Odds update every 5 minutes | Back to Election Betting Odds
Election Betting Odds FAQ
Why trust these odds?
Studies find that political prediction (betting) markets tend to be better at predicting elections than polls. Some reasons:
- Bettors take into account important factors besides polls.
- Unlike pundits, bettors put their money where their mouths are.
- People involved in the event might trade before news breaks publicly
- The "wisdom of crowds".
Why this site uses odds from Betfair.com
This site's odds are entirely from the site Betfair.com. After Intrade.com was shut down, Betfair became the only prediction market in the world that has a lot of trading and is efficient. It is based in the U.K., where regulations are less onerous. It does not accept American traders due to regulations. Betfair, however, expresses their odds in a very technical gambling format; hence the need for this site.
How people bet
American regulations are strict, but PredictIt.com has been allowed by the federal government's "Commodity Futures Trading Commission" and is easy to use. However, see the following section for why PredictIt's prices are systematically too high. Non-Americans also have the option of using Betfair.com.
Why this site's odds are not from PredictIt.com
PredictIt.com is good for Americans to trade on, but the odds are not efficient: As of this writing, the odds on PredictIt for who will win the Republican nomination add up to 191%.
For example: PredictIt says there's a 38% chance Rubio will win, a 32% chance Bush will win, a 22% chance Trump will win, an 18% chance Cruz will win -- just between those four candidates, PredictIt's odds add up to 110%. So we know the odds are off. The problem is caused by regulations that limit the amount people can bet on any race to $850. Betfair, which this site uses, does not have that problem; their odds add up to about 100%.
When/if PredictIt's odds become more efficient and get close to 100%, this site's main page will include the odds.
Likewise, if other prediction markets like Augur or Predictious take off, I will include those.
Note: PredictIt does not have the above problem for state-by-state betting; because of that, this site averages both PredictIt and Betfair for the state pages. Sometimes only PredictIt has a liquid market for a state, in which case only PredictIt is used and that is noted at the bottom of the state's page.
Why these odds are better than bookies' odds
Betfair is a market -- in other words, candidates' shares are traded just like stocks in the stock market, and the prices that result tell you the probability.
Bookies are similar, but the prices are not set by a market. This makes them more random, they often are only willing to take one side of the bet, and the "cut" is generally greater.
Is there enough trading on Betfair to give reliable odds?
So far, over $2 million has been bet on Betfair on the races covered here on ElectionBettingOdds.com.
In 2012, Betfair traders predicted the election outcome and called it with more certainty than Intrade.com.
How precise are these odds?
The average bid-ask spread is roughly 1%, meaning that a difference of less then 1% should generally not be viewed as significant. The markets are not always very deep, either; I would suggest assuming a "margin of error" of 3 or 4% (similar to the margin of error of polls, but the odds are still much more indicative of who will win compared to just looking at a poll.)
Comparison with other sites
ElectionBettingOdds.com is the only site that shows the odds as well as their recent percentage increase/decrease.
The NY Times' "Upshot" uses the "PredictWise" average. But if you read the fine print, their "prediction market" odds are actually an average of prediction markets AND polls.
Eli Dourado's Primary Guide is good and uses Betfair odds. It is useful if you are interested in contingent odds, and it also weights odds by liquidity.
Both of these sites, however, also use a formula for averaging the odds the skews them somewhat (detailed below.)
Methodology *very technical*
Short answer: This site simply averages the market "bid" and "ask" prices to come up with the implied odds.
Long answer: The odds on Betfair are not expressed as percentages and have to be converted. The conversion formula is 1/x, where x is the Betfair price. For example, on Betfair a candidate will trade at "50" -- that means the candidate has a 1/50 = 2% chance of winning.
ElectionBettingOdds.com converts the Betfair odds to percentages, and then averages them. Specifically, the formula is: ((1/Bid + 1/Ask)/2)
This generally gives intuitive results. For example, on Betfair, Chris Christie may have a "bid" price of "50" and an "ask" price of "25". In other words, the bid is (1/50)=2% and the ask is (1/25)=4%. It seems logical that the implied chance is the average: 3%.
But the sites mentioned above do not do that. Instead of calculating the percentages and then averaging those, they average the Betfair numbers first. Their formula is 1 / ((Back + Lay)/2) which in the Chris Christie example above would yield 2.66% instead of 3%.
This is a particular issue for low odds. For example, if a candidate on Betfair has a spread between "1000" (.1%) and "100" (1%), this site's formula would conclude that the odds are the average: .55%. But the other sites' formulas would compute the odds to be just .18%. In general, their formula skews odds towards 0% probability and 100% probability. Although it is theoretically possible that skewing the odds in that way makes them better at predicting events, there does not seem to be hard evidence for that notion.
So ElectionBettingOdds.com uses the simple averaging method because it most faithfully depicts the odds themselves without alteration.
One tricky thing is the rare occasion when liquidity is very low for a candidate. For example, one could imagine a candidate who has no chance -- so low that almost nobody is bothering to trade on the person's chances. In that case you might see a wide bid-ask spread. In the most extreme possible case, the bid might be 1% and the ask 99%. In that case the average would compute as 50%, which would be pretty clearly misleading for a candidate so obscure that nobody bothers to trade. It would not be good if this site suddenly put that candidate in the lead at 50% just because nobody was trading that person. The solution this site uses is: In cases where the bid-ask spread is greater than 10%, the bid alone is used, because it provides a solid lower-bound. In cases where the bid is greater than 50% and the bid-ask spread is greater than 10%, the ask alone is used. (These cases are extremely rare; the average bid-ask spread is around 1% and nearly all the time all of the candidates' spreads are well under 10%.)
More questions or comments? Maxim Lott can be reached on Twitter here.
Back to Election Betting Odds