Hacker News new | comments | show | ask | jobs | submit login
Show HN: Football Prediction API (boggio-analytics.com)
36 points by boggio 9 months ago | hide | past | web | favorite | 29 comments

This seems too good to be true. If they could actually predict football results with better-than-random accuracy, then why wouldn't they just use this themselves to make loads of money on betting markets?

I don't think they're claiming their statistics-based predictions are necessarily more accurate than the bookmakers' odds

Good point, I suppose "69%" accuracy doesn't actually mean that much unless you consider the odds.

I'm pretty sure I could predict Real Madrid vs. my local home town team with better than 69% accuracy, but that doesn't mean much

Well sure, but there are over 2200 games per month and less than 3-5% are the kind of games you are talking about. Super good team vs terrible team (assuming your local home town team is not Barcelona)

It has 69% accuracy, that means you still lose 31% of the time.

69% is a huge number in terms of sports betting. Typically to be profitable you need to be in the 52% - 56% arena to be profitable, so this is significantly profitable

Neat! What is it's past performance / accuracy like?

There is an endpoint that measures accuracy.

Long-term accuracy is 69%.

What is the baseline accuracy if you just took a measure like team ELO and predicted based on that?

Over what period and how far have you back-tested?

1 year and a couple of months.

No Champions league? or UEFA cup? Just local leagues?

No.. maybe in the future.

I think you should call it "soccer" or at least "association football" in this context.

No we should call american football "Hand Egg" instead

American Full-Contact Hand "Ball"

In Germany (one of the countries with long standing football tradition) it is called "Fußball" which splits into "Fuß" (foot) and "Ball" (ball). It makes much more sense than calling "American Football" Football, as in "American Football" the feet are less dominant than in "soccer" (where hand usage is actually a foul).

Hmm, I though its called football in the rest of the world except the US. Apparently there are other countries that call it soccer too.



Because it's always better to be more specific in all cases.

In this case, especially because it's an international, but heavily American audience.

Because most native English speakers call it soccer.

Really? Everyone I know calls the sport football. The second F in FIFA is football.

I know Americans call it soccer, but that is I assume because of American football, a sport no other country seems to play, or care about.


The name soccer came from Britain, not the US. It was used to differentiate Association Football (soccer) from Rugby Football (rugby). Ireland, South Africa, Australia and New Zealand all still call the sport soccer. Even in England, it continued to be a popular name for the sport until the 1950s.

Pretty mixed in Ireland to be honest, and we have our own sport Gaelic Football which gets confusing.

> I know Americans call it soccer

Correct, and most native English speakers are American. That doesn't even account for Canada or Australia where it has also traditionally been called soccer.

The fact that the "F" stands for "Football" in the name of an organization with a French name seems irrelevant.

I don't think many people likely to be interested in an association football API call it "soccer". They're marketing to that audience, not the minority of the world that assumes "football" refers to a gridiron sport.

Because any good developer should, when presenting a product in an international market, consider i18n (which involves more than just "translate from language A to language B").

Interesting but football is such a terrible sport to predict statistically outside of a few teams (the outliers) in each league. Kind of hope there would me tools out there to mess with play patterns and tendencies more than the actual result, I think that it would be actually useful plus such a resource paired with a decent analyst would net a way better prediction rate. But betting.jpg

>mess with play patterns and tendencies

Maybe you should get involved with Data Science, this field is all about playing with patterns (and hopefully finding a result), and even thinks like SVM's, random forests and genetic algorithms aren't really hard to grok and apply in that area.

There are some scientists involved with betting, but this field has really low profit margins IMHO. Personally, I view algorithmic trading as more interesting but YMMV.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact