Top value · model vs market
edge = model_p − implied_pAll fixtures · sortable
click row to expand| Time | Home | Away | H | D | A | Edge | ||||
|---|---|---|---|---|---|---|---|---|---|---|
| Loading… | ||||||||||
No fixtures with both odds and predictions today.
Try a different date or league.
How it works · model vs market
two ways to price the same outcomeThe core idea
Every match has two prices.
We bet when they disagree.
Borrowed straight from quant finance. Bookmakers publish a market price (the odd you can actually click on). Our model publishes a model price (what we think the outcome is really worth). When the model says an outcome is more likely than the market implies, there's an edge — and that's a value bet.
Example · Home win
Market
40%
Model
52%
Edge
+12%
Market underestimates the home win. The 12-point gap is what makes this a value bet.
Comparison · the two valuation methods
| Feature | Mark-to-Market aka "Market Price" · what books charge |
Mark-to-Model aka "Model Price" · what we compute |
|---|---|---|
| Definition | Values an asset based on the current price quoted on an active exchange or market. | Values an asset using financial theories, mathematical formulas, and historical data. |
| Best used for | Highly liquid and widely traded assets (e.g. publicly traded stocks, top-flight football). | Illiquid, highly complex, or unique derivatives (e.g. correlation swaps, unlisted bonds, lower-division matches). |
| Advantages | Reflects exact, real-time capital availability; highly objective and transparent. | Allows valuation of assets where no direct market exists, or where the market is suspected to be wrong. |
| In this app | The odd at Bet365 / Unibet / 10Bet — converted to 1/odd, then normalised by removing the bookmaker overround. |
API-Football's ML prediction — historical form, attack/defence ratings, H2H, Poisson goal expectancy. |
| Trusted when | Many sharps are pricing the same match — the market converges on truth. | The market is thin or slow to react — books haven't priced in form, injuries, or fixture context yet. |
Try it · interactive simulator
① Market · bookmaker odds
② Model · your prediction (must sum to 100%)
auto-normalised so they always sum to 100%
Σ = 100%
③ Edge · model − market
Best value outcome
Home win
Edge
+12.0%
Model says home win is worth backing — 12 points more likely than the book implies.
Side by side · all outcomes
Book overround8.2%
Sum of raw implied prob108.2%
Raw
1/odd probabilities always sum to MORE than 100% — that's the book's margin (vig). We divide each by the sum so they total 100%, then compare to the model's probability.
Show the math
Market → implied prob
raw_prob(outcome) = 1 / odd
margin = Σ raw_prob (always > 100%)
implied_prob = raw_prob / margin
Removing the margin gives a fair "no-vig" probability — what the book really thinks before adding its cut.
Model → edge
model_prob = ML output (normalised)
edge = model_prob − implied_prob
value bet when edge > 0
Color thresholds: ≥ 5% hot · 2–5% warm · < 0 cool.
Edge ≠ guaranteed profit
A 10% edge means positive expected value over MANY bets, not a guaranteed win on the next one. Variance is real.
Models lag reality
If a star striker is injured 2h before kickoff, the market re-prices instantly. The model still thinks last week's form. Always cross-check.
When the market wins
In top-flight, deeply-traded matches, the market is usually right. Persistent edge tends to show up in less-watched fixtures.