Daily Bet Research Report — Wednesday, September 16, 2026

Generated 3:55 PM ET

Bankroll: $55.67 | Stake (12.5%, one play): $6.96 Record: 1-0-0 (Main bets, main leg, since 9/14 reset) | Hit rate: 100.0% Slate: 15 games across MLB · 10 shortlisted · Verified against MLB StatsAPI at 3:47 PM ET

⚠ THIS RUN FIRED LATE. The scheduled 11:00 AM ET trigger did not execute; the run began at 3:45 PM ET, nearly five hours late and long past both the 11:45 AM research cutoff and the 12:00 PM site deploy. The noon deploy therefore published yesterday's report. Everything below was researched between 3:45 and 4:05 PM ET against a live board, and the four early games (1:10, 1:15, 1:40, 3:07 PM ET) were already underway and are excluded as unbookable. See Research Limitations.

Section 1: Verification

Verified against MLB StatsAPI (schedule?sportId=1&date=2026-09-16&hydrate=probablePitcher) at 3:47 PM ET — 15 games.

Cross-checks run against the league schedule before any pricing:

• ESPN core API (NFL, NCAAF, NHL): all returned count: 0 for 20260916. MLB is the only sport in season today. No NFL (Wed), no college football (Wed), no NHL (preseason not started), no NBA/NCAAB.
• VegasInsider MLB board: contents intersect today's participant list on all 11 remaining games, and the listed starters match today's probables on 9 of 11 (Messick, Liberatore, Rodón, Matthews, Scherzer, Snell, Abbott, Henderson, Jones, Wheeler, Cavalli, Bassitt, Stock, Imanaga, Bennett, Javier, Ray, Kirby, Kikuchi, Kelly all confirmed on both sources). Freshness check PASSED — priced from.
• Two starter conflicts flagged, neither priced: VegasInsider lists Smith-Shawver for Atlanta where StatsAPI lists JR Ritchie; VegasInsider lists deGrom for Texas where StatsAPI lists MacKenzie Gore. Per the MLB framework ("confirm starters on 2 sources"), ATL @ CHC and BOS @ TEX are carried on the shortlist as rejects only and would not have been recommended even on positive edge.

Excluded as unbookable (inside/past first pitch at run time): CWS @ CLE (1:10), SF @ STL (1:15), NYY @ MIN (1:40), DET @ TOR (3:07) — 4 games. Excluded by rule: SD @ COL (8:40) — Coors Field. Screened: 10 games.

Section 2: Shortlist

Games left unscreened: 5 of 15 — four (CWS @ CLE, SF @ STL, NYY @ MIN, DET @ TOR) had already started by the time this late run began and are unbookable; one (SD @ COL) is excluded by the standing Coors Field rule. All 10 remaining games were screened on the moneyline across both books; three of those ten (LAD @ CIN, PHI @ WSH, BOS @ TEX) were additionally priced on spread and total.

Section 3: Recommendations

No qualifying plays today — 10 games screened, no positive edge after calibration.

Allocate nothing.

Hit Rate is discounted for measured overconfidence; raw estimates in Section 4.

Section 4: Reasoning

Zero-play day. Below are the five closest candidates and what disqualified each, followed by the two structural findings that shaped the result.

Closest candidate 1 — New York Mets ML vs BAL, bet365 1.71 (−140)

The best number on the entire board, and still not close.

• Consensus: BAL +123 / NYM −148 → raw implied 44.84% / 59.68%, sum 104.52% → de-vigged fair value NYM 57.10%.
• Best price: bet365 −140 = decimal 1.71, implied 58.33%. DraftKings is worse at −148 (1.68, 59.52% implied), so bet365 is the shop winner by 8 cents.
• Raw estimate 57.10% → discounted 54.10% (odds < 1.80, subtract 3 points).
• Edge = 54.10 − 58.33 = −4.24%. EV = (0.5410 × 1.71) − 1 = −7.26%.

