How to build an odds comparison: a developer guide

New to odds data? Five terms first

Odds comparison

A page or app that shows the best available price for each outcome across bookmakers.

Normalise

Convert every bookmaker’s naming into one shared format.

Line

The number in a market, such as 2.5 in Over/Under 2.5 goals.

Stale price

A price that has not been confirmed recently and may no longer be available.

Arbitrage

Best prices that together guarantee a profit. Rare, short-lived and often a data error.

Start with one row per price

Whatever your source, flatten it into one row per price. Every later step, from matching to display, becomes a simple group-by on these fields.

Step 1: match events across bookmakers

Step 2: normalise markets and lines

Step 3: pick the best fresh price

With clean rows, the comparison itself is short. This example uses illustrative data, not a live feed. It drops prices older than five minutes, then keeps the highest price per selection.

from collections import defaultdict
from datetime import datetime, timedelta, timezone

# One row per price: illustrative data in a normalised shape
now = datetime(2026, 11, 2, 18, 0, tzinfo=timezone.utc)
rows = [
    # event_id, market, line, selection, bookmaker, odds, updated_at
    ("ars-che", "1x2", None, "home", "Bookmaker A", 2.10, now - timedelta(seconds=40)),
    ("ars-che", "1x2", None, "home", "Bookmaker B", 2.18, now - timedelta(seconds=75)),
    ("ars-che", "1x2", None, "home", "Bookmaker C", 2.25, now - timedelta(minutes=14)),
    ("ars-che", "1x2", None, "draw", "Bookmaker A", 3.50, now - timedelta(seconds=40)),
    ("ars-che", "1x2", None, "draw", "Bookmaker B", 3.40, now - timedelta(seconds=75)),
    ("ars-che", "1x2", None, "away", "Bookmaker A", 3.60, now - timedelta(seconds=40)),
    ("ars-che", "1x2", None, "away", "Bookmaker B", 3.75, now - timedelta(seconds=75)),
    ("ars-che", "total", 2.5, "over", "Bookmaker A", 1.95, now - timedelta(seconds=40)),
    ("ars-che", "total", 2.5, "over", "Bookmaker B", 1.91, now - timedelta(seconds=75)),
    ("ars-che", "total", 2.5, "under", "Bookmaker A", 1.87, now - timedelta(seconds=40)),
    ("ars-che", "total", 2.5, "under", "Bookmaker B", 1.93, now - timedelta(seconds=75)),
]

MAX_AGE = timedelta(minutes=5)

def best_prices(rows, now):
    """Best fresh price per selection, grouped by event, market and line."""
    markets = defaultdict(dict)
    for event, market, line, selection, book, odds, updated in rows:
        if now - updated > MAX_AGE:
            continue                       # stale price: never show it as best
        key = (event, market, line)
        best = markets[key].get(selection)
        if best is None or odds > best[0]:
            markets[key][selection] = (odds, book)
    return markets

for (event, market, line), sels in best_prices(rows, now).items():
    book_sum = sum(1 / odds for odds, _ in sels.values())
    label = market if line is None else f"{market} {line}"
    print(f"{event} {label}  (best-price book: {book_sum:.1%})")
    for sel, (odds, book) in sels.items():
        print(f"  {sel:6} {odds:.2f}  {book}")

Output:

ars-che 1x2  (best-price book: 101.1%)
  home   2.18  Bookmaker B
  draw   3.50  Bookmaker A
  away   3.75  Bookmaker B
ars-che total 2.5  (best-price book: 103.1%)
  over   1.95  Bookmaker A
  under  1.93  Bookmaker B

Bookmaker C’s 2.25 on the home win would have been the best price, but it was 14 minutes old. A stale best price is the most common way an odds comparison misleads its users: the price is gone when they click.

Step 4: sanity-check the result

Launch checklist

Using ValueOddsAPI for this

Frequently asked questions

Related reading