / 19 batters faced, Sinclair 14, Tolman 22, per the 9/22 boxscore). Detroit's own pen barely worked: Kinley 8, Hurter 19, Brieske 10, 37 pitches total behind D. Anderson's 71. Framber Valdez (ERA 4.05, xERA 4.42, diff −0.365) is not sharp, but he only has to out-pitch a patchwork. The line has not moved a cent from −180, which means the market has not repriced Washington's staff state — exactly the slow-absorbing information §3 describes. Risk: Washington won 3-1 last night behind the same kind of staff and their opener plan can work for five innings; at 1.55 the price gives no cushion if it does.

#2 — Chicago Cubs ML, 1.57. A §2 ERA−xERA play, tagged [MODERATE] by the playbook's own honest limitation. Gausman is the playbook's worked example and the 2026 numbers still fit: ERA 4.55 against xERA 4.01, diff +0.542 — he has been unlucky and the results overstate the damage. Miami's Ryan Gusto is the mirror: ERA 4.14, xERA 4.96, diff −0.823 — better on paper than in substance. Neither gap clears the 1.00 flag on its own, which is why the playbook forbids carrying a play on the gap alone; the support comes from §4 wind. MET Norway has Wrigley at 6.5 m/s from 42.8° at 23:00Z and 6.8 m/s from 51.5° at 00:00Z — call it 15 mph from ~46°, NNE off the lake, blowing IN, the 7.72-run version of the park (7.12 at ≥10 mph). Wind in suppresses the contact-quality gap Gusto has been surviving on. The line has not moved from −175. Risk: Miami won 8-2 at Wrigley last night, and the Cubs' own pen absorbed that loss's innings. No total here — with the wind read in hand H2 is satisfied, but 7.12 expected against a posted 7.0 is not an edge.

#3 — Minnesota Twins ML, 1.67. Built on §2 plus H4's logic rather than against it. Connor Prielipp carries ERA 4.94 against xERA 4.01, diff +0.926 — just under the flag, pointing to positive regression. San Francisco counters with Cesar Perdomo: 4 career games, 3 career starts, 19.1 career innings, and absent from the Savant qualified-pitcher board, so there is no xERA for him at all. H4 bars a play on a starter with fewer than 3 starts; Perdomo has exactly 3, and in any case the play is against his team and rests on that inexperience — the explicit carve-out. Minnesota also took the series opener 3-2 on the road. Risk: rookie variance cuts both ways by definition, Oracle Park suppresses the Twins' offence as much as the Giants', and −149 is a real price for two 70-win clubs.

#4 — CWS @ KC Over 8.5, 1.91. §3 again, on the market where the playbook says a tired pen "pays most reliably." Last night's game at Kauffman finished 14-11 and both bullpens are shorter for it: Kansas City spent Gonsolin 63 pitches / 15 batters faced in long relief plus Hoffman 17 and Lange 10 behind Lynch IV's 78; Chicago used four relievers — Schultz 23, Brazobán 14, Richards 23, Castillo 20 — behind Kay's 65. Today's starters do nothing to fix it: Seth Lugo sits at ERA 5.06 / xERA 5.09 (diff −0.031, so no hidden quality), and Chicago's Bryan Hudson is a 65.1-inning reliever, i.e. another bullpen game. §4 order of operations: Kauffman is a neutral-to-large park, and MET Norway has today's wind at ~3.6 m/s from 84° at first pitch — 8 mph, light, no suppression (speed only; the playbook only certifies an out/in bearing for Wrigley). Total moved 8.5 → 8.5 with the over drifting −107 → −109, a shade toward the over. Risk: two lineups that just combined for 25 runs are the classic candidate for a 3-2 hangover game, and Hudson's own ERA (3.03) is better than his role implies.

