Daily Bet Research Workflow (updated 2026-07-01)

Bankroll stake: 20% of available bankroll (AGGRESSIVE mode, rollback disabled per 2026-06-11 override) Odds band: Decimal 1.50+

The 7-step research process

Step 0: Confirm date & time of request (added 2026-07-23)

Before ANY analysis, explicitly confirm: - Today's date (day of week + full date) — never assume - Current time in ET — needed to know which games are still bookable - If user hasn't stated the time explicitly and it's ambiguous, ASK before running the report

Why: Timing confusion has caused multiple report errors — writing off Sunday day games as "off the board" when user pinged at 10am, mixing up Sunday vs Monday, misidentifying which games were still bookable. A 30-second confirmation prevents 30 minutes of wasted analysis on the wrong games.

Format the confirmation at top of every report:

Report generated at [TIME] ET on [DAY], [DATE]. Games with first pitch after [TIME+30min] are bookable.

1. Bankroll check — pull current balance, confirm stake at 20%, note record
2. Yesterday recap — settle open bets, note wins/losses/net impact
3. Schedule verification — pull today's full slate (MLB + WC when active), verify times and starters against Step 0 timing to determine what's actually playable
4. Per-game research — for candidate games only, gather confirmed starters (cross-referenced 2+ sources per 7/23 rule), ERAs, records, park factors, injury reports, betting lines
5. Edge math & pick selection — apply the two-lens framework below (3 options per lens)
6. Recommendation — present picks with confidence tier and stake sizing

Timing (added 2026-07-06, updated 2026-07-13)

Standard run window: 3-5 PM ET for MLB evening slates. Sunday day-game slates: 10 AM-12 PM ET ping (most Sunday games start 1-2 PM ET) WC afternoon kickoffs (3-4 PM ET): ping 10 AM-1 PM ET Tuesday/Wednesday

Data-driven basis (May 21 – July 6, 44 settled bets):

Evening reports use confirmed lineups, confirmed starters, weather locked in, sharp money already moved lines. Late-morning window (10am-noon) has been our worst — actively avoid.

User will ping when ready. Auto-run scheduled tasks are disabled.

Two-lens framework (added 2026-07-01)

Every daily report should present picks through two lenses:

Lens 1: Max-EV Play

The pick with the highest expected value. This is the long-run growth engine. Over 100+ bets, this lens wins. - Includes: single-leg edges, manufactured-boost SGPs, structural-edge plays - Typical hit rate: 50-65% - Typical edge: +5-20%

Lens 2: Max-Hit-Rate Alternative

The highest-probability positive-EV play available. This is the psychological engine — wins matter to avoid tilt, maintain confidence. - Usually the biggest favorite with any positive edge - Or a manufactured-boost SGP with near-lock amplifiers keeping combined hit rate above 55% - Typical hit rate: 60-80% - Typical edge: +2-10% - NEVER a chalk parlay — adding legs always reduces hit rate

When they converge vs. diverge

• Most nights they'll be the same pick or overlap heavily
• When they diverge (e.g. max-EV is a longshot SGP, max-hit-rate is a straight favorite), user chooses based on mood/streak/preference
• Both must have positive expected value — a max-hit-rate play with negative EV is not worth including

Probability estimation — triangulation rule (added 2026-07-01)

Never anchor on a single source. When estimating true probability, gather from at least 3 sources when available: - Opta supercomputer - Sportytrader / algorithm-based models - Kalshi implied prob (prediction market) - FiveThirtyEight (when active) - DraftKings network model - Expert consensus from CBS, Yahoo, ESPN

Take the median or thoughtful weighted average. Note the range in the report.

Recommend-against calls — quantify closeness (added 2026-07-01)

When advising against a play, quantify: - Clear pass (edge < -5%): "clear negative edge, skip" - Close call (edge -2% to -5%): "roughly break-even, do it if you want the swing" - Marginal (edge -0% to -2%): "coin flip on EV, hit-rate preference decides"

Don't tell user "don't do it" when the math is genuinely close — give them agency.

