Daily Bet Research Report — Saturday, September 26, 2026

⚠ THIS RUN FIRED 90 MINUTES LATE — THE 12:00 DEPLOY HAS ALREADY PASSED

Scheduled for 11:13 AM ET; the scheduler fired it at 12:43 PM ET. The 11:50 build cutoff and the 12:00 deploy were both gone before the first action of the run. The site's 12:00 deploy published stale content (yesterday's report). This report was written and built after that deploy, so Mitch must run rebuild_and_deploy.bat manually for it to reach the site. Two early games were also lost to the delay (see Section 2).

Start: 12:43 ET | End: 12:50 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: 13 games · 11 analysed · 27 sources · Playbooks: MLB only · Verified at 12:44 ET

Section 2: Game Analysis

Two of 13 games went unanalysed, both because of the 90-minute scheduler delay: NYM @ WSH (12:35 PM, already in Warmup when the run started) and PIT @ DET (1:10 PM, 23 minutes away at the 12:43 start — inside the 30-minute cutoff). On an on-time 11:13 run both would have been bookable.

Section 3: Today's Plays

⚠ All four plays carry negative edge. Nothing on today's board prices above my honest read of it. The least-negative is #3 (−0.5%), effectively a fair price; #1 and #2 are roughly 3-4 points of vig against. This is a thin card, and a pass is a legitimate answer to it — the penultimate day of the season is a day when clinched teams rest, eliminated teams field call-ups, and the market has more information about which is which than I do at midday.

Section 4: Reasoning

#1 — Toronto Blue Jays ML, 1.61

The argument is §6 (Lineup, Rest and Travel) reinforced by §2 (Starting Pitcher), and it is the one game today where I have the lineups rather than guessing at them. Cincinnati's posted order runs four call-ups or fringe bats — Carlos Jorge, Sal Stewart, Héctor Rodríguez and Juan Brito — around De La Cruz and Suárez, while Toronto posts something close to its real nine: Guerrero Jr., Kirk, Okamoto, Giménez, Clement. Both clubs are eliminated (TOR 77-82, CIN 74-85), so the §6 asymmetry isn't about motivation, it's about who is actually in the box. On top of that sits a genuine starter gap: Yesavage 3.67 ERA / 3.36 xERA versus Lowder 6.01 ERA / 5.50 xERA — a 2.14-run difference in expected run prevention, and Rogers Centre is a dome, so §4's weather chain drops out entirely and this is the pure pitching-and-lineup read the playbook says a dome total should be. Evidence tags: §6 is [WEAK to MODERATE] and §2's xERA is [MODERATE], so this is a moderate-strength case, not a strong one, and neither pitcher's ERA−xERA gap clears the 1.00 threshold on its own. The line moved 8 cents toward Toronto (-156 → -164), which by the prompt's own rule means the market agrees and I am late — that is most of why edge is negative. The risk that beats me: Toronto is eliminated too, and a club with nothing to play for can pull Guerrero and Kirk after four innings regardless of what the pre-game card says; if Toronto empties its bench in the fifth, the lineup edge I'm buying evaporates mid-game. Source: statsapi boxscore 822759 (lineups), Baseball Savant expected_statistics.

#2 — Houston Astros ML, 1.62

This rests on §6, the part of it the playbook actually endorses — "a club that clinched or was eliminated this week" — and it is the sharpest version of that on the board. Houston is 79-80, tied with Texas atop the AL West, and has not clinched anything; the Athletics are 63-96 and eliminated. Houston's bullpen is also in shape to be used like a must-win pen: last night it threw only 58 pitches across four arms (Sousa 5, Blubaugh 15, VanWey 34, Hendrickson 4) behind Hunter Brown's 100, so no arm is on a third straight day per §3's breakpoint. Wesneski brings a 3.50 ERA / 3.89 xERA. The honest problem, and I am not going to bury it: §2's strongest signal on today's entire board points the other way. Jack Perkins carries ERA 5.90 against xERA 4.03, a +1.867 gap — past the playbook's 1.75 "headline-worthy" line — and a positive gap is a buy signal on his team, meaning Savant thinks the A's starter is a run and a half better than his ERA advertises. The line has also already moved 19 cents toward Houston (-143 → -162), so the motivation story is fully priced and then some. Those two facts are why confidence sits at 61 rather than the high 60s. The risk that beats me is exactly that: Perkins pitches to his xERA, the A's win the game they won 6-5 last night, and I paid -162 for a narrative the market already bought. Source: statsapi standings, statsapi boxscore 824947, Baseball Savant.

