August 2026
Monthly bet strategy review
Monthly Bet Strategy Review — August 2026
Window: 2026-08-01 to 2026-08-31 (ET) · Settled bets: 36 (9-27-0)
Bottom line
August was a drawdown month — bankroll fell from $236.06 to $39.86 (-83%) — but the damage is concentrated almost entirely in the Issue Era (8/3-8/21), which is now closed. The Rebound Era (8/23 onward, post-process-rewrite) posted a materially better ROI (-12.5% vs -26.4%) on similar bet counts, though the sample (n=18) is still too small to call it a confirmed turnaround. The most actionable finding: this month's Pick Rank calibration data shows Option 2 has been dramatically overconfident (22% actual vs. 57% average stated hit rate) and Option 4 dramatically underconfident (78% actual vs. 58% stated) — which matters immediately because the new Kelly Weight staking column (added 9/2) trusts stated hit rate at face value, so as built it will systematically over-allocate to Option-2-type picks and under-allocate to Option-4-type picks until this is investigated.
Headline numbers
| Metric | Value |
|---|---|
| Bankroll start (7/31 close) | $236.06 |
| Bankroll end (8/31 close) | $39.86 |
| Bankroll change | -$196.20 (-83.1%) |
| Total wagered | $653.82 |
| Total P&L | -$196.20 |
| ROI | -30.0% |
| Win rate (decided bets) | 25.0% (9-27) |
Era comparison
| Era | Window (within Aug) | N | W-L | Hit % | Staked | P&L | ROI |
|---|---|---|---|---|---|---|---|
| Good Era | thru 8/2 | 1 | 0-1 | 0.0% | $47.21 | -$47.21 | -100.0% |
| Issue Era | 8/3–8/21 | 17 | 4-13 | 23.5% | $525.28 | -$138.85 | -26.4% |
| Rebound Era | 8/23–8/31 | 18 | 5-13 | 27.8% | $81.33 | -$10.14 | -12.5% |
Sample tag: Issue Era (n=17) and Rebound Era (n=18) are both Notable tier. The ROI gap (-26.4% vs -12.5%) is directionally consistent with the process changes made 8/23-8/24 (self-check step, narrowed Section 2 scope, Option 1 = straight/anchor convention), but at this sample size the confidence intervals on a ~25-28% hit rate still overlap — not yet a statistically confirmed turnaround, just a real and encouraging signal. Stakes also shrank sharply as bankroll fell (Good/Issue Era bets averaged ~$34/bet; Rebound Era ~$4.50/bet), which is why Rebound Era's dollar P&L looks small in absolute terms even though the underlying ROI improved.
Calibration check
Since 2026-08-23 (the date explicit OptN:W/L tagging actually began — one day earlier than workflow.md's documented "since 8/24," a minor doc fix noted below), every settled bet's Notes column was matched against that date's daily report Section 3 stated Hit Rate for the same rank.
| Rank | N | Actual W-L | Actual Hit % | Avg Stated Hit % | Gap | Note |
|---|---|---|---|---|---|---|
| 1 | 9 | 5-4 | 55.6% | 62.4% | -6.9 pts | Within tolerance |
| 2 | 9 | 2-7 | 22.2% | 57.4% | -35.2 pts | Flagged — badly overconfident |
| 3 | 9 | 6-3 | 66.7% | 59.5% | +7.2 pts | Within tolerance |
| 4 | 9 | 7-2 | 77.8% | 57.9% | +19.8 pts | Flagged — badly underconfident |
| 5 | 9 | 5-4 | 55.6% | 56.6% | -1.0 pt | Well calibrated |
Excluded (untagged, pre-convention, dated before 8/23): 18 bets.
Sample tag: Suggestive (n=9 per rank) — real gaps, but each rank's confidence interval at n=9 is wide (a 22% or 78% observed rate could plausibly regress toward the 55-60% region other ranks show with more data). Don't fully re-trust the model yet; do start watching Option 2 and Option 4 specifically next month.
