Daily Bet Research Report — Friday, September 18, 2026

Generated 11:40 AM ET

Bankroll: $61.03 | Stake (12.5%, one play): $7.63 Record: 2-0-0 (Main bets, main leg, since 9/14 reset) | Hit rate: 100.0% Slate: 15 games, all MLB · 8 analysed · Playbooks: mlb.md (v1) · Verified at 11:12 AM ET

Playbooks read: mlb.md only. NFL has no games today (Week 3 Thursday game was 9/17); NHL and NBA are out of season. nfl.md, nhl.md and nba.md are stubs and were not needed.

⚠ Freshness check — one source failed, and it failed badly

Verified against statsapi.mlb.com/api/v1/schedule at 11:12 AM ET — 15 games, all night starts (earliest 6:40 PM ET). Nothing on the card is inside 30 minutes.

VegasInsider's Las Vegas MLB board is serving a different day's slate and has been discarded in full. It listed Brewers–Pirates, Dodgers–Reds, Athletics–Rays, Padres–Rockies and Angels–Twins. None of those are today's matchups (today: Brewers @ Orioles, Giants @ Dodgers, Athletics @ Guardians, Marlins @ Padres, Twins @ Angels). The pitchers attached to it — Harrison, Dotel, Wrobleski, Rasmussen, King, Urena — match none of today's probables either. This is precisely the 9/15 failure mode, caught this time before anything was written on it.

Consequence: no bet365 prices today. Every number below is DraftKings via the ESPN core API, which cross-checked clean against the statsapi participants (team IDs verified game by game). It also means Hard Rule H3 (both starters confirmed on two independent sources) is satisfied on the league's own ground-truth feed but not on a second one — see Research Limitations. Confidence figures are trimmed for it.

Game Analysis

Kansas City Royals @ Pittsburgh Pirates — 6:40 PM ET, PNC Park Key inputs: Paul Skenes (3.78 ERA, 162.0 IP, 10.28 K/9, 2.56 BB/9, 0.94 HR/9, 1.13 WHIP) vs Randy Dobnak (2.73 ERA, 69.1 IP, 6.10 K/9, 2.99 BB/9, 1.27 WHIP, .249 BAA). Component ERA estimates from the season lines: Skenes ≈ 3.29, Dobnak ≈ 4.00. Line: PIT −193 (1.51), opened −173 (1.57). KC +158 (2.58), out of band. Run line PIT −1.5 at +113 (2.13), opened +119. Total 8, unchanged, over −112 / under −107. Read: Dobnak is the clearest mispricing on the board and it points against his own team. A 2.73 ERA sitting 1.27 runs below his component estimate, on a 1.27 WHIP and 6.10 K/9, is the exact ERA-lower-than-xFIP profile §2 flags as a fade — and 1.27 is past the 1.25 "headline of the writeup" threshold. Skenes is the mirror in miniature: 3.78 ERA against a ~3.29 component line, a 0.49 gap, mildly unlucky, buy side. The problem is that the market got here first — PIT shortened 20 cents from open, so this is a read I am late to, not one I found. Playbook §7 explicitly bars converting a starter-quality case into a run line, so PIT −1.5 at 2.13 is off the table despite being the better price. Play is PIT ML, priced thin.

Boston Red Sox @ Tampa Bay Rays — 7:10 PM ET, Tropicana Field (dome) Key inputs: Ranger Suarez (3.33 ERA, 137.2 IP, 8.50 K/9, 2.62 BB/9, 0.59 HR/9) vs Ian Seymour (4.27 ERA, 122.1 IP, 11.04 K/9, 2.50 BB/9, 1.69 HR/9, 23 HR allowed, 1.14 WHIP). Seymour's home runs are the story: 23 HR against 130 air outs is roughly a 15% HR/FB rate, well north of the ~11% norm. Normalising that to league rate drops his component ERA to ≈3.44 against a 4.27 actual — a 0.83-run gap, past §2's 0.75 flag. Suarez runs the opposite way: a 0.59 HR/9 is unsustainably low, so his true level is worse than his 3.33. Line: TB −137 (1.72), opened −131 (1.76) — five cents toward Tampa, inside tolerance. BOS +114 (2.14). Total dropped a full run, 8 → 7. Read: The better pitcher today is the one with the worse ERA, and the market is pricing Tampa off the record (92-59) rather than off Seymour. A full-run collapse in the total says the market does respect the pitching here even if the moneyline hasn't fully caught up. Dome, so no weather input needed. Real risk: Seymour has 44 appearances against 14 starts — he is a swingman on a short leash, and the Rays pen decides this game by the fifth. This is the play I like most on substance.

