Today’s post is over at the new and, yes, improved Edgehogs.com. Our launch was a couple of months ago. Since that time we’ve got a good chunk of games and predictions for those games. Patterns are beginning to emerge.
Both Experts, true sports pros and professional bettors, and regular Users make predictions for the NFL, College football, and Mixed Martial Arts—and now College basketball and NHL hockey (maybe even the NBA, if they ever begin playing). We track and rate the picks based on a secret-sauce formula that accounts for the difficulty and number of the picks.
For the NFL it’s not surprising that all Experts aren’t better than all Users, but the best pickers are Experts. But for College Football, Users do almost as well as Experts.
Head on over to see the analysis.
Sports “news†consists of reporting the recent past (The Indians lost to Boston three to two last night), predicting the near future (The Indians will start a three game series in Detroit tonight), predicting the not-so-near future (Boston will win the Stanley Cup this year) and telling stories from sports history.
The first two are usually reliable and provide some utility. Historically they were covered by the box score on page two of the sports section. Predicting the not-so-near future is futile and for me not very interesting. Stories from history can be entertaining and educational if not always completely truthful.
If one wants to predict the future, better to do it at the track or Vegas where you can profit from it and where the pile of chips remaining provides a reliable running score. I guess that for some, predicting the not so-near-future is a sport in itself.
For me, the skill and entertainment of sports is in the game … and the SI Swimsuit edition.
Briggs,
accounts for the difficulty and number of the picks
Would it be revealing too much if you were to define what you mean by “difficulty”? Obviously, you don’t mean from, say, a comatose state. Offhand, I would think making a correct pick when P(win|whatever)=0.5 would be the most difficult. But if that’s what you mean, you would need to have the probability of winning already. That would make you a super-handicapper in your own right.
Speed,
The best handicappers on thoroughbred races make their money by selecting bets vs. selecting the winners. In fact, handicapping contests use profit as the measure of success. In one, I lost even though I had more picks than the winner. I lost by $0.60.
I am curious. If you run the same contest next year will the same experts score near the top? If a different group score high each time then you are showing that expects make broader predictions as a group, not better.
Assuming Dr. Briggs’s rating is showing true quality, you can see (at least for NFL) that experts dominate the upper 25-30%. It’s hard (an understatement) to tell from the graph if more experts are better than the number of users since the user/expert ratio isn’t given. Knowing nothing else for any given randomly selected pair (user, expert) go with the expert.
DAV,
Good point. The numbers are about even, Experts and Users. These are frequency histograms. Most of the people in the pics have 100s of picks.
William Sears,
We have data going back about a decade. Accuscore, for instance, has been in the top consistently. We only have ordinary User data for a season, so I’d expect more fluctuation.
DAV again,
Be glad to. I just didn’t want to make the post too long. Difficult is largely measured in terms of odds, using the “official” Vegas line. Odds are converted to probability. Events which are close to 50/50 are more difficult than events which are (say) 5/95 or 95/5. That is, if you pick competitor A who the odds say had a 95% chance of winning, you will gain in your X-Rating, but not as much as if the odds say you have a (say) 5% chance. Picking long shots correctly boosts your score the most.
But we also look at other measures of game difficulty (score, if there is one, etc.). And we also look at the odds. Just how I’ll reveal in another post.
Speed,
What’s interesting is that most people pick around about 65% correctly, both Experts and Users. Experts are better at picking more difficult games, which accounts for the difference.
most people pick around about 65% correctly
That’s pretty impressive. In horse racing the crowd favorite wins roughly 30% of the time — about the level of a decent handicapper. Getting 65% accuracy in horse racing puts one in the top 1-10%. Maybe that 65% user/expert accuracy says something about the relative ease of sports prediction?
Selected at random, the thoroughbred win odds pretty much match the win ratio. Have you ever evaluated the Vegas Line performance?
@ DAV: The difference between the track and NFL results might be a motivational thing as far as competitors are concerned. After all, “They Shoot Horses, Don’t They?”
DAV,
But there are about 8 horses per race and only two competitors in the NFL or MMA.
A few points…
1. there is no “official†Vegas line. Each casino can have a different point spread, and the point spreads at any given casino can vary over time. Therefore your degree of difficulty for any game is subjective and may change over time, so when the selection is made may be a factor as well.
2. Briggs is talking about picking who will win a game straight up, when actual betting one uses either a point spread or the money line. And the point spread has money odds attached to it, usually (but not always) risking 11 to win 10.
3. Briggs says “most people pick around about 65% correctly” but that is straight up. Very few can hit anything close to that against a point spread.
robert burns,
1. Exactly so. We actually store a time line of lines from a prominent source. You get the odds that are in force at the time of your pick.
2. Yes, picks are straight up, and the odds are money line. We use the moneyline to award X-Bucks, just as in real bets. We have the spread but don’t yet allow picks against the spread; this is coming.
3. Very true; point spread picks are much closer to 50%, as you’d expect.