Bets · edges ·
blowout-correlation
🎯 Blowout Correlation Edge — validation framework
The strongest pattern we've found in the backtests, broken out as its own validated edge. 4 trigger conditions, 3-phase validation plan, KV-backed bet tracking. Currently UNPROVEN — we need 30-50 attempts to know if real.
📌 The honest framing
We found something interesting in the backtest data. Blowout Correlation Edge (BCE) is a real mechanism — books price correlated outcomes as independent. The pattern hit 5/5 in our small sample. This isn't proven yet. We need 30-50 attempts to know if it's real edge or backtest noise. But it's the most promising thing we've found. The math: $10 bets, 10-15% hit rate, 100:1 average payout = +$5,000-15,000/yr if pattern holds. -$1,500/yr if it doesn't. Worth testing? Sam decides.
1. The hypothesis — plain language
When specific conditions align in NBA games, three outcomes happen together that books price independently:
- Favorite covers the spread
- Game total goes over
- Favorite's star player hits over on points/stats
Books price each leg as if independent. But these outcomes are correlated — all happen because favorite blew out dog. This mispricing creates a parlay edge of estimated 5-15% over the book line.
The evidence
In our backtest of 20 NBA playoff games, 5 matched this pattern. All 5 hit the 3-leg correlated parlay. That's 100% hit rate on a 25% sample (5/20 games triggered the conditions).
NYK vs PHI G1 (5/4) · NYK 137-98 · NYK -15.5 + over 220 + Brunson 30+ · ~100:1
SAS vs MIN G2 (5/6) · SAS 133-95 · SAS -10.5 + over 215 + Wemby 30/15 · ~100:1
LAL vs OKC G3 (5/9) · OKC 131-108 road · OKC -7.5 + over 220 + SGA 35+ · ~100:1
PHI vs NYK G4 (5/10) · NYK 144-114 (sweep) · NYK -8 + over 230 + Brunson 35+ · ~100:1
SAS vs MIN G5 (5/12) · SAS 126-97 · SAS -7 + over 215 + Wemby 30/15 · ~100:1
The caveat
Small sample. Pattern needs 30-50 validation bets to confirm. We may have unconsciously cherry-picked. Real hit rate when triggered might be 25-40%, not 100%. Still profitable IF pattern holds. Could also be variance — 5 from 20 is statistically noisy.
2. The 4 trigger conditions (precise)
All 4 must be met before placing the full $10 bet. 3 of 4 → smaller stake ($5) for testing. 2 or fewer → skip.
Cond 1
Talent gap
Win-rate difference between teams: 10%+ regular season or 15%+ playoffs. Filters out coin flips.
64-18 (.780) vs 50-32 (.610) = 17% gap ✓
Cond 2
Favorite status
Favorite at home, OR favorite has top-15 scoring star with soft matchup.
OKC on road but SGA + dominant team ✓
Cond 3
Motivation factor (any one)
Bounce-back · series momentum (2-0 or 3-0) · pivotal game (G5 series tied) · closeout · statement (back from injury, vs rival).
Cond 4
Pace / scoring profile
Favorite in top-10 NBA pace, OR both teams in top-15 pace. Total line at 215+ (suggests high scoring expected).
Decision table
| Conditions met | Action | Stake |
|---|
| 4 of 4 | BET — trigger fires | $10 |
| 3 of 4 | TEST — smaller stake | $5 |
| 2 or fewer | SKIP | $0 |
3. Parlay construction
When trigger fires, build a 3-leg same-game parlay:
- Leg 1: favorite covers spread (typically -110)
- Leg 2: game total over (typically -110)
- Leg 3: star player over points line (typically -110 to -130)
Combined parlay odds at -110/leg: ~7:1. But SGP discount typically applies: actual offered ~5-6:1.
To reach 100:1, add one of:
- Star player over assists/rebounds (~+125)
- Specific team to reach X points first
- First-half spread (correlated with full-game cover)
Example parlay
- Pistons -5.5 (-110)
- Over 220.5 (-110)
- Cunningham over 28.5 pts (-115)
- Cunningham over 6.5 ast (-110)
- Combined: ~25:1 (closer to 30-40:1 in SGP)
For TRUE 100:1, add 5th leg or use longer spread.
Realistic offered odds
| Parlay shape | Offered odds | $10 → |
|---|
| Bare 3-leg correlated | 6-12:1 | $60-120 |
| 4-leg correlated | 20-40:1 | $200-400 |
| 5-leg correlated ← target | 80-150:1 | $800-1,500 |
4. Validation methodology — 3 phases
Phase 1 · 10 bets (weeks 1-3)
$10/bet · $100 total exposure
Target hits: 2-3. Track: hit rate, ROI, which legs hit/missed.
Decision point: 0-1 hits → pattern likely noise, abandon. 2+ hits → continue to Phase 2.
Phase 2 · 20 more bets (months 1-3)
$10/bet · $200 total exposure
Cumulative target: 5-8 hits. Refine trigger conditions based on what hits vs misses.
Decision point: hit rate < 8% → pattern not real. Hit rate 12-20% → real edge found.
Phase 3 · Scale (months 4+)
$20-25/bet · increased selectively
If pattern confirmed, increase stake. Bet more selectively to maintain edge. Track closing line value (CLV) as ongoing edge proof.
