DavidMRothschild on April 18, 2014 @ 3:41PM
Introducing a new section called www.PredictWise.com/NBA. We are going to start building these for all of the major perennial events. For the playoffs this section will include:
1) Updating table of the likelihood of winning the NBA championship and conference.
2) Updating table of the likelihood of winning the next game in each series and the series.
3) Static table of the likelihood of winning each game on the morning of the game.
4) Blog posts that will take a deeper dive into interesting things.
What is interesting right now is Brooklyn (6) v. Toronto (3) series. The prediction markets are more bullish on the Brooklyn Nets than the panel of experts we are consulting over at ESPN. That is the only major difference between the markets and the experts.
DavidMRothschild on April 09, 2014 @ 9:30AM
I am stoked to introduce a new game that was developed at Microsoft Research for the Indian Elections. The game is easy to find at aka.ms/elections and provides the ability to wager on any of the states in the upcoming Indian elections. The game will produce the most accurate predictions of the Indian elections by allowing the crowd to input their idiosyncratic knowledge of the different parties in the different states through their wagers and an extremely robust market maker aggregate those wagers. Just a few of the cool features:
1) Ranges: all bets are on ranges of seats for a given party in a given state. Or they are on the combinations of ranges for multiple parties in a given state.
2) Combinatorial: it is a fully combinatorial system so that every time you pick a range it affects all of the other possible outcomes, not just for that party in that state, but all parties across all states.
So, if you have some knowledge of the Indian elections, please give a try.
DavidMRothschild on April 08, 2014 @ 9:04AM
Previous Posts: Likelihood of Winning, Game by Game, First Round, Second Round (Day 1), Second Round (Day 2), Third Round (Day 1), Third Round (Day 2), Calibration, Sweet Sixteen, Elite Eight, Final Four
DavidMRothschild on April 05, 2014 @ 1:13PM
There are a few things you need to know about the Final Four
1) Florida (1) out of South is about 75% to beat Connecticut (7) out of East.
2) Amazingly, Kentucky (8) out of Midwest is about 51% to beat Wisconsin (2) out of West. For those that do not follow college basketball closely the fact that Kentucky is 8th in their region and (slightly) favorite to win should be surprising. But, Kentucky takes advantage of a peculiar policy of the NBA that does not let high school seniors jump directly to making money, but forces them to wait until they are 19 and/or played 1 year of high school basketball. Some schools are reluctant to take advantage of the rule and "enroll" players who will go to the NBA after one season, but not Kentucky. Thus, will a team full of freshman, they have the talent, but not necessarily the cohesion early in the season. Thus, the bad start and the low seed, but the strong run.
DavidMRothschild on March 29, 2014 @ 10:00AM
There are a few things you need to know about the Elite Eight:
1) I told you we would get 5 of 8 games right (in the binary sense) in the Sweet Sixteen and we did. If only we knew which 5 it would be we would have gotten 8 of 8 game right! Of course, those were some tight games. We had Michigan (2) has our main upset watch and they pull it out by 2 points over Tennessee (11). And we had Louisville (1) carrying the rivalry game against Kentucky (8), but they lost a tight one 74 to 69. That being said, we were happy to see Michigan St. (4) beat Virginia (1) as expected and Arizona (1) handle San Diego St.. Going into the Elite 8 we have 3 of 4 pre-tournament Final Four teams still alive: Michigan St. (4), Florida (1), and Arizona (1).
2) Florida (1) is a heavy favorite over Dayton (11). Arizona (1) is a tighter favorite over Wisconsin (2).
3) For tomorrow's games Michigan St. (4) is going to be a heavy favorite over Connecticut (7) and Kentucky (8) will be a slight favorite over Michigan (7).
DavidMRothschild on March 26, 2014 @ 1:59PM
There are a few things you need to know about Sweet Sixteen:
1) The favored team has an average likelihood of victory of just 61% in the Sweet Sixteen. This means we only expect to get about 5 of 8 games right (in the binary sense). We have been extremely well calibrated so far this tourney, so expect an interesting Sweet Sixteen. But, not all games are equally uncertain ...
2) Virginia, the top ranked team in the East, if favored to lose. Not by much, but Michigan State (4) is 52% to win the game against Virginia (1). That is the only game where we have the lower ranked team actually more likely to win the game.
3) Florida (1) in South, Arizona (1) in West, and Louisville (4) in Midwest, our other pre-tournament favorites to get to the Final Four (along with Michigan State in the East), are all favored with between 65% and 75%. To single out the Louisville and Kentucky game, Louisville is as highly likely at 68% to beat the surging Kentucky squad.
DavidMRothschild on March 24, 2014 @ 10:14AM
When we say there is a 75% likelihood of an outcome occurring, we mean that if the event were to happen 4 times, 3 times the outcome would occur and 1 time it would not. Brier scores, mean square errors, mean absolute errors, all have their place. But, errors are one thing and calibration is another. We are extremely proud of how we our likelihoods are calibrated and this year's NCAA tournament is no exception.
Below I have placed all 52 games into buckets based on the likelihood of victory for the team we predicted was most likely to win: 50-59.9, 60-69.9, etc., The x-axis is the average prediction in the bucket and the y-axis is the percent of games where the most likely team won. What the chart illustrates is that the predictions are nearly perfectly calibrated. Overall, the average probability of victory for the most likely team has been 73% and, of those 52 teams, 38 have gone on to win or precisely 73%.
DavidMRothschild on March 23, 2014 @ 8:55AM
There are a few things you need to know about Round 3, Day 2:
1) Again, we had a well calibrated day with two upsets out of eight games. We had one our two most likely as Villanova (2) losing to Connecticut (7) and we had Dayton (11) at 27% to beat Syracuse (3) who came limping into the tournament.
2) Upsets predicted for today: North Carolina (6) is a virtual toss-up against Iowa State (3) at 48% to 52% respectively to win. Similarly, Creighton (3) is just 59% against Baylor (6). The 3-6 matchups are the tightest in seeds and tightest expected outcomes.
3) The most likely upset outside of the 3-6 matchups is Wichita State (1) falling to Kentucky. We have Kentucky at 35% to knock out the undefeated Shockers. Interestingly, the next closest game is Virginia (1) at just 71% to advance to the Final Four. Of course, we are not that concerned about these games in the long run, because we both Virginia and Wichita State falling to very strong 4th seeds (Michigan State and Louisville respectively) in the Sweet Sixteen ... if they make it!
DavidMRothschild on March 22, 2014 @ 8:11AM
There are a few things you need to know about Round 3, Day 1:
1) The "big call" yesterday was Tennessee (11) over UMass (6) at 62% and Tennessee crushed UMass 86 to 67. The other double digit upset we said had a high probability was Stanford (10) at 40% to beat New Mexico (7) and they did at 58 to 53.
2) Obviously, Duke (3) was heavily favored to beat Mercer (14), but while we did not know which upset was going to happen, we knew upsets would happen! So far we have had 13 games where the favored team was 80% or greater (average 94%) and one upset in that group is about in expectation.
3) Actually, we are perfectly calibrated through 36 games, with average probability of the favored team at 75% we have had 75% games, exactly, where the favorite won. You can see the full list of static pre-game predictions here.
DavidMRothschild on March 21, 2014 @ 9:14AM
There are a few things you need to know about Round 2, Day 2:
1) If you have not alread, you will be eliminated from Buffett's bracket challenge. There are 16 game today and my rough estimate is that there is a 1% chance of escaping day 2 intact (for those lucky few who escaped day 1!).
2) Yesterday went just as expected: Pitt crushed Colorado, Harvard and North Dakota St. won, while Saint Louis and Texas both won in thrillers. And, the round out the light day of upsets, we had Dayton at 28% to overcome Ohio State.
2) Upsets predicted for today: Oklahoma State (9) is favored over Gonzaga (8) to the tune of 58% and the big call is Tenessee (11) at 62% over over UMass (6).
3) Most likely upsets are the games where the higher seed is less than 60% to win: Memphis (8) is 59% against George Washington (9) and New Mexico (7) is just 60% against Stanford (10).