San Francisco Giants @ Los Angeles Dodgers — 10:15 PM ET, Dodger Stadium Key inputs: Tyler Glasnow (3.50 ERA, 61.2 IP, 11.24 K/9, 0.97 WHIP, .181 BAA; component ≈2.99, so mildly unlucky) vs Cesar Perdomo — MLB debut 2026-09-01, 3 appearances, 2 career starts. 14.2 IP, 9.82 H/9, .286 BAA, 1.36 WHIP, 6.75 K/9. Line: LAD −324 (1.30), out of band. SF +255, out of band. LAD −1.5 at −137 (1.72), opened −131 (1.76). Total 8. Read: H4 applies — Perdomo has two career starts, so the only permissible play is one built explicitly against his inexperience, and this is that play. It also clears the §7 run-line test in a way the Pittsburgh case does not: the argument is offensive mismatch (a 92-60 Dodgers lineup against a rookie lefty being hit at 9.82 per nine) plus a thin pen behind him on a 64-89 club, not starter quality. Glasnow suppresses the counter-scoring that kills a −1.5. Risk is the ordinary one — great pitching on the favourite's side produces 3-1 wins as happily as 8-1 ones, and the line already moved toward LA.

Washington Nationals @ St. Louis Cardinals — 8:15 PM ET, Busch Stadium Key inputs: Cade Cavalli (3.12 ERA, 164.2 IP, 10.60 K/9, 2.62 BB/9, 1.17 WHIP) vs Kyle Leahy (3.33 ERA, 135.0 IP, 8.53 K/9, 1.31 WHIP, .266 BAA). Component estimates 3.38 and 3.56 — both running slightly ahead of their skill, Leahy more so; a 1.31 WHIP and .266 opponent average producing a 3.33 ERA is strand-rate work. Line: STL −117 (1.85), WSH −102 (1.98). Washington opened −105, drifted three cents. Total 7.5 flat, over −118. Read: The clearly better starter is on the underdog side of a near-pick'em, and the market has barely moved all morning. Cavalli's 10.60 K/9 against Leahy's 8.53, with a 14-point WHIP edge, is a real if modest gap. It is modest, though — both are mildly lucky, the price is honest, and there is no second signal underneath it. Playable, not strong.

Seattle Mariners @ Colorado Rockies — 8:10 PM ET, Coors Field No play — Hard Rule H1. Coors Field is a house no-play regardless of the matchup. Not analysed further.

Minnesota Twins @ Los Angeles Angels — 9:38 PM ET, Angel Stadium Key inputs: Connor Prielipp vs Grayson Rodriguez. Both listed on statsapi. Read: Set aside on H4. Prielipp is a 2026 call-up and I could not confirm inside the research window that he clears the three-career-start threshold. The rule's default is no-play unless the case is built explicitly against his team, and I have no independent read on the Angels' offense to build that on. Passing rather than guessing.

Chicago Cubs @ Cincinnati Reds — 6:40 PM ET, Great American Ball Park Key inputs: Chase Burns listed for Cincinnati; Chicago's starter is TBD on statsapi as of 11:12 AM. Read: No play — H3. Half the dominant input is missing. Great American is a strong hitters' park and the Cubs (85-68) are still playing for position, so there is probably a total here worth reading once the Cubs name someone, but not from this desk at this hour.

Milwaukee Brewers @ Baltimore Orioles — 7:05 PM ET, Camden Yards Key inputs: Dustin May for Milwaukee (95-57, best record in baseball); Baltimore TBD. Read: No play — H3. Worth flagging for tomorrow: Milwaukee at a neutral-to-hitter park with a named starter against an unnamed one is the shape of game where a stale line can sit, but an outdoor total at Camden also needs wind I did not have time to pull (§4). Passing.

