BKFC Betting Glossary
Moneyline, implied probability, vig, edge, CLV, and every other term you need to read our research pages.
Moneyline
The price you pay to bet a fighter to win. American format:
- Negative (-180): favorite. Risk $180 to win $100.
- Positive (+150): underdog. Risk $100 to win $150.
The more negative a favorite's number, the bigger the favorite. -1000 means the book thinks this fighter wins 90%+ of the time.
Implied probability
The win-probability embedded in a moneyline. Formulas:
- Favorite:
|ml| / (|ml| + 100) - Underdog:
100 / (ml + 100)
Example: -180 implies 64.3%. +150 implies 40.0%. Add them up and they exceed 100% because of the vig (see next).
Vig / juice
The sportsbook's built-in edge. If two sides of a two-outcome market sum to more than 100% implied probability, the excess is the vig. Typical BKFC markets carry 7-10% vig — higher than mainstream sports because the market is less competitive.
De-vigging = normalizing the two sides to sum to 100% so you can see the book's "true" estimate of each fighter's chance. We use de-vigged probabilities when computing edge vs. the model.
Edge
The gap between the model's probability and the market's de-vigged probability, expressed in percentage points. A 15-point edge on Fighter A means our model gives them 60% to win while the market gives them 45%.
Bigger edge = more value IF the model is right. The catch: our model is wrong 40% of the time, so big edges on low-confidence picks are not automatic wins.
Advantage Score
Our primary pick display, 0-100 per fighter. Directly equals model probability × 100. A 67 means the model gives this fighter a 67% win chance. The two advantage scores in a bout always sum to 100.
Predictability
A separate 0-100 trust metric for how much to rely on a given pick. Blends data quality (50%), signal strength (30%), and factor agreement (20%). A 67/33 pick with Predictability 85 is much more bettable than a 67/33 pick with Predictability 25.
We filter the /picks/ aggregator to Predictability ≥ 40 so the board isn't flooded with noise.
Closing Line Value (CLV)
The classic pro-bettor metric. If you bet Fighter A at +150, and the line closes at +120 (book moved toward Fighter A), you've captured positive CLV. If it closes at +180, you got negative CLV.
Over a large sample, CLV is a better predictor of long-term profit than win rate. The book's closing line is the sharpest estimate of true probability. Beating it consistently = real edge.
We track CLV passively via odds history; once enough events have cycled we'll surface CLV stats per edge type.
Shrinkage
A calibration technique. Our model's probabilities tend to over-estimate at the 65-70% range (claims 67%, delivers ~60%). Shrinkage pulls all probs toward 50% by a factor: p' = 0.5 + 0.85 × (p - 0.5). Pre-shrinkage 67% becomes post-shrinkage ~64%. More honest.
Visible in the fact that most Advantage Scores cluster in the 35-65 range rather than getting extreme.
Book consensus
When we have moneylines from multiple sportsbooks (xbet.ag, BestFightOdds, etc.) for the same bout, we compute the median of each side to get a consensus line. Then we surface the best individual line as a "line-shopping" callout. The best ML beating consensus = free value that nobody's priced in yet.