Show HN: Predicting the winner of the US Presidential Election | Hacker News Reader

Show HN: Predicting the winner of the US Presidential Election

7 pointssendos11 comments

In 2008, I did some calculations to estimate the probability of Obama or McCain winning the election, using state poll results, and plotted the probabilities vs time, right up to the election. The results were quite interesting.

So, this year, I'm doing the same for the 2012 elections, and have put it all up on:

http://prespredict.com

I include both 2012 and 2008 results, and make an attempt to see how they correlate with campaign events and national news stories. There is also a section where I describe the methodology.

Let me know what you think

[P.S. I used bootstrap to get the site up and running quickly, but I'm not too pleased with how it looks. Any suggestions or templates I can use to improve the look?]

View on HN

If you're unhappy with the default Bootstrap look, http://bootswatch.com/ has nice free themes. It made a world of a difference when I ditched the default and used one of their themes on a current project.

99.7% for Obama in March, and 82.6% now?

According to you, the odds of Obama losing are now higher by a factor of 58. That is huge. What has happened in the last four months that justifies such an overwhelming shift?

> According to you, the odds of Obama losing are now higher by a factor of 58

Just to be clear, this is not strictly speaking "according to me", that is I don't use my personal opinion to come up with these numbers. I just take daily state poll results from electoral-vote.com and pass them through some equations. No human intervention.

Having said that, I think what was happening in the March timeframe and before was that the field of Republican potential nominees was very uninspiring to Republicans, which can be seen by how often we saw a new front-runner, only to lose their front-runner status and usually bow out shortly thereafter.

So, perhaps, against this amorphous group of uninspiring candidates, Obama attracted a lot of voters. Then, as the Republicans started settling on one candidate, and started to slowly accept him as the nominee, he now has a much better chance against Obama.

If your predictions are so good, why not keep the data to yourself, get on Intrade, and make a ridiculous amount of money?

Even if we assume my predictions are very accurate, I'm not familiar enough with Intrade to know whether it's mostly guided by fundamentals vs technical analysis.

That is, Obama's value may go up because many people think it will go up, due to some technical indicators, and not necessarily because Obama's probability of winning went up.

This doesn't make sense to me.

Indeed, the more irrational you believe Intrade's users to be, the more incentive you have to use it. Same for the stock market, or any other market.

Does this always work? If you think that company X should be valued at $10 billion, but is currently valued at $5 billion because the market is being irrational, are you claiming that it's always a safe bet to invest in this company? Are you claiming that in the long term the market becomes rational and will surely value the company at the value you think it is worth?

You're missing your parent commenter's point. He/she means to say if more people are wrong then the odds are stacked higher against the candidate you expect to win therefore the payout in favor of your candidate would be higher. Sorry I think my comment is even more confusing but I hope you arrive at this conclusion anyway.

Clickable: http://prespredict.com

What are the main differences between you and your methods and FiveThirtyEight?

I'm not too sure, since I can't find the exact model used by FiveThirtyEight. If any of you can point me to a post of theirs where they describe their model in detail, that would be great.

From the few things I have found out, some differences are that (1) they take into account correlations between states, which I don't as of now, but plan on adding and (2) they use Monte Carlo simulations whereas I have a closed-form expression for calculating the probability.

Beyond the above two differences, I'm not sure what model they use to convert the poll results into final probability estimates.

FYI, in case you haven't seen it yet, my methodology is described at http://prespredict.com/methodology.html

Reply on news.ycombinator.com