Even before the calibration discount this side is −1.24% against its own price — bet365 is 8c better than DraftKings but still on the wrong side of fair. Framework applied (MLB, starting pitcher dominates): Robert Stock starting for a 69-81 Mets club against Chris Bassitt and a 73-78 Baltimore team is not a matchup that justifies laying −140 in the first place; the market is pricing home field and bullpen, not the starter. Nothing here argues the true number is 4+ points better than the field's consensus. Risk factor: Stock is an unusual name to see in a September rotation slot, which normally signals an opener or bulk-relief plan — that widens the outcome distribution and argues for less confidence, not more.

Closest candidate 2 — Houston Astros ML vs KC, bet365 1.71 (−140)

Consensus KC +119 / HOU −144 → fair HOU 56.38%. bet365 −140 (1.71, 58.33% implied) beats DraftKings −144 (1.69). Discounted 53.38%. Edge −4.95%, EV −8.49%. Cristian Javier vs Daniel Lynch IV, two .440–.500 clubs; consensus is tight and no book is meaningfully off the field.

Closest candidate 3 — Chicago Cubs ML vs ATL, bet365 1.71 (−140)

Consensus ATL +119 / CHC −143 → fair CHC 56.31%, discounted 53.31%, vs 58.33% implied. Edge −5.03%, EV −8.61%. Disqualified twice over: negative edge, and the two sources disagree on Atlanta's starter (StatsAPI: JR Ritchie; VegasInsider: Smith-Shawver). Imanaga at Wrigley against an unconfirmed Braves arm is not a spot to guess.

Closest candidate 4 — Texas Rangers ML vs BOS, bet365 1.91 (−110)

Consensus BOS −105 / TEX −115 → fair TEX 51.08%, discounted 47.08% (odds in the 1.80–2.10 band, subtract 4), vs 52.38% implied at bet365's −110. Edge −5.30%, EV −10.11%. Same starter conflict problem: StatsAPI says MacKenzie Gore, VegasInsider says deGrom — a difference that would move this line 15–20 cents. Pass on principle regardless of price.

Closest candidate 5 — Texas Rangers +1.5 (run line), DraftKings 1.53 (−187)

ESPN's DraftKings record: BOS −1.5 at +154 (2.54), TEX +1.5 at −187 (1.53). De-vigged: TEX +1.5 fair 62.41%, discounted 59.41%, vs 65.36% implied. Edge −5.95%, EV −9.11%. Line moved from −205 to −187 on the Texas side and the total from 7.5 to 8.0 since open — drift toward, not away from, the favorite side of the total. Note the run line is a fixed 1.5 runs and is distorted by the frequency of one-run games; that distortion is already in the price here.

Structural finding 1 — the 2.10 ceiling removed the only real book outlier

The one genuine price dislocation on today's board is bet365 posting +120 on three separate underdogs (Pittsburgh, Washington, LA Angels) where the seven-book consensus is +112 / +113. That is a real 7–8 cent edge over the field, exactly the kind of outlier worth hunting. All three sit at decimal 2.20 — above the hard 2.10 ceiling — so none is bettable. This is the ceiling working as designed: the band was set because estimates collapsed above 2.10 (5 wins from 16 picks), and a 7-cent price edge does not repair a probability estimate that is unreliable at that price. Recording it so the pattern is visible if it recurs.

Everywhere else the two books were within 3–5 cents of each other and of consensus. On the favorite side, bet365 was the better price in 6 of 10 games; DraftKings in 4. Neither was ever far enough off the field to generate edge.

Structural finding 2 — spreads and totals cannot currently produce a positive edge

This is worth flagging explicitly because it is a method limitation, not a property of today's slate.

Fair value is defined as de-vigged consensus. For moneylines, consensus is real: eight books are readable on VegasInsider. For spreads and totals, only one book (DraftKings, via the ESPN core API) is readable. De-vigging a single book's two sides returns that book's own no-vig line. The edge of any side against its own de-vigged price is therefore mathematically bounded at −(half the vig), before the calibration discount is applied at all. Every one of the six spread/total sides priced today landed between −5.6% and −6.5%, clustered exactly where that arithmetic predicts.