5. Ledger schema — KV-tracked per bet
{
"bet_id": "BCE-001",
"date": "2026-05-15",
"game": "DET vs CLE G6",
"trigger_conditions": {
"talent_gap": true,
"favorite_status": true,
"motivation_factor": "elimination_game",
"pace_profile": true,
"all_4_met": true
},
"parlay_construction": {
"leg_1": "DET +4.5 (-110)",
"leg_2": "Over 218 (-110)",
"leg_3": "Cunningham over 28.5 pts (-115)",
"leg_4": "Cunningham over 6.5 ast (-110)",
"combined_odds": "+2500",
"stake": 10,
"potential_payout": 260
},
"outcome": null,
"legs_hit": null,
"legs_missed": null,
"final_result": null,
"methodology_tag": "blowout_correlation_edge_v1"
}
Stored under KV key bce-bet-{bet_id}. Index of all BCE bets under bce-feed. Phase progress under bce-phase-state.
6. Upcoming games scanner
Each upcoming game gets evaluated against the 4 trigger conditions. Output: BET / TEST / SKIP recommendation.
⚾ Pirates vs Phillies G2 (Sat 5/16)Skip
Conditions met: 2/4. Talent gap moderate, no clear motivation factor, baseball doesn't fit this NBA-specific pattern.
talent gap weakno motivationwrong sportpace N/A
⛳ PGA Championship Sun finalSkip
Conditions met: N/A. Golf doesn't fit this pattern — different correlation structure. Use SameSHOT Framework instead.
🏀 Cavs vs Pistons G6 (TONIGHT)Test stake $5
Conditions met: 3/4. Talent gap weak (similar records), elimination motivation ✓, pace OK (regular-season totals ~225). Run as Phase 1 test with smaller stake.
talent gap weakeliminationpace okfav home (CLE)
Scanner updates per slate. Future: auto-populate from /bets/games/ manifest with structured fields.
7. Honest expectations — annual math
| Scenario | Triggers/yr | Hit % | Hits | Spent | Returned | Net |
|---|
| Optimistic (backtest rate) | 150-180 | 25% | 40 | $1,800 | $32,000 | +$30,000 |
| Realistic (50% decay) | 150 | 12% | 18 | $1,500 | $14,400 | +$12,900 |
| Conservative (75% decay) | 150 | 8% | 12 | $1,500 | $9,600 | +$8,100 |
| Pessimistic (noise) | 150 | 4% | 6 | $1,500 | $4,800 | +$3,300 |
| Worst case (cold year) | 150 | 0-1% | 0-2 | $1,500 | $0-1,500 | -$1,500 |
Year 1: -$1,500 to +$10,000 (learning phase). Year 2+: +$5K to +$25K if pattern holds.
8. Why books may be vulnerable
Books price each parlay leg independently:
- Cover probability: 50% (-110)
- Over probability: 50% (-110)
- Star prop probability: 50% (-110)
If independent: 0.5 × 0.5 × 0.5 = 12.5% true probability. At independent pricing parlay should pay 7:1; books typically pay 5-6:1 (their margin).
But these outcomes aren't independent
When favorite blows out underdog:
- Cover happens BECAUSE of blowout
- Over happens BECAUSE of high scoring (often blowouts)
- Star prop hits BECAUSE star drove the blowout
True correlation: maybe 80-90% conditional probability:
- Given cover happens, over happens ~70% of time (not 50%)
- Given over happens, star prop hits ~75% of time (not 50%)
Real probability of all 3: maybe 25-35% (not 12.5%). Books still pay 5-6:1. That's where the edge lives.
This is similar to "correlated parlay" strategies used by sharp bettors. The key is identifying WHEN correlation will be highest — that's what the 4 trigger conditions filter for.
9. Risk management — what never to do
- Chase losses — stake stays $10 regardless of recent results
- Increase on hot streaks — variance bites, stay disciplined
- Bet when only 2/4 conditions — reduces edge below profitability
- Bet on sports that don't fit — MLB has different correlation structure
- Ignore bankroll — need $500 minimum to weather variance
Bankroll management
- Min bankroll: $500 (50 bets at $10)
- Stop loss: $300 drawdown → pause, reassess
- Take profit: every $2,000 won, withdraw 50%
Weekly discipline
- Max 3 BCE bets per week (selective)
- Skip weeks with no triggers
- Track every bet honestly (wins AND losses)
10. How to know if we found the missing edge
After 30 bets
- Hit rate > 15%: PROVEN EDGE
- Hit rate 10-15%: SUGGESTIVE, continue
- Hit rate < 10%: PROBABLY NOISE
After 50 bets
- ROI > +100%: STRONG EDGE
- ROI +20-100%: REAL EDGE
- ROI 0 to +20%: BORDERLINE
- ROI < 0: NOT AN EDGE
After 100 bets
- Hit rate stable across last 50: PATTERN CONFIRMED
- Variance reduced: TRUST THE PROCESS
- Can increase stakes confidently
Closing Line Value (CLV) — the gold standard
If we consistently bet at better odds than the closing line, we're beating the market. CLV across 50+ bets is the strongest evidence of real edge — stronger than ROI (which is noisier).
📌 Worth testing?
Sam decides. The framework is built. The 4 trigger conditions are precise enough that a chat or this page can scan any slate. The validation phases are defined. If pattern holds, this is the most promising thing we've found. If it doesn't, we'll know in 30-50 bets at $300-500 total exposure. That's a fair learning cost for a potential $5K-25K/yr edge.