The Wild bowed out early yet again, the Flames looked every bit the team with a <1% chance of winning the cup, and the Model went 7 for 8 in first round series predictions. The big (and only) miss came on the heels of the Chicago Blackhawks, who apparently decided they would rather be golfing than playing hockey. Despite the addition of Bruce Boudreau, the superstar-less Minnesota Wild proved once again that they lack a star goal scorer needed to make a deep run in the playoffs, such as a Vladimir Tarasenko. While Round 1 presented zero game sevens, the Model predicts we might see a nail biter or two in the second round.
NBA Role Probability Model is used to predict the likelihood that a given college basketball player becomes an NBA All-Star, Starter, Bench player, or does not make the NBA. The model uses individual college basketball season-long box score statistics, team-level statistics (e.g. strength of schedule), physical measurements, high school scouting ranking, position, and age/experience to predict the probability of each NBA role. For more detail on this model, see here. In the following table, you can find our predicted probabilities for the 2016 NBA Draft prospects landing in each category:
Continue reading NBA Role Probability Model 2016
With the 28th overall pick in the 2016 NBA Draft, the Sacramento Kings selected Skal Labissiere, who perfectly fit the bill of a modern-day NBA big man: nearly 7’0″ and roughly 215 lbs, armed with a 7’3″ wingspan and a smooth jumper. However, Labissiere’s production in his one season with Kentucky was extremely minimal. Though he was Draft Express’s preseason number 1 overall pick, he averaged just 6.6 points per game, 3.0 rebounds per game, and 1.6 blocks per game while playing a measly 15.8 minutes per game. Given this production, Labissiere seemingly didn’t warrant any draft pick at all. But not only was he drafted, he went in the first round. Why? Potential. The idea was that Labissiere could develop his tantalizing tools and become the all-star caliber player many thought he would be prior to his time at Kentucky. While that would have been a great outcome, it was still more likely that Skal would not reach that all-star potential at all. With the combination of this potential and an unproven track record, it seemed that Labissiere’s role in the NBA would be either be an all-star or a bench warmer—or maybe even out of the league! Contrast Labissiere with Frank Kaminsky, who earned the Wooden Award in his senior season at Wisconsin. Most did not envision Kaminsky as an all-star, but rather a 4th, 5th or 6th man in the NBA. He had more polish than Labissiere, but a lower ceiling. These two seven-footers had very different profiles coming out of college.
In order to capture the likelihood that players like Skal Labissiere become NBA all-stars and players like Frank Kaminsky become NBA starters, we have created an NBA Role Probability Model that seeks to predict what role an NBA prospect will play in the NBA. Adding this to our previous prospecting work, we now have three components to help evaluate NBA prospects:
(1) PNSP, which answers the question, “How valuable will a player be?”
(2) Similarity Scores, which tell us about playing style by comparison to similar players
(3) NBA role probability model, which answers, “what roles might this player fill in the NBA?”
For a glimpse of how the 2016 draft class scored on this model, check out NBA Role Probabilities for 2016 NBA Draftees here.
Using the average probability from our NHL Playoff Models, we have run a large-scale simulation for the 2017 NHL Playoffs. This allows us to estimate how often each team makes the semifinals, conference championship, Stanley Cup Finals, and ultimately win the Stanley Cup. In order to do this, we took each series and generated a random number between 0 and 1. If the random number is less than our predicted probability of the home team winning, then we advance the home team as the winner of that series (and vice versa if the random number is greater than our probability). For example, Model 1 has a probability of 0.74 that the Chicago Blackhawks will beat the Nashville Predators. If the random number is 0.95, we would pick the Predators to advance, and if the random number is 0.55, we would pick the Blackhawks to advance. This methodology was applied to the entire NHL Playoff bracket, generating 10,000 brackets and 10,000 Stanley Cup champions.
Our NHL playoff models use team-level and individual player statistics to predict the probability that a team will win a given series. We build our bracket by advancing the team with the higher win probability for each series. For example, this year, our models give the Chicago Blackhawks a 74% chance of winning their first round series against the Nashville Predators, so we advanced the Blackhawks to the following round in our bracket. While we utilize multiple models to generate predictions for each series, the bracket below represents the average probability of all models for the 2017 NHL Playoffs. For more information on our NHL Playoff Models, see our methodology article here.
The NHL Playoff Model uses team-level and individual player statistics to predict the probability that a given team will win a series (rather than predicting each game individually). Theoretically, one would think predicting a winner of a series would be easier than predicting the probability a team wins a single game, but many things can happen throughout the course of a series, most significantly, injuries that make predicting the series winner difficult.
The following tables display our predicted point spreads and point totals for all 2017 first round games, separated by region. The play-in teams are marked with an * and will be updated as needed. As an example of how to interpret the tables, in the second row of the East Region table, our models have Wisconsin (-2.8) beating Virginia Tech by 2.8 points, and Vegas has Wisconsin (-5.5) beating Virginia Tech by 5.5 points. Positive numbers for both spread columns indicate that Team2 is favored. Additionally, our models have the total combined points scored in the Wisconsin-Virginia Tech game as 144.1 while Vegas has 137.
For more on the historical performance of our point spread and point total models, click here.
Do you think you can beat the Model? Well, here is your chance to prove you are smarter than the Model. Enter a March Madness bracket into the group Model 284 for a chance to take on the Model. The winning bracket will receive a $25 gift card to Raising Canes Chicken Fingers along with the bragging rights of beating the Model.
Select any 2016 drafted NBA player from the dropdown menu to view their top 10 most similar college basketball players historically. Similarity Scores provide insight into how a player will translate to the NBA based on how similar college basketball players in production, physical measurements and experience have performed in the NBA. For a background on the calculation of the similarity scores, see here.
“Buddy Hield is the next Stephen Curry”
“Brandon Ingram is a poor man’s Kevin Durant”
“Andrew Wiggins’ upside is Carmelo Anthony, but his floor is James Posey”
So often when we talk about NBA players, we do it through comparisons to other players, and with good reason—comparisons are a good way to quickly convey lots of information about a player. For example, if I tell you that a player had a Box Plus/Minus of 7.8 last season, you might get a vague idea of how good he is. If I then tell you that player had 19.5 points per game, you might have a slightly better idea, but it’s still far from the full picture. But if I claim that this player is the next Chris Paul, it immediately brings to mind an idea of not only how good he is, but also his strengths, weaknesses, and overall playing style. Maybe your mind also queues up a mental highlight reel of Chris Paul-like plays, for good measure. Comparisons quickly give a complete picture of a player which would otherwise require taking the time to slowly digest each number in his stat line.
Also, they’re lots of fun!
We’ve created our own set of similarity scores to make comparisons using math for the purpose of prospecting players coming out of college. Our goal is to produce a useful complement to our PNSP model. Where our PNSP model answers the question, “How valuable will this player be?”, our similarity scores aim to answer the question, “Who will this player be like?”