How to find value bets: a step-by-step method

What you need before you start

Step 1: pick a fair reference

Every value bet is measured against a fair price, so the reference is the most important choice you make.

Read more about using Pinnacle as a reference in Pinnacle odds API, and see the full list in Sharp vs soft bookmakers.

Step 2: match exactly the same bet

Step 3: remove the margin from the reference

Step 4: calculate the expected value of every price

Now compare each soft bookmaker with the fair probability: EV = odds × fair probability − 1. With a fair price of 2.00 (50%):

More worked examples in How to calculate expected value.

Step 5: filter before you bet

Step 6: log every bet and check closing line value

Where value bets usually show up

Main markets in top leagues close to kick-off are the most efficient. Value there is rare and small.

Finding value bets in football

Three ways to do it

A minimal value bet scanner in Python

This script does steps 2 to 5 against the ValueOddsAPI response format: each row has the bookmaker’s original price, the margin-free fair probability and the age of the observation. Use the bookmaker IDs your key is entitled to.

import os
import requests

BASE = os.environ["VALUEODDS_BASE_URL"]
KEY = os.environ["VALUEODDS_API_KEY"]
REFERENCE = "pinnacle"                 # the sharp price you trust
SOFT = ["bet365", "unibet", "betano"]  # the prices you can take
MIN_EV = 0.02                          # flag +2% or better
MAX_AGE = 10                           # seconds; older prices are skipped

def fetch(bookmakers):
    r = requests.get(f"{BASE}/v1/odds",
                     params={"phase": "prematch", "bookmakers": ",".join(bookmakers)},
                     headers={"Authorization": f"Bearer {KEY}"}, timeout=10)
    r.raise_for_status()
    return r.json()["data"]

def same_bet(row, outcome):
    m = row["market"]
    return (row["event_id"], m["id"], m["period"], m["line"], m["settlement"], outcome["id"])

rows = [r for r in fetch([REFERENCE, *SOFT])
        if r["market"]["status"] == "open" and r["data_age_seconds"] <= MAX_AGE]

# Step 3: fair probabilities from the reference (the API has removed the margin)
fair = {same_bet(r, o): o["fair_probability"]
        for r in rows if r["bookmaker"] == REFERENCE and r["market"]["fair_status"] == "available"
        for o in r["market"]["outcomes"]}

# Steps 4-5: expected value for every soft price, filtered by threshold
for r in rows:
    if r["bookmaker"] == REFERENCE:
        continue
    for o in r["market"]["outcomes"]:
        p = fair.get(same_bet(r, o))
        if p is None:
            continue                   # no matching reference: skip, never guess
        ev = o["bookmaker_odds"] * p - 1
        if ev >= MIN_EV:
            print(f'{r["event_id"]} {r["market"]["id"]} {o["id"]} @ {r["bookmaker"]}: '
                  f'{o["bookmaker_odds"]:.2f} vs fair {1 / p:.2f} -> EV {ev:+.1%}')

In production you would run it on a WebSocket stream instead of polling, store every flagged bet, and add the closing price after kick-off. See the integration guide and the value betting API.

Frequently asked questions

Related reading