Also on the card, not analysed: Athletics @ Guardians, Phillies @ Mets, Tigers @ White Sox, Blue Jays @ Rangers, Braves @ Astros, Marlins @ Padres, Yankees @ Diamondbacks. Six of those seven have a TBD starter on statsapi and fail H3 outright; Marlins @ Padres (Phillips vs Pivetta) and Yankees @ Diamondbacks (Cole vs E. Rodriguez) are both fully named and were dropped only for time against the 11:45 cutoff.

Today's Plays

• #1 Pirates ML — Dobnak's 2.73 ERA sits 1.27 runs below his component line; that is the biggest single regression flag on the board. ⚠ Negative edge (−3.2%) — the market already moved 20 cents to PIT and the price is thin. Highest calibrated confidence, worst price.
• #2 Rays ML — Seymour's 4.27 ERA masks a ~3.44 component line once his 15% HR/FB is normalised; the only positive-edge play today.
• #3 Dodgers −1.5 — an offensive-mismatch run line against a two-start rookie, which is the one construction §7 permits. ⚠ Negative edge (−1.1%).
• #4 Nationals ML — the better starter on the dog side of a pick'em. ⚠ Negative edge (−0.5%), and the thinnest case of the four.

Reasoning

#1 — Pittsburgh Pirates ML, 1.51. This rests on §2, [STRONG] (xFIP/SIERA beat ERA as predictors) with the gap thresholds at [MODERATE]. Dobnak's actual ERA of 2.73 against a component estimate near 4.00 is a 1.27-run gap in the fade direction, and the underlying line — 6.10 K/9, 2.99 BB/9, 1.27 WHIP, .249 BAA — offers nothing to suggest the run prevention is real. Skenes runs the gap the other way by about half a run, so both signals point the same direction. Line movement works against me: PIT went −173 → −193, so the market reached this conclusion overnight and I am buying after the move, which is exactly what §3's movement rule says should cost confidence rather than be waved through. The risk that beats me is Skenes' own line — 3.78 ERA and 17 home runs allowed in 162 innings is good, not untouchable, and at 1.51 I need him to be roughly two-thirds certain. Source: statsapi.mlb.com/api/v1/people season pitching splits, opened this run.

#2 — Tampa Bay Rays ML, 1.72. Also §2 [STRONG], and this is the cleaner version because the market has not moved on it. Seymour's 1.69 HR/9 on 23 home runs against 130 air outs implies a HR/FB rate around 15%; that is the single most regression-prone rate statistic in pitching, and normalising it drops his component ERA from 4.07 to roughly 3.44 against a 4.27 actual — a 0.83-run gap, past the flag threshold. Underneath it sits an 11.04 K/9 against a 2.50 BB/9, which is genuine skill, not sequencing. Suarez is the inverse at 0.59 HR/9 — unsustainably low, so his 3.33 flatters him. Movement: TB −131 → −137, five cents, well inside the ~5-cent tolerance; the total's 8 → 7 collapse says the market respects the pitching without having repriced the side. The risk that beats me is usage, not talent — Seymour has made 14 starts in 44 appearances, and a §3 short-leash exit by the fifth hands the game to a bullpen whose state I could not check today. Source: same statsapi pitching splits; odds from sports.core.api.espn.com/.../401816983/odds.

#3 — Dodgers −1.5, 1.72. This one is governed by H4 and by §7's run-line rule. Perdomo debuted on September 1 with two career starts, so H4 permits a play only if it is explicitly against his team and built on that inexperience — which is what this is. §7 then demands a run line be justified by offensive mismatch and bullpen weakness rather than starter quality, and that fits: 14.2 major-league innings at 9.82 hits per nine and a .286 opponent average, against a 92-60 lineup at home, with a 64-89 Giants bullpen behind him once he exits. Glasnow's 0.97 WHIP and .181 BAA are what make the margin plausible rather than just the win. Movement: −131 → −137 toward LA, so again slightly late. The risk is the standard run-line risk and it is real — Glasnow at 11.24 K/9 produces plenty of 3-1 games, and a 3-1 win cashes the moneyline I cannot bet and loses this one. Source: statsapi (Perdomo mlbDebutDate 2026-09-01, 2 GS), ESPN odds endpoint for event 401816996.