#3 — CLE @ KC Over 8.0, 1.87

This is the §3 bullpen play — the playbook's designated "most under-priced factor available," and it explicitly says a tired pen "is worth more on the total than on the moneyline." Last night's game at Kauffman was a 12-9 Cleveland win in which Cleveland used six relievers for 102 pitches (Herrin 18, Armstrong 19, Espino 13, Sabrowski 18, Gaddis 13, Festa 21) behind Gavin Williams' 74 — comfortably past §3's "4+ innings from the pen is a genuine next-day disadvantage." Kansas City's two highest-leverage arms both crossed the 25-pitch workload marker: Erceg 25 and Pearson 23. §2 adds support on the KC side: Wacha's ERA 3.33 against xERA 3.97 is a −0.636 gap, i.e. he has been lucky and regression is owed, which is a fade signal on the Royals however good the line looks; Bibee is neutral at +0.062. §4's weather chain is benign rather than helpful — Kauffman at first pitch forecasts 75°F with wind 3.4 m/s (7.6 mph) from 131°, light enough not to suppress anything, and per the playbook I report speed only because no bearing is verified for that park. Tags: §3 is [MODERATE] with a [STRONG] supporting mechanism, §2 [MODERATE] — the best-evidenced case on the card, which is why this is also the least-negative edge. The total has not moved at all (8.0, Over -115 at both open and current), so there is no steam in either direction to fight or follow. The risk that beats me: Cleveland has clinched the AL Central and will rest regulars, so the lineup that scored 12 last night may not be the one that shows up, and two rested starters can make a 3-2 game out of any bullpen situation. Source: statsapi boxscore 824058, MET Norway forecast, Baseball Savant.

#4 — HOU @ ATH Under 10.0, 1.85

A total of 10.0 is the highest number on the board by a run and a half, and the case for the under is that the two inputs driving it that high are both softer than they look. Perkins' 5.90 ERA is what makes this a double-digit total, but §2 says that ERA overstates how badly he has pitched — xERA 4.03, a +1.867 gap, the largest on the slate — and Wesneski is at 3.50 ERA / 3.89 xERA on the other side. §4's environmental input gives the over no help either: Sutter Health Park forecasts 25.2°C (77°F) with wind of only 3.5 m/s (7.8 mph) from 206° at first pitch, and per the playbook I read that as speed only, since Sacramento has no verified home-plate-to-centre bearing — light air, not carrying air. Neither bullpen is exhausted (58 and 62 pitches respectively last night), so late-innings run expectancy isn't inflated by §3 either. Tag: this is [MODERATE] at best and leans on an estimator the playbook itself calls weaker than it appears. Line movement agrees with me and that is a warning, not a comfort: the under went from -112 to -118 while the over cheapened from -108 to -102 — money is on the under, so I am following rather than leading, which caps the edge. The risk that beats me is the park itself: Sutter Health Park has played as a genuine offensive environment, the market set 10.0 knowing that, and if the A's are running call-ups against a Houston club that needs runs, 11 comes easily. Source: MET Norway forecast, Baseball Savant, ESPN odds 401817100.

Sources

• https://statsapi.mlb.com/api/v1/schedule?sportId=1&date=2026-09-26&hydrate=probablePitcher
• https://www.mlb.com/probable-pitchers/2026-09-26
• https://statsapi.mlb.com/api/v1/standings?leagueId=103,104&season=2026
• https://statsapi.mlb.com/api/v1/schedule?sportId=1&date=2026-09-25
• https://statsapi.mlb.com/api/v1/game/824058/boxscore (CLE@KC pitch counts)
• https://statsapi.mlb.com/api/v1/game/824947/boxscore (HOU@ATH pitch counts)
• https://statsapi.mlb.com/api/v1/game/822760/boxscore (CIN@TOR pitch counts)
• https://statsapi.mlb.com/api/v1/game/822759/boxscore (today's CIN@TOR lineups)
• https://statsapi.mlb.com/api/v1/people?personIds=695549 (career-starts check — wrong ID, see limitations)
• 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=20260926&limit=100
• https://site.api.espn.com/apis/site/v2/sports/baseball/mlb/scoreboard?dates=20260926
• https://site.api.espn.com/apis/site/v2/sports/baseball/mlb/summary?event=401817096
• ESPN DraftKings odds, one per game: events 401817090, 401817087, 401817093, 401817094, 401817095, 401817096, 401817097, 401817098, 401817099, 401817100, 401817101 (.../competitions/{id}/odds and .../odds?limit=50)
• https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=38.5804&lon=-121.5130 (Sutter Health Park)
• https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=41.8299&lon=-87.6338 (Rate Field)
• https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=39.05&lon=-94.48 (Kauffman Stadium)

Research limitations

The three counts: 11 of 13 games analysed · 33 distinct URLs fetched (27 returning usable data) · 2 games where a Tier-1 factor could not be obtained.

Playbooks read: mlb.md only. Every game on today's slate is MLB, so no stub playbook was used — NHL and NBA were not needed and neither sport had a bookable slate here.

The scheduler delay is the run's biggest limitation. It fired at 12:43 ET against an 11:13 schedule. Consequences: two games lost to the 30-minute cutoff, the 11:50 build cutoff missed outright, and the 12:00 deploy published stale content. rebuild_and_deploy.bat needs running manually.

🔴 A genuine source finding worth folding into the playbook — the ESPN odds endpoint needs ?limit=50. The bare .../competitions/{id}/odds URL returned {"count":0,...,"items":[]} for four games (401817087 ATL@MIA, 401817093 TB@PHI, 401817097 TEX@MIN, 401817100 HOU@ATH). Re-fetching the identical endpoint with ?limit=50 appended returned full DraftKings data for all four. These were first-time fetches, not cache artifacts. Without the retry I would have reported four games as unpriced and fallen below the odds-for-every-game requirement. §9 of mlb.md should specify ?limit=50 as the default form.

Tier-1 factors that could not be obtained:

• TB @ PHI and ARI @ SD — starters not announced on either statsapi or MLB.com. These are the two games counted above as Tier-1-unavailable, and both are H3 no-plays rather than analytical failures.
• Career-starts verification for H4 failed. I queried personIds=695549 expecting Jack Perkins and got Jackson Jobe — wrong ID, and I did not spend a second fetch correcting it. So I have no explicit gamesStarted figure for Perkins, Ralston, Wilkinson, Kade Anderson, Ryan Johnson or Quinn Mathews. For plays #2 and #4 I am relying on Perkins appearing on Savant's min=q qualified leaderboard, which implies a full-season workload well beyond three starts — reasonable, but it is inference, not the direct check H4 asks for. ATL @ MIA and LAA @ SEA were marked no-play specifically because I could not run this check on their rookie starters.
• Lineups: obtained for one game only (CIN @ TOR, the 3:07 PM start). §6 says to always confirm today's actual lineup, and for the 7:10 PM and 9:40 PM games nothing was posted at 12:50 ET. On the penultimate day of the season, with five clinched clubs likely to rest regulars, this is the single largest hole in today's card — the rest/call-up reasoning behind plays #2, #3 and #4 rests on standings position rather than a posted nine.
• Umpire assignments: not obtained. §9 lists this as unverified and §5 values it at 0.3-0.5 runs. Not attempted this run; it remains an open item in the evidence ledger.
• Savant coverage gaps. ABSENT from the min=q leaderboard: Blake Snell, Matt Wilkinson, Quinn Mathews, Justin Verlander, Jack Ralston, Kade Anderson, Ryan Johnson. Note a near-miss worth flagging: grepping for "Anderson" returns Drew Anderson, not Kade — do not read that row as the Mariners' starter. Per §9 I did not substitute FIP for any of them.

Freshness-check result: PASSED, on two independent sources. statsapi (12:44 ET, 13 games) and MLB.com's probable-pitchers page agree on all 13 matchups and all 24 named starters, and agree that the PHI and ARI starters are TBD. No stale-board symptoms.

Other source failures:

• site.api.espn.com/.../mlb/standings?season=2026 — returned only a "Full Standings" nav link, no data. statsapi standings was used instead and worked.
• The long field-filtered statsapi standings URL was rejected by the fetch proxy (HTTP 403, "url exceeds the maximum fetchable length"). The unfiltered URL worked fine.
• curl is unusable in this container this run — every attempt died with connect_rejected (the egress proxy denied the CONNECT). All data above came through WebFetch. This is new behaviour worth noting; any playbook step that assumes shell curl will fail.
• MET Norway is coordinate-cache sensitive. lat=39.0517&lon=-94.4803 (Kauffman, full precision) returned a timeseries covering 2026-08-05 to 2026-08-15 — stale August data, no September entries. Rounding to lat=39.05&lon=-94.48 returned a correct forecast starting 2026-09-26T16:00Z. If met.no returns dates that don't include today, re-request with rounded coordinates before declaring it failed.
• ESPN exposes DraftKings only (count: 1), as §9 documents. VegasInsider was not used. No prices came from memory; every quoted number came from an ESPN odds fetch this run.

One record note: no Main bet has been logged since 9/19 — a week with no graded plays, so the 3-1-0 record is a four-bet sample and the 75.0% hit rate should be read as such. The Dashboard also shows Over/Under at 0-1 all-time for main bets (−$41.26 on a single bet); that is a sample of one and not a reason to discount plays #3 and #4, but two of today's four ranked plays are totals and it is worth your knowing that this bankroll has no O/U track record behind it.

Calibration caveat, as the prompt requires: the 3-and-4-point deductions were measured on a mostly-prop sample and there is little history on game lines. Treat them as a starting assumption, not a fitted model.