Ask HN: I have 15 years of horse racing data what should I do with it? | Hacker News Reader

Ask HN: I have 15 years of horse racing data what should I do with it?

7 pointsmcgin17 comments

The data covers 203,916 races. I have info trainer, jockey, individual horses, betting odds, and a multitude of other things. I have a few ideas for what I can do with it, but thought I'd get some input before starting work on it.

View on HN

Do you have a background in econometrics or regression analysis? You might be able to develop a model to predict winning horses.

That is one of the plans I have in mind

Do you have info on the way the lines change from initially being set to post time? That shit is totally fixed, if you could figure out what certain line changes mean in terms of predicting a winner you could perhaps make some money.

Unfortunately no, its purely the prices at the off which I have available. However I am planning on monitoring future races to see how their prices change over time

post it on http://infochimps.com/

Just off the top of my head, I'd say it's pretty much useless without a betting strategy. I'd recommend starting with "Dr. Z's Beat the Racetrack". It's out of print, but you should be able to find a copy.

How about creating a simple site where you simulate a horse race. i.e. give users part of the history data for horses, ask them to select the winners, and then reveal the true outcome. People would be betting for boast points.

A discussion board where people can discuss their heuristics for selecting winners would be great. For the famous horses, you can link to its Wikipedia page, Flickr photos, etc. if they exist to create a more immersive experience.

I, for one, would play with such a site for a while.

I think this is a cool idea.

Just to add to this, I would center it around the 'process' of trying to select the winners if this is all being done off historical data, otherwise people will just game the system via google/wikipedia.

That being said, I definitely like the idea of having it be a social sort of tool where people compete on the winners.

Perhaps using the data you can somehow figure out how to generate random horse races based on real historical data and real horses and then people can really try to pick the winners instead of game the system for points through a simple google search.

Intersting idea, will add it to the list of potentials

Share it!

Are there any free/commercial sites with comparable datasets? I can see the data itself having a not-insignificant market value if it's of sufficient quality.

Unfortunately it is scraped from various sources on the web so it cannot be used commercially

Use it as a training data set for a bunch of machine learning algorithms and go pro, or discover that it's a really hard problem.

Share it!

Also consider ANN.

Share it on Bit Torrent! Large, accurate real-world datasets are difficult to find for the purposes of testing and experimenting with various machine learning algos.

Good idea, Ill create a torrent once I get the chance

Reply on news.ycombinator.com