Two bets per game. Tier 1 = safe play ($5, 60-70% hit). Tier 2 = refined longshot ($10 → target $1,000+, 5-15% hit). Every bet tagged with methodology so we learn which approaches actually work.
For every game: generate two bet options. Tier 1 protects the bankroll (consistent small wins). Tier 2 swings for the fence (5-15% hit on $10 → $1,000+). Every bet tagged with methodology so we learn which approaches actually work over months.
1. The vision
This is the operational version of everything we've learned from backtests (see /bets/backtests/ — 48 games, 4 iterations).
Tier 1: high-probability play (ML/spread, $5 stake, 60-70% hit rate)
Methodology tags on every bet — over time you learn which approaches work for YOUR betting
2. Tier 1 — high-probability bets
Methodology · selective discipline + talent gap
For each game, evaluate:
Talent / record gap between teams
Home court advantage
Recent form (last 5-10 games)
Situational factors (rest, B2B, travel)
Line vs my probability estimate
BET when ALL of these:
My probability ≥ 60%
Line agrees with my pick (no contrarian)
Edge of at least 4% over implied probability
No "narrative-only" reasons (must have data)
SKIP when ANY of these:
My probability 50-59% (coin-flip range)
Line requires >65% to break even (-185 or worse)
Major injury uncertainty
Game has unusual circumstances I can't price
Stake: $5 flat (no confidence tiers — calibration unproven). Expected hit rate: 60-70%. Expected ROI: 2-5%. Annual profit at $25/wk: $100-300.
This is the boring but real part.
3. Tier 2 — refined longshots
The point: $10 bets that can return $1,000+ if they hit. Need ~3-5 hits per year to break even, 7+ to profit. Five methodologies, each with proven mechanism.
Mechanism: outcomes are linked, books often misprice correlation. Star player going off correlates with team winning + total going over.
Cunningham 30+ pts · Pistons ML · over 220
Use when: 4 trigger conditions met — see Blowout Correlation Edge for the precise framework + validation tracker.
Backtest: 5/5 hits (cherry-picked, in validation). Target: 10-15% real hit rate.
Method 2 · 4-Leg Straight Parlay (selective games only)
~50:1 to 100:1 with props
Mechanism: pick 4 games where you have STRONG conviction (70%+ each). Combined ~16:1. Add 1 prop at +200 = ~50:1; 2 props = ~100:1. Each leg has small edge, compounds.
4 high-confidence favorites · 1 player prop · 1 game total
Use when: multiple games on slate have STRONG plays — especially high-confidence favorites with similar line ranges.
Hit rate target: 5-10%
Method 3 · Contrarian Parlay (when public is heavy)
8:1 to 25:1 per leg; combine for 100:1
Mechanism: find 2-3 games where 75%+ of public is on the favorite. Line shifts toward favorite, value on dog.
3 underdog MLs where public 75%+ on favorite
Use when: public betting trends visible (most apps show this) AND sharp action is opposite of public.
Hit rate target: 10-20%
Method 4 · Live-Betting Longshot
~6:1 to 12:1 per leg
Mechanism: live lines slow to update; early-game momentum is partly predictable. Bet comeback ML when favorite is down 10-15+ at half. Combine 2-3 for 100:1.
Comeback ML × 2 games + game-prop hit
Use when: favorite expected to fall behind in early-game scripts AND you can watch the game live.
Hit rate target: 8-15%
Method 5 · Specific Score / Exact Outcome
3:1 to 20:1 per leg
Mechanism: a specific outcome with clear value from situational factors. "Decided by 1 run" MLB ~3:1; "team wins by 11+" NBA spread parlay; "first scorer + game winner" NHL.
"PIT wins by 1 run" · "Reynolds first hit" · over 8.5
Use when: one particular outcome has clear value AND can be paired with normal bets.
Hit rate target: 5-12%
4. Per-game framework
For ANY game, generate this structure (this is the template that should appear on every game-analysis page going forward):
GAME: [Team A] vs [Team B] @ [venue]
TIME: [start time]
TIER 1 PLAY · the safe bet
Pick: [Team or Spread]
Confidence: HIGH / MEDIUM
Reasoning: [2-3 specific factors with data]
Stake: $5
Expected payout: $X
TIER 2 PLAY · the longshot
Methodology: [Method 1-5]
Parlay legs: [list legs]
Stake: $10
Target payout: $1,000+
Hit probability: [estimated %]
Reasoning: [why this specific construction]
METHODOLOGY TAGS · for tracking
Tier 1 tags: [talent_gap | situational | line_value | ...]
Tier 2 method: [1-5]
POST-GAME LEARNING NOTES
Did Tier 1 hit? [yes / no]
Which legs of T2 hit? [list]
Factor that mattered: [free text]
5. Realistic financial expectations
Annual budget breakdown (recommended)
Tier
Schedule
Exposure
Hit-rate cases
Annual outcome
Tier 1
$25/wk × 52
$1,300
65% hit, +3% ROI
+$39/yr
Tier 1
$25/wk × 52
$1,300
60% hit (break-even)
$0/yr
Tier 1
$25/wk × 52
$1,300
55% hit (vig wins)
-$130/yr
Tier 2
$10 × ~150 selective
$1,500
5 hits/yr at 100:1
break-even
Tier 2
$10 × ~150
$1,500
8-12 hits/yr
+$3K-7K
Tier 2
$10 × ~150
$1,500
0-3 hits/yr (cold)
-$1,200-1,500
Combined expectation
Case
Annual P/L
Best case (top 5%)
+$3,000 to +$7,000
Average case
-$200 to +$500
Worst case (cold year)
-$2,800
Realistic first year: -$500 to +$1,000 while learning patterns.
Realistic year 3+ (if disciplined): +$1,500 to +$3,000/year.
Key metrics to track
Tier 1 ROI separate from Tier 2 ROI (so they don't pollute each other)
Methodology hit rate (which longshot method works for YOUR bets)
Closing Line Value (CLV) — gold-standard for real edge
6. Methodology tracking schema
Every bet gets tagged so we can isolate which approaches actually win.
The dual-tier vision works mathematically. Method 1 (correlated parlays on blowout setups) is the strongest longshot pattern in our backtest — see sections 19-20.
But the math requires DISCIPLINE ($10 longshots only when conditions align, not on every game) and ACCEPTANCE OF VARIANCE (cold streaks of 20-30 bets will happen before hits cluster).
Expected first-year outcome: -$500 to +$1,000 while learning patterns. After 6-12 months of disciplined tracking: real edges become visible. This is a 3-5 year project to build real edge. Quick money isn't possible with positive expected value. Slow money is.