Review Of The Bookie Odds System

in sports •  9 months ago


Now that we're done with our sample of matches, it's time to review the system and see what worked and what didn't. Overall, i'm a little disappointed with the results and think there's room for further improvement.

The Results

Total PredictionsCorrectWinningsROI
30178.41 units28%

Getting 17/30 right is just 57%. Given that odds were all above 2, we'd minimally need a strike rate of 50% to make this work. So where did it all go wrong?

Asian Leagues

The system went 0/3 in the Asian leagues - Japan, China and Korea. Personal experience has taught me that football in Asia tends to be a little unpredictable, with form and home advantage often going out the window.

Normally i avoid betting on the smaller Asian leagues like Singapore, Malaysia and Indonesia but it seems that this system doesn't work for the bigger Asian leagues as well. Going forward, i'd avoid them no matter how tempting the data looks.

Home and Away

Breaking down the predictions by Home and Away - the system predicted 25 Home wins and 5 Away wins. 60% of the Home team to win predictions ended up being right while only 40% of the Away team to win predictions were correct.

Couple of things to note - All the Asian losses were on the Home team. Also, a sample size of 5 Away matches is small. That being said, i think focusing on Home wins provides better value, especially if the odds are all in the same range of between two to three.

Lack of matches

One of the frustrations with the system is that there aren't many matches that fit the criteria to bet on, especially with many of the major leagues winding down. So here's a thought i had while doing this test run - what if we just focus on the team's probability of clearing the odds rather than the difference?

Let's just look at Home teams who clear odds greater than two more than 50% of the time. In theory i believe we ought to get similar results with more matches to choose from (I think).

Authors get paid when people like you upvote their post.
If you enjoyed what you read here, create your account today and start earning FREE STEEM!
Sort Order:  

I'm glad that you could demonstrate profitability from your test, it's hard to know exactly what goes into what sofascore say a team wins at these odds a certain % of the time, but it is good that there is some validity to what they are saying anyway. I agree it is difficult to get a decent sample of matches or find matches to bet on that on a given day. Some leagues, like southern hemisphere leagues may not be winding down just yet, but the calibre of players in top leagues may be lower than usual over June due to world cup.

Sofascore is definitely a good resource to consider for opportunties like what you illustrated in your test of 30 games, I was looking at it for NBA for some of the boston home matches in the playoffs. Boston were marked around evens on exchanges ( 1.8-1.9 on bookmaker websites) at home, and sofascore was saying they'd win these matches 75% of the time.
Boston are now 10-0 at home in the playoffs ( 1-6 on the road) and the sofascore resource has definitely made a difference on calls I made. That being said, it's small sample size of what I took, and there have only been fewer than 20 teams in NBA history who have won 10 home games in a row in playoffs, so it probably has some luck and positive variance in the results
I will be looking at Boston to win (depending on price) against cleveland at home if they lose the game 6 tonight in Cleveland, which I won't be betting on.

Another resource I will be looking at is nate silver's elo rankings of various sports, on I think he acknowledges that on some sports, predictions based off of elo rankings don't beat the market, but I'm not sure if that is due to bookmaker overround, or not. ( maybe profitable on betting exchanges) I will have a post about this early next week I'd say and start a similar test over 30 to 50 matches to see if there is anything to it on baseball.


Unfortunately, sofascore doesn't give data on clearing handicaps by home and away or the sample size. I'm not too familiar with US sports other than basketball and my impression was that the odds on handicaps are a lot better compared to just the one on Win/Lose.

Thanks for the fivethirtyeight links - i didn't know they did sports analysis as well. Looking forward to the results from your test, if you decide to do a post about it.

@numpypython congrats!

I think that the model will hold and that with some tweaking you can improve the profitability of it.

I can't wait for the Big 5 to start 💰💰💰


Thanks @beat-the-bookies. Hope we make some money with this haha

A roi of 28% is just amazing. Although this could fluctuate a lit because of the small sample of bets. To be sure you will need at least 1000 bets with the system to be sure!

Probably the 40% of away games still present a profit because the odds for an away win are usely higher than a win for the home team! Any figures about this?

Great analysis!



Thanks, Peter. Yeah i think a thousand sample bets would provide a lot more confidence as well as a lot more new ideas along the way. Some dodgy Central Limit Theorem hopefully gives us a little confidence.

The Away matches still ended up with a loss of 0.8 units in the end. Not the highest odds on these unfortunately.

Back to the drawing board lol.


It takes time and learning from the mistakes when building a profitable model!
Very interested in the model!