Microsoft claims Bing can predict the NBA Draft -- do you agree with the predictions?
I am a huge fan of the NBA -- the New York Knicks to be precise. Basketball is a great sport as it does not require too much equipment. You can visit any park with a hoop and bring nothing but the basketball and spend hours just fooling around. This is in contrast to something like baseball, where everyone needs a glove and you have to find a bat, or hockey where everyone needs a stick. This makes basketball the ideal global sport, especially in low income areas.
Besides watching games on TV or live at Madison Square Garden, there is something else I love -- the draft. If you aren't familiar, every year NBA teams are ranked from worst to best based on record, except for the top 3 which are chosen by lottery. Also, since draft picks can be traded, this can alter the list too. Hell, some teams don't even have picks as a result of trades, such as my Knicks this year. Then, one by one, teams select eligible amateur players to join their teams, for a total of 60 new players (30 in the first round, 30 in the second round). Today, Microsoft announces that Bing has gained the ability to predict the NBA Draft. The question is, how accurate will it be?
"Harnessing the wealth of signals available to Bing and training the model on results from past seasons, we estimate the relative potential of amateur players in this year’s draft. We take publicly available data such as player statistics, player profile information, team needs, combine statistics, and expert evaluations to generate a projection of a basketball player's on-court capabilities. We then add search data and social signals into the model to determine the projected associations between teams and players and player trends to reach the final ordering", says Walter Sun, Principal Development Manager, Bing Predicts.
Sun further explains, "unlike the voting shows, where we had high confidence in the predictions and final rank order, because the outcomes depended heavily on popularity and were determined incrementally one night a week over one to two months, the NBA draft picks are made by team management and happen over the course of a few hours, where selections are made in minutes and trades happen in real-time. As a result, reaching 100% accuracy for each of the top 14 picks, while ideal, is realistically less likely. For the cases where we don't get the exact position, our goal is to be as close as possible to the exact position. We think our model will compare favorably in relation to other expert picks".
To 100% correctly pick the top 14 picks is a near impossibility, especially by computer algorithm, so Microsoft is smart not to claim perfection. The NBA draft is very different from the NFL for example, where teams often pick based on needs. In the NBA, more teams employ taking the BPA, or Best Player Available, even if the player duplicates the skill-set of an existing player. One of the reasons for this is because of the 1984 draft. You see, the Portland Trailblazers infamously drafted for need, choosing the lackluster Sam Bowie because of his height over Michael Jordan. Yes, that Michael Jordan, arguably the best player of all time.
If you want to see Bing's predictions, simply go to Bing.com and search for "NBA Draft Predictions". Tell me if you agree with them in the comments.