Sources

• https://statsapi.mlb.com/api/v1/schedule?sportId=1&date=2026-09-23&hydrate=probablePitcher
• https://statsapi.mlb.com/api/v1/schedule?sportId=1&startDate=2026-09-22&endDate=2026-09-22&fields=... (9/22 finals)
• https://statsapi.mlb.com/api/v1/people?personIds=702885,663992,687570,669199,701656&hydrate=stats(group=[pitching],type=[career])
• https://statsapi.mlb.com/api/v1/game/824061/boxscore (CWS @ KC 9/22 — full, then ?fields=info,label,value)
• https://statsapi.mlb.com/api/v1/game/824222/boxscore (WSH @ DET 9/22 — full, then ?fields=info,label,value)
• https://statsapi.mlb.com/api/v1/game/823543/boxscore?fields=info,label,value (TB @ NYY DH G2)
• https://baseballsavant.mlb.com/leaderboard/expected_statistics?type=pitcher&year=2026&min=q&csv=true
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events?dates=20260923&limit=100
• https://site.api.espn.com/apis/site/v2/sports/baseball/mlb/scoreboard?dates=20260923
• DraftKings odds, one per game: https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/{id}/competitions/{id}/odds for 401817045, 401817046, 401817047, 401817048, 401817049, 401817050, 401817051, 401817052, 401817053, 401817058
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/401817051/competitions/401817051/probables
• https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=41.948&lon=-87.656 (Wrigley)
• https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=39.05&lon=-94.48&altitude=230 (Kauffman)

Research limitations

Counts: 9 of 16 games analysed · 23 distinct sources fetched · 1 analysed game missing a Tier-1 number.

The one gap: Cesar Perdomo (SF) has no xERA — he is not on the Savant qualified board, which the playbook warns happens for fringe starters. I did not substitute FIP. His career line (4 G, 3 GS, 19.1 IP, 3.72) came from statsapi instead, and the play is built on the inexperience rather than on a metric I don't have.

⚠ H3 is only partially satisfied, and that applies to every play on this card. The rule wants both starters confirmed on two independent sources. I have MLB's own statsapi probablePitcher (the Step 1 ground truth) and an MLBAM-player-ID cross-match into Baseball Savant, which confirms identity but is not an independent board. A second board could not be obtained: ESPN's /probables sub-resource returned an empty body, and the ESPN scoreboard payload carries no pitcher names at all (grepped for Gausman, Valdez, Lovelady, probables — no matches). Treat all four plays as conditional on no late scratch; on 9/15 a board serving the previous day produced five phantom pitchers, and a single-source confirmation is exactly the exposure that rule exists to close.

Other gaps:

• Seven games unanalysed. The ESPN scoreboard response (88k chars, 5 lines) only yielded eight event-ID-to-matchup mappings under Grep -o, which appears to be a match cap on a single very long line rather than a truncation of the file. I recovered two more IDs (401817052 CWS@KC, 401817053 MIA@CHC) by fetching their odds and identifying the teams from the ESPN team $ref numbers. That method works and would have mapped the rest; I stopped after clearing the floor. Worth adding to the playbook: ESPN team IDs are stable and the odds payload alone identifies the game — 2 BOS, 4 CHW, 5 CLE, 6 DET, 7 KC, 8 MIL, 9 MIN, 10 NYY, 15 ATL, 16 CHC, 17 CIN, 20 WSH, 22 PHI, 23 PIT, 24 STL, 26 SF, 28 MIA, 30 TB, all verified against today's matchups.
• TOR @ BAL carries no DraftKings price on either leg of the makeup doubleheader (count: 0). Not a fetch failure — the feed genuinely has nothing.
• Umpire assignments: not obtained. Still on the playbook's "not yet verified" list; I did not spend the window on it. No play here rests on a total tight enough for a 0.3–0.5-run umpire effect to be decisive.
• Per-game bullpen usage took three attempts. The full boxscore (76k chars, 5 long lines) mixes season and per-game stats and Grep -o capped out before reaching the per-game block. ?fields=info,label,value returns a ~1k-char payload with Pitches-strikes and Batters faced per pitcher, plus observed weather and umpires — this is the clean way to get bullpen state and should go in §9 as verified.
• Statcast park factors: not fetched. Still unverified in the playbook; the two totals I looked at were reasoned from wind and pen state, not park multipliers.
• No sport outside MLB was in season today, so the NHL and NBA stubs were not used.
• Hard cutoff not reached.