In other words: under the current source set, no spread or total can ever be recommended, regardless of the game. That is not a good reason to relax the discount — it is a reason to add a second readable source for spreads and totals, or to accept that these two markets are research-only until one exists. Flagging for the strategy review rather than working around it.

Calibration caveat (carried over, as required)

The 3-point and 4-point discounts were measured across 70 graded picks that were overwhelmingly player props. There is still effectively no calibration history on moneylines, spreads or totals — the post-9/14 sample is a single graded leg. The discounts applied to every line above are a starting assumption carried over from prop data, not a measured property of game lines. They should be re-measured once roughly 30 game-line picks have been graded. Until then, a day like today — where the best side on the board misses by 4.24 points and the discount accounts for 3 of them — is genuinely sensitive to that assumption. Even so: at the raw, undiscounted estimate the best play on the board is still −1.24%, so today is a no-play on the undiscounted numbers as well. The conclusion does not depend on the caveat.

Sources

• https://statsapi.mlb.com/api/v1/schedule?sportId=1&date=2026-09-16&hydrate=probablePitcher
• https://www.vegasinsider.com/mlb/odds/las-vegas/
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events?dates=20260916&limit=100
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/401816962/competitions/401816962/odds (LAD @ CIN)
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/401816964/competitions/401816964/odds (PHI @ WSH)
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/401816966/competitions/401816966/odds (BOS @ TEX)
• https://sports.core.api.espn.com/v2/sports/football/leagues/nfl/events?dates=20260916&limit=100
• https://sports.core.api.espn.com/v2/sports/football/leagues/college-football/events?dates=20260916&limit=100
• https://sports.core.api.espn.com/v2/sports/hockey/leagues/nhl/events?dates=20260916&limit=100

Research limitations

1. The run fired ~4h45m late. The 11:00 AM ET trigger did not fire; work began at 3:45 PM ET. The 11:45 AM research cutoff and the 12:00 PM deploy had both already passed, so the noon deploy published yesterday's report. The site has been rebuilt after this report was written, but it is late. Four games that would normally have been in scope (CWS @ CLE, SF @ STL, NYY @ MIN, DET @ TOR) had already started and were dropped as unbookable — a real loss of slate coverage attributable entirely to the late fire, not to research choices. Worth checking why the schedule did not trigger.

2. Spreads and totals have only one readable book. As detailed in Section 4, this makes "de-vigged consensus" collapse to de-vigged DraftKings and bounds every spread/total edge at negative before the discount is even applied. This is a standing structural limitation, not a today problem.

3. Two-source starter conflicts on two games. ATL @ CHC (StatsAPI: JR Ritchie / VegasInsider: Smith-Shawver) and BOS @ TEX (StatsAPI: MacKenzie Gore / VegasInsider: deGrom). Neither was resolvable in the time available. Both games were carried as shortlist rejects only and neither would have been recommended.

4. Unresolved price conflict on PHI @ WSH. VegasInsider shows all eight books at PHI −135 to −140 (consensus −136). The ESPN DraftKings record for the same game shows PHI −198, with the total moved 7.5 → 9.0 and the Phillies run line from +129 to −115 — internally consistent within the ESPN record, but ~60 cents off the entire VegasInsider board. A search for a Cade Cavalli scratch found no such report; he is listed as Washington's starter on both sources. Neither source was priced into a recommendation for this game. Note that this conflict could not have produced a play either way: at −198 Philadelphia is a worse price than the one already rejected, and Washington at +162 (2.62) is above the 2.10 ceiling.

5. No source failed to return data, and nothing was abandoned for truncation. The VegasInsider fetch exceeded the token limit and was read in full from the saved file via Grep and Read; the full eight-book moneyline board for all 11 bookable games was recovered and priced.

Star scale (on Edge %)

★★★★★ +10%+ · ★★★★ +5-10% · ★★★ +2-5% · ★★ 0-2% · ★ negative