Skip to content

Building a predictive model for the World Cup

kimonolabs.com
22 pointspranade7 comments
On HN

Comments

The analysis provides a derivation of the model, but it does not even hint at the accuracy of a finished model, and there's plenty of data to do cross-validation.

Good point - thanks for the feedback! Yes, it's been about 75% on the R16 games. It's also fully interactive - so you can change the weights to explore relationships with historical data that might be interesting.

France is beating Nigeria with 100% certainty, and Brazil beats Chile with 35% probability i.e. loses with 65% probability.

I don't think the model is working great.

Yes, that is with high certainty, but not 100%. We'll fix the messaging hear - thanks for flagging. Also, it's interactive so you can change the weights and see what happens

They appear to have built just an empty black web page for me (Firefox/Mac)

Thanks for catching - for now you'll have to view it in Chrome.

france beats germany with 100% certainty?! D:

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.