Performance breakdowns
By Category
| Category | N | W-L | Hit % | Staked | P&L | ROI |
|---|---|---|---|---|---|---|
| Main | 24 | 8-16 | 33.3% | $623.72 | -$194.09 | -31.1% |
| Fun Parlay | 12 | 1-11 | 8.3% | $30.10 | -$2.11 | -7.0% |
By Type
| Type | N | W-L | Hit % | Staked | P&L | ROI |
|---|---|---|---|---|---|---|
| SGP | 16 | 4-12 | 25.0% | $533.63 | -$147.20 | -27.6% |
| Parlay | 12 | 1-11 | 8.3% | $30.10 | -$2.11 | -7.0% |
| ML | 7 | 4-3 | 57.1% | $85.79 | -$42.59 | -49.6% |
| Total | 1 | 0-1 | 0.0% | $4.30 | -$4.30 | -100.0% |
Note on ML's odd-looking line (57.1% hit rate but -49.6% ROI): this is a sequencing artifact, not a red flag — the 3 ML losses landed earlier in the month on a larger bankroll (bigger $ stakes), while the 4 ML wins landed later on a much-shrunk bankroll (small $ stakes), so the dollar P&L doesn't track the win rate. Worth remembering as bankroll rebuilds: win rate and dollar ROI can diverge sharply during a drawdown-then-recovery stretch.
Findings
- SGP bled the most in dollar terms (-$147.20, Notable tier n=16) and carried the Issue Era's damage — 12 of its 16 bets landed in the Issue/Good Era window, before the Option-1-anchor convention simplified Main bets to mostly straight ML/Total plays.
- Fun Parlay stayed nearly break-even (-$2.11 on $30.10 staked, Notable tier n=12) despite an 8.3% hit rate — the $2 flat-stake sizing is doing its job of capping downside on a structurally low-hit-rate bet type (4-leg parlays).
- Rank #1 is not meaningfully outperforming Ranks #2-5. Rank 1 hit 55.6% (n=9); Ranks 2-5 combined hit 55.6% (20-16, n=36) — identical. The daily report's "pick the single best-EV/hit-rate convergence play as Option 1" selection isn't showing an edge over the average of Options 2-5 yet. Suggestive tier (n=9 for Rank 1 specifically) — worth re-checking at a larger sample, but if this holds, it questions whether Option 1 deserves its position as the anchor for the (now larger) Main bet.
- Option 2 is badly overconfident, Option 3/4 are underconfident — see Calibration check above. This is the month's most concrete, actionable finding, and it directly affects the new Kelly Weight staking system (see Proposed methodology updates below).
- Rebound Era ROI improvement is real but not yet statistically confirmed (-12.5% vs -26.4%, both Notable tier, overlapping confidence intervals at this sample size).
- Report rigor spot-check passed. Cross-checked Section 4 math derivations and Sources lists across the 8/23-8/31 window (already reviewed in depth this cycle, e.g. 8/31's report shows full Implied%/true-estimate/Edge%/EV derivation for every pick with real, checkable sources) — no reversion to the Issue Era's assert-without-deriving pattern.
Proposed methodology updates
- Investigate Option 2's calibration gap before trusting Kelly Weight staking on it. Kelly Weight (added 9/2 to the three daily prompts) computes weight directly from each day's stated Hit Rate. If Option 2 is structurally overconfident (22% actual vs 57% stated, this month), Kelly Weight will over-allocate budget to it until the underlying report-generation bias is found and fixed. Proposed: add a note to the three daily prompts' Kelly Weight section flagging Options 2 and 4 as under active calibration review, and re-run this calibration check monthly with a lower flag threshold (>15 pts) specifically for Ranks 2 and 4 until the gap closes or the pattern is confirmed structural. (awaiting approval)
- Re-examine whether "Option 1 = Main anchor" selection logic actually beats an unranked pool. Rank 1 hit 55.6% this month, statistically identical to the Ranks 2-5 pool. Proposed: next month, additionally track a simple "would a randomly-chosen pick from the top 5 have done as well as Option 1" comparison, to see if this holds at a larger sample before considering any change to which option anchors the Main bet. (awaiting approval)
- Fix workflow.md's documented tagging start date. It currently says the
OptN:W/Lconvention started 8/24; the Log shows the first tags on 8/23 (Bet #78). Proposed: update workflow.md's dated note from "since 8/24" to "since 8/23." Cosmetic, no behavior change. (awaiting approval)
How to approve
Reply in the pinned chat with, e.g., "apply 1 and 3, skip 2" or "apply all" or "skip all."
Data source
computer://C:\Users\mitch\Documents\BettingTracker\bet_tracker.xlsx