Signal Score Methodology — How We Rank Research Signals
Signal Score Methodology — How We Rank Research Signals
> Signal Syndicate Research Library · Methodology · Educational intelligence · Not betting advice
What is Signal Score?
Signal Score is Signal Syndicate's ranking metric for research signals on the Signal Board. It measures how large the disagreement is between our model's estimate and the market's pricing — not your chance of winning.
Think of it like a volume dial for model-market gaps:
- Higher score = Larger difference between what the model sees and what the market prices
- Lower score = Smaller difference
Signal Score helps you scan the board quickly and see which matchups have the biggest research gaps worth exploring. It does NOT tell you what to bet or predict outcomes. It ranks the research gap behind what Signal looks for in MLB markets.
How Signal Score Works (Plain English)
The Formula (Simplified)
Signal Score combines two factors:
1. Gap Size — How far apart is the model from the market? (measured in runs, points, etc.)
2. Strength Tier — How confident is the model in its estimate? (Strong, Moderate, Slight)
Signal Score = |Model Edge| × Strength Multiplier
Example:
- Model sees total at 8.2 runs
- Market prices 9.0 runs
- Difference: -0.8 runs (model favors UNDER)
- Strength: Moderate
- Signal Score: ~60-70 (rankable research gap)
What Scores Mean
| Score Range | What It Means |
|---|---|
| **80-100** | Large model-market gap · Strong research signal |
| **60-79** | Moderate gap · Worth exploring |
| **40-59** | Slight gap · Context research |
| **< 40** | Minimal gap · Market agrees with model |
Important: These are research hierarchy rankings, not win probabilities.
- Score 85 does NOT mean 85% chance of winning
- Score 45 does NOT mean 45% chance of winning
The Three Strength Tiers
Signal Board cards show one of three confidence levels:
*Strong*
- Model sees large structural advantage (2+ run edge)
- Multiple factors align (pitcher, defense, park, weather)
- Historical data supports the gap
- Multiplier: 1.2x-1.5x
*Moderate*
- Model sees clear advantage (1-2 run edge)
- Key factors favor the signal
- Some historical support
- Multiplier: 1.0x
*Slight*
- Model sees edge but conditions are noisy (0.5-1 run edge)
- Mixed factors or smaller sample size
- Research context, not promotional
- Multiplier: 0.7x-0.9x
What Signal Score Is NOT
❌ Not Win Probability
- Score 75 does NOT mean 75% chance of winning
- It measures gap size, not outcome likelihood
❌ Not Betting Edge
- High score = large disagreement, not guaranteed profit
- Market may be right; model may be wrong
❌ Not a Pick Recommendation
- Signal Board shows research context
- Not telling you what to bet
❌ Not Consistent Across Sports
- MLB run totals use different scale than NBA point totals
- Compare scores within a sport, not across sports
How We Use Signal Score
On the Signal Board:
- Pinned signals (top 3) have highest scores for the day
- Cards sort by score descending (biggest gaps first)
- Helps users scan research quickly
In Daily Intelligence:
- Featured matchups may reference signal strength
- Context for why a game is highlighted
NOT Used For:
- ❌ Pick promotion ("Lock of the day")
- ❌ Marketing claims ("95% win rate")
- ❌ Performance tracking (that's ROI/CLV)
Transparency: What Affects Scores
Signal Scores can change based on:
1. Line Movement — If market moves toward model, gap shrinks, score drops
2. New Information — Injury, weather, lineup changes update model edge
3. Route/Lane Changes — Different model lanes use different edge caps
4. Model Updates — Refinements to model logic adjust estimates
Current Board Route: ROUTE_B (v1 broad totals, |edge| ≤ 2.5 cap)
Example Walkthrough
Game: Yankees vs Red Sox
- Market Total: 9.0 runs
- Model Estimate: 7.8 runs
- Gap: -1.2 runs (model favors UNDER)
- Strength: Moderate (both starters have strong recent form)
- Signal Score: 68
What This Means:
- There's a 1.2 run disagreement between model and market
- Score 68 = moderate-sized research gap
- Worth exploring in Daily Intelligence
- NOT a guarantee, NOT a pick
Limitations & Disclaimers
Sample Size
- Small slates (2-3 games) have fewer comparison points
- Scores more meaningful on 8+ game slates
Market Efficiency
- Closing lines aggregate sharp action and late info
- High Signal Score doesn't mean market is wrong
- Model may be missing context market has priced
Route B Edge Cap
- Plays with |edge| > 2.5 runs excluded from public board
- Shadow lanes may show different score ranges
Not Performance Validation
- Signal Score is for board hierarchy only
- See Performance page for ROI, CLV, hit rate
- Historical backtests separate from score methodology
Related Reading
- Signal Methodology — Full model overview
- CLV Validation Study — How we validate process
- 75-Bet Gate Policy — Why we require large samples
FAQ
Q: If Signal Score is high, should I bet that game?
A: No. Signal Score shows research context. We don't tell you what to bet. Use it to explore matchups in Daily Intelligence, not as betting instructions.
Q: Why did a score change during the day?
A: Lines moved, or new information (injury, weather) updated the model. Scores reflect current model-market gap.
Q: Can I trust high-scoring signals more than low-scoring signals?
A: High scores mean bigger gaps, not better accuracy. Market may be pricing info the model doesn't have. Treat all signals as research context.
Q: What's a "good" Signal Score for betting?
A: We don't promote betting thresholds. Signal Board is educational research, not pick guidance.
Q: Do all sports use the same scoring system?
A: No. MLB uses run-based edges, NBA uses point-based edges. Compare scores within a sport only.
Educational research only · Estimates only · Not betting advice · Signal Score is for board hierarchy, not betting recommendations