PredictWise Blog

NCAA Basketball Tourney - Round 2 (Day 1)

Bookmark and Share

There are a few things you need to know about Round 2, Day 1:

1) You will be eliminated from Buffett's bracket challenge. There are 16 game today and my rough estimate is that there is a 0.5% chance of escaping day 1 intact.

2) Upsets predicted: Pitt (9) over Colorado (8) (does that count?)

3) Most likely upsets are the games where the higher seed is less than 60% likely to win: Harvard (12) is 42% to win today over Cincinnati (5), N.C. State (12) is 41% to win over Saint Louis (5), and Arizona St. (10) is 46% to win over Texas (7). Those 12 seeds are dangerous with North Dakota St. (12) at 39% to win over Oklahoma (5).

4) Full list of static, pre-game predictions.

5) Full list of real-time updating predictions. These are updating about every 2-3 minutes. I apologize in advance for any real-time quirks!

Now ... go have a productive day at work!

NCAA Basketball Tourney - Likelihood to Win Each Game

Bookmark and Share

This table will show the percent likelihood of winning each of the 67 games at about 8 AM on the day of the game (games at future dates will be updated on the day of the game of the game).

NCAA Basketball Tourney - Progession of Likelihood to Win

Bookmark and Share

This table will be updated daily and show the percent likelihood of winning the tournament for all 68 teams at about 8 AM ET each morning.

NCAA Basketball Tourney - First Round Blog Post

Bookmark and Share

Florida is the most likely team to win the NCAA basketball tourney at 16%, but that is not very high and that is not why you are reading this article. You want to fill out a bracket and here are my easy tips.

1) The smaller the group with which you are competing, the less deviations from expectations. Do not try to pick too many upsets, because you do not need too many upsets to win. Upsets, by definition, are less likely to happen.

2) There are more than enough surprises in my numbers for you to be competitive in your bracket.

South Region: Florida (1) is heavily likely to win the South.

East Region: Michigan State (4) is favored to come out of the East and Virginia (1) as second most likely. They play in the Sweet Sixteen. I have Villanova (2) as slightly more competitive than Iowa St. (3) in the bottom half of the region.

West Region: Arizona (1) is heavily likely to win the West.

Midwest Region: I have Louisville (4) is favored to come out of the Midwest with Wichita St. (1) as second most likely. Again, since this is a 4 versus 1 situation, they will meet in the Sweet Sixteen. That leaves Duke (3) as the most likely come out of the bottom half of the draw over Michigan (2).

These predictions are mainly derived from prediction market data, properly de-biased and normalized to 100%. Why do we need to de-bias? Because, the much less likely teams are heavily overpriced at approximately $0.005 on the $1.00 due to the favorite-longshot bias. We use the same de-bias technique shown in the Oscars. Well, actually, we use that same forecasting technique for everything, we just update the weights by domain! (That explains our scalability in providing accurate, real-time predictions for so many things). In further posts I will talk about the fundamental data for NCAA basketball in more detail.

Oscar Night!

Bookmark and Share

12:35 PM ET: My final thought for the night is that this is another case of pundits and insiders advertising a close event when the proper aggregation of data said it was not. As I noted on Twitter earlier, my acceptance speech is short. I would like to thank prediction markets for efficiently aggregating dispersed and idiosyncratic data. And, thank you to Miro Dudik and Deepak Pathak for crunching the data with me on the paper that determined the methodology for these predictions. And, thank you to Andrew Leonard and David Pennock for helping to make PredictWise a destination for data-driven prognostication! Good night … and, one last time, 21 for 24!!!

12:15 PM ET: There are two ways to judge accuracy. First, and most obvious, you want a small error. Correctly predicting 21 out of 24 categories and having non-negligible probability for the three other winners is indicative of a small error. Second, you want the predictions to be well calibrated. The top nominee in each category averaged 86.5% likelihood of victory. Thus, multiplied out over 24 categories we expected to pick the winner 20.76 categories. Winning 21 out of 24 means that our predictions were perfectly calibrated. If we picked any more categories we should have had higher probabilities of victory (i.e., if you get 24 out of 24, you should have 100% likelihood for the winner in each of the categories) and if we picked any less we should have had lower probabilities.

12:00 AM ET: 21 for 24.

11:54 PM ET: With the leading actress and actor categories we are up to 20 out of 23 with just the picture award left.

11:40 PM ET: 12 Years a Slave is now 7 percentage points more likely to win Picture since start of the show. Currenly 18 for 21 with 3 awards left!

11:30 PM ET: Excited about original screenplay ... that was the tightest of all of he categories! Now we are 17 for 20 with 4 awards left!

11:19 PM ET: Got the two music categories right; Gravity for score and Let it Go for song ... now 15 for 18 and breathing a little easier ... 6 more awards to go!

10:48 PM ET: Now 13 for 16, perfect in the technical categories … getting stoked for the big awards!

10:37 PM ET: Now 12 for 15 … just one more of the less prominent categories and then the big ones!

10:11 PM ET: Up to 9 for 12 after Gravity's double win in sound mixing and editing.

9:40 PM ET: Rough patch, now 6 for 9. 1 for 3 on short films.

9:29 PM ET: Now 5 for 6. Missed on Animated Short Film with surprise win by Mr. Hublot, but big wins with Frozen for Animated Feature Film and Gravity for Visual Effects.

9:02 PM ET: Two less prominent categories go our way, another Oscar for Dallas Buyers Club in Makeup and Hairstyling and an Oscar for The Great Gatsby in Costume Design.

8:43 PM ET: They started with a pretty secure category, best supporting actor ... stoked to start out 1 for 1 though ... phew!

8:20 PM ET: I am on the couch, I got a couple of beers, and I am looking forward to a fun Oscar night. I will update live here all night long, so stay tuned!