#4 — Washington Nationals ML, 1.98. Weakest case of the four and I would rather say so than dress it up. It rests on §2 [STRONG] but with small numbers: Cavalli's 10.60 K/9 and 1.17 WHIP against Leahy's 8.53 and 1.31, with Leahy's .266 BAA and 1.31 WHIP producing a 3.33 ERA that his components (≈3.56) do not support. Both pitchers are running slightly hot, so this is a relative edge, not an absolute one. Movement: essentially none — WSH −105 → −102 overnight, which means no one else has found anything here either, and that cuts both ways. The risk that beats me is that this is simply a coin flip correctly priced as a coin flip, and a 1.98 line on a 71-82 road team is the market being right. Source: statsapi pitching splits; ESPN odds endpoint for event 401816991.

Sources

• https://statsapi.mlb.com/api/v1/schedule?sportId=1&date=2026-09-18&hydrate=probablePitcher — slate ground truth, 15 games
• https://statsapi.mlb.com/api/v1/people?personIds=694973,677976,624133,693855&hydrate=stats(type=season,group=pitching) — Skenes, Dobnak, Suarez, Seymour
• https://statsapi.mlb.com/api/v1/people?personIds=607192,702885,676917,681517&hydrate=stats(type=season,group=pitching) — Glasnow, Perdomo, Cavalli, Leahy
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events?dates=20260918&limit=100 — ESPN event ID map
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/401816985/competitions/401816985/odds — KC @ PIT (DraftKings)
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/401816983/competitions/401816983/odds — BOS @ TB
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/401816991/competitions/401816991/odds — WSH @ STL
• https://sports.core.api.espn.com/v2/sports/baseball/leagues/mlb/events/401816996/competitions/401816996/odds — SF @ LAD
• https://www.vegasinsider.com/mlb/odds/las-vegas/ — fetched, read in full, and rejected as stale (see below)
• https://site.api.espn.com/apis/site/v2/sports/baseball/mlb/scoreboard?dates=20260918 — fetched; contains no odds block, used only for event IDs

Research limitations

1. VegasInsider failed the Step 1 freshness check and was discarded. Its board carried five matchups not on today's slate and six starting pitchers not pitching today. Nothing from it appears above. This removes bet365 from today's report entirely — every price quoted is DraftKings. The playbook lists this source as verified-working; it should be re-tested tomorrow and downgraded if it fails again.

2. H3 is satisfied on one source, not two. statsapi's probablePitcher is the league's own feed and is the ground truth the skill instructs me to validate everything else against — but the intended second source is the one that failed, and I did not have cutoff time to pull ESPN's per-event probables for four separate games as a replacement. Confidence figures carry a trim for this. Worth eyeballing the lineup card before placing.

3. xFIP and SIERA were not available. FanGraphs leaderboards are listed in the playbook as not yet verified and I did not have the research window to test them. Every "component estimate" above is a FIP calculation I ran by hand from the statsapi season line, with a manual HR/FB normalisation for Seymour and a directional note for Suarez. These are my arithmetic, not a published xFIP, and the Seymour normalisation in particular uses air outs as a fly-ball proxy, which overstates the denominator. The direction of each gap is solid; the magnitudes are approximate.

4. No bullpen usage data — §3 went entirely unchecked. This is the playbook's designated most-under-priced factor and I have nothing on it today. It matters most for the Tampa play, where Seymour's swingman profile makes an early exit likely.

5. No weather, no park factors, no umpire assignments. All three are listed as unverified sources in the playbook and none was attempted. Consequence: no totals are recommended today, including the two that looked interesting (KC @ PIT Under 8 at PNC, and the BOS @ TB number that fell a full run). Recommending an outdoor total without wind would violate §4's order of operations in spirit and H2 in letter at Wrigley.

6. Seven games were not analysed. Six of them (Athletics @ Guardians, Phillies @ Mets, Tigers @ White Sox, Blue Jays @ Rangers, Braves @ Astros, and the Cubs and Orioles halves of their matchups) have a TBD starter on statsapi and fail H3 outright. Marlins @ Padres and Yankees @ Diamondbacks are fully named and were dropped for time only — both are legitimately unanalysed rather than unanalysable.

7. Calibration remains largely unevidenced on game lines. The 3- and 4-point subtractions come from a mostly-prop sample. Today's card is four game lines and the post-reset record is 2-0 — there is no meaningful calibration history behind these numbers yet.

8. Cutoff: research stopped at 11:40 AM ET to make the 11:45 build deadline. Points 4, 5 and 6 are all consequences of that window. No fetch was abandoned for truncation.