Manufactured-boost SGP template

When bet365's 30% profit boost is unavailable, manufacture equivalent value with near-lock amplifier legs: - Anchor: real-edge play (ML, Under, or Over) at 50-65% hit rate - Amplifiers: 1-2 near-lock props (3+ K, Under 13.5 runs, batter 1+ hit, etc.) at 90%+ hit rate - Combined boosted decimal typically 2.0-2.9 - Net effect: raises payout by 30-70% at cost of ~5-10% hit rate

Winning historical examples: - 6/29 Brewers ML + Under 13.5 + Gasser 3+ K @ 2.05 → WIN +$27.75 - 6/18 CLE/MIL Under SGP @ 3.05 → WIN +$48.75 - 6/19 CIN/NYY Under SGP @ 3.05 → WIN +$48.75

Reports to avoid

• Yankees ML the day after a Yankees loss — tilt-bet risk
• Coors Field games (MIA/COL etc.) — variance too high
• Heavy chalk WC knockouts blindly — post-6/29 (Germany lost, Netherlands lost) knockouts have shown real upset variance
• Any two-leg chalk parlay — reduces hit rate below either individual leg
• WC knockout ML on 90-min-only lines (added 2026-07-12) — Argentina beat Switzerland but drew 90 min, went to ET/PK. 90-min ML LOST despite Argentina advancing. Use "To Advance" or "Team ML including ET" markets for knockout stage.
• SGP amplifier upgrades without checking actual leg prices (added 2026-07-13) — Recommended moving Red Sox Team Total from O2.5 → O3.5 based on 76% true estimate; book was aggressively pricing at 87% implied. Upgrade LOWERED edge from +10% to +5%. Before recommending line upgrades, verify actual bet365 prices. If book implied > true estimate, DON'T upgrade — stick with tighter amplifier.
• Pitcher K prop estimates using only season K/9 average (added 2026-07-19) — Estimated Skubal P(9+K) at 25-30% using Poisson from 6.8 K/game average; book priced at 45-55% implied and was closer to right (Skubal actual 9 K). When bet365 offers "generous" pricing on a specific pitcher prop, they usually know something the raw season stats miss. Weight recent form (last 3-5 starts), opponent whiff rate, and park factor. If book's implied deviates significantly from raw-stat estimate, respect the market — don't over-lever on that leg.
• Stale pitcher confirmations from single-source expert articles (added 2026-07-23) — Reported Ian Seymour as Rays starter based on one NBC Sports article; actual starter was Casey Legumina (pitching profiles materially different). Rule: cross-reference at least 2 sources dated within 24 hours OR check ESPN/Baseball-Reference/MLB.com previews directly for same-day confirmations. Never rely on a single expert-analysis article for starter identity — expert articles often lag lineup changes by 12-24 hours.
• Line drift beyond model expectation is a sharp-money signal (added 2026-07-31) — Yankees Over 7.5 was at 2.00 at bet365 vs 1.91 (Dimers). I noted "sharp money possibly on Under" but treated it as minor concern. Game came in 3 total runs (2-1 in 11 innings) — both mediocre starters had career starts. When a total line drifts 5+ cents beyond the model in the "better for us" direction, treat that as a positive signal for THE OTHER SIDE, not just noise. Rule: if line is materially better than expected AND the other side is a defensible bet (Under with quality pitching potential, Over in offensive matchup), reduce stake or pass. The book knows something the public models don't. Cost us $40 on Bet #58.

Chase-prevention protocol (added 2026-07-20)

User self-flagged "chasing losses" on Bet #52 and explicitly requested: if this loses, don't let me chase again.

Rule: if a user acknowledges they're chasing, and that specific chase play loses, the next report MUST: 1. Flag "no-chasing" at the top of the report 2. Recommend the highest-hit-rate positive-EV play only (or rest day) 3. Explicitly push back on any speculative plays (SGPs with lower hit rates, longshot MLs) 4. Reference the previous chase acknowledgment as accountability

Not paternalistic — the user asked for it. Honor the request.

KPIs to track

• Overall record (W-L-P)
• Bankroll trajectory
• Hit rate (added 2026-07-01) — % of bets won, tracked separately from P&L
• Net P&L
• ROI (P&L / total staked)
• Max drawdown

Discontinued strategies

• WC side hustle — closed 2026-06-26 at +$0.13 P&L after 2-8 record on 10 parlays. Small-stake parlays weren't producing edge.
• Rollback rule — disabled per user override 2026-06-11. All plays at 20% AGGRESSIVE.