I love odds and win probabilities, especially when those numbers are driven by market forces. I’m a believer in the ‘wisdom of crowds,’ particularly if the risk/reward is monetary. Unfortunately, there is specific regulatory language around betting on sports events that prevents a full-fledged sports prediction market based in the U.S. (that will have to wait for now).
GamePredict is a social experiment. How accurate would a sample of sports fans be about their favorite sports events? Given all users’ predictions, the GamePredict system will generate both a score/rating of accuracy and a point spread or line. GamePredict’s scoring and line algorithms were designed by an expert statistician and decision theorist. Users make precise predictions on sports events, and the system will rank every user against each other depending upon how accurate they are over time. The GamePredict community line is a weighted average of the community’s predictions for a game, weighted by the user’s history of accuracy.
I expect GamePredict to have 3 types of users:
Every type of user and sports fan is welcome, and I hope you enjoy this ‘gameified’ prediction market. If you have any questions, comments, or feedback, please let us know. You can also contact me directly at admin [at] gamepredict [dot] us.
Why the point spread?
The point spread, commonly referred to as the line, is a precise measure for how favored a team is over the other at a specific venue. It’s essentially an odds or probability measure, in the context of the game, of what the market expects to happen if two teams were to play at a specific location. The precision of the line (as well as the implied win probability) and how it moves over time, is especially interesting. We will track those movements for both the Vegas and GamePredict communities or ‘markets.’
How is this different than Vegas?
The current betting market is a fascinating study. Evidence shows that sportsbooks can maximize profits by strategically setting lines where they can exploit the market’s biases, rather than setting that line at where they actually think the outcome will be. At GamePredict, we are interested in precision in predicting the final margin of victory. You are not making a 50/50 guess (cover or not cover) on a specific line; you are guessing what that line will or should be. The GamePredict algorithms are designed to reward and punish predictions based on how accurate they are.
Why even show the Vegas lines?
While many argue that the Vegas line is merely a perception tool, it still remains the most accurate indicator of future sports events.
What does ‘Vegas’ mean anyhow?
All of the legally sanctioned U.S. (brick and mortar) sportsbooks are located in Las Vegas, NV. With the advent of the Internet, most of sports gambling is now done online on the ‘black market’. However, all of these sportsbooks operate very similarly with very similar prices. For many, the term ‘Vegas’ could refer to any sportsbook or oddsmaker literally in or out of Vegas.
Which sportsbook are your ‘Vegas’ lines from?
BetOnline. They consistently release their lines earlier than others. In steady-state, they're practically the same as all other lines.
How is the GamePredict community point spread/line generated?
The community line is a weighted average of every prediction for that game, weighted according to the predictor’s GamePredict score. The more accurate you are, the more impact you have on the community line. Hopefully this will produce a feedback loop that can help us generate more accurate lines over time.
How should I interpret my GamePredict score and rank?
Your score and resulting rank is relative to other users in the community. You will see how your score and rank changes over time.
You can also measure your performance on a game-by-game basis; your predict button for past games is color-coded: , , , and .
What if I make a few bad guesses? Will that kill my score?
Everyone will make terrible guesses, and your score is relative to others, so no. Plus, the scoring algorithm is designed to lessen the impact of extreme outliers.
Any credit for picking the right team? Why can I choose a point spread of 0?
GamePredict is all about precision. Being only 1 point off is impressive whether the resulting line is -10 or -1. If you feel a game is truly a toss-up, then give it even odds.
I made some predicitons but I don't see myself in the Leaderboards?
Your score is not calculated until you've made at least 5 predictions in at least one league.
Some of your lines are way off. What gives?
This is merely a small sample problem. As the site gets more users, and more predictions, the algorithms are designed to become more accurate over time. Please invite your friends! If someone picks a -99 line to be funny and they are the only prediction on that game, then the GamePredict line will inevitably be -99. However, that funny person’s GamePredict score will then become so small that he/she will effectively have 0 impact on a line going forward (until they start becoming more accurate).
How are your win probabilities calculated?
The Vegas win probability is derived from the moneyline. The GamePredict win probability is derived from the community line. We created functions for each sport that converts the point spread into a win probability, based on historical betting data. By comparing past spread and moneyline trends, we estimated the given win probability.
Is this like Fantasy Sports?
In some ways, perhaps. I have a love/hate relationship with Fantasy Sports. I hate the Fantasy aspect; many Fantasy Sports fanatics end up caring more about individual player performance than they do about the actual game. They create their own fake team that obfuscates the real reason for why the game is even played to begin with. However, I love the Sports aspect; the data, analysis, strategy, competition, excitement, disappointment, etc.
Why only football and basketball?
We are starting with the most popular U.S. sports and hope to add others in the future.
When are the leaderboards updated?
The leaderboards are updated late each night after the day’s games have ended.
What does a premium user get?
For now, the premium user can create private groups, invite friends to compete, add specific games that the group must predict, set up rules for how long the contests last (weekly, daily, etc) and the prize for winners, etc. It also tracks the winner of each contest. The same scoring algorithm is applied to each group. These groups are fully customizable and self-regulated.
Why does every private group participant have to pay?
We will continue to develop more features and benefits for our premium users. In some ways the monthly/annual fee will also help weed out those of your friends that aren’t really interested in competing. Suppose you were competing for more than bragging rights; your friend who is subscribed already has some skin in the game, is serious about competing, and is more likely to “pay up.”
How much does a premium membership cost?
Only $1.67 or $2.00 per month (depending upon annual or monthly subscription).
What other features are you considering?
We’d love to get some feedback from the community. Some ideas we have are: more predictions (e.g. guessing the vegas line beforehand, total points, playoff series winners, etc) as well as some more interesting data, like a calculated power ranking for every team based on the community’s aggregate predictions.