Opening Odds, Falling Odds
Back when i was doing my internship, many years ago, the husband of a colleague of mine used to be into sports betting and would make a decent sum every month with this theory of his. Basically, he tracks the odds of a match from the time they're released till just before they close (in-game bets not included).
His theory was that you want to bet on the team with falling odds. With one caveat, if the odds drop too much, that indicates a 'heavy boat' and is a match that's likely to be fixed.
Let's put some filters
I've got no idea how he filtered which matches to bet on but for our purposes, let's set the following rules:
- Team must be favorite (lowest odds)
- Team should have falling odds from $2+
- Odds must still be greater than 2
- Team must have overall winning H2H or Winning Manager H2H
Today's sample
Match | Prediction | Odds | Outcome |
---|---|---|---|
Al Riffa vs Al Shabbab | Al Riffa | 2 | L |
Real Oviedo Vs Valladolid | Real Oviedo | 2.04 | L |
Union Espanola vs Huachipato | Union | 2.1 | W |
Ebbsfleet vs Sutton United | Ebbsfleet | 2.2 | L |
SJK vs Honka | SJK | 2.39 | W |
Boulogne vs Pau | Boulogne | 2.2 | L |
Kyoto vs Kumamoto | Kyoto | 2.04 | L |
Tokyo Verdy vs Mito Hollyhock | Tokyo | 2.1 | W |
Queen of the South vs Dundee | Dundee | 2.2 | W |
St. Mirren vs Greenock | St. Mirren | 2.04 | W |
Review
So overall, we got 5/10 correct, which is better than E(X) of 3 or 4 out of 10. Given that we're looking at odds over $2, getting a 50% win ratio is needed to break even / make a small profit.
This round's return was 5.4% which is not worth the risk in my opinion. For the next round of 10, i'm going to modify the filter by using a form filter instead of historical / manager H2Hs.
An interesting approach...
I'm waiting for the results to come in, and if the theory proves itself, that's a moneymaker in your hand.
What data sources do you use? Is there a dataset with the opening and closing odds?
If there is, applying a bit of ml magic to find out what filters are most profitable wouldn't be hard.
So far two losses from the opening matches lol. I'll track this for 30 matches and see if this makes any money over the period.
For today, i got the info from sofascore. The format's not clean enough to download or copy/paste but it can be scraped.