Skip to Main Content

Case Study: How VQ’s Player Predictions Handled Turmoil with the Trail Blazers

By: Simon Smith

Carmelo Anthony signed with the Portland Trail Blazers on Nov. 19 after a year of not playing. The same day, and hours before Portland’s game against New Orleans, news surfaced that Blazers star Damian Lillard would miss the game due to injury.

Stats Perform’s new VQ player prop probabilities API quickly processed both pieces of news and republished predictions for all stats of every player on both teams.

One of the use cases for VQ is for betting odds compilers to calibrate and set their player prop betting lines. It has historically been difficult for them to account for the impact such injuries and signings have on predictions for every other player on the court. There are bottlenecks accessing the information and bottlenecks processing its implications, given the vast data involved and the number of players to process it for.

This is why most non-VQ sportsbooks we checked after the news broke that day had either Lillard markets still up, or all Blazers and Pelicans player prop lines suspended – in some cases for hours, in others permanently. At best, only one or two player markets were available. Very few produced any options for Carmelo until shortly before tip-off. The lack of player prop coverage for this game starkly contrasted the other games on that day’s slate.

The challenge: Anthony hasn’t played for a year and had never played with this group of players: How will he perform relative to his new teammates and his opponents?

VQ is able to quickly generate predictions for how each player will perform in a scenario without Lillard and with Anthony.

Stats Perform’s depth of data means we objectively understand through AI not just how the relevant teams have been playing but how players perform from game to game, year to year, and even how a player performs in his 10th season vs. his 11th season.

Our model learns about players with similar ages, heights, weights and playing style attributes.

It understands how the player’s statistics should be scaled based on likely lineups, it understands how career performance increases or decreases with age, and it deeply understands the expectations of someone in that specific position.

This applies for established pros like Lillard and Anthony and even for rookies: Our deep college data archive enables VQ to predict the impact of a rookie who has never set foot on an NBA court.

The other bottleneck is anticipating that a lineup will change, which is the domain of our expert, experienced journalists judiciously reacting to player news. This team is made up of experts who have built careers on knowing which sources to trust and how to read between the lines.

The result:

The combination of deep data, expert journalists and sophisticated AI means we can help our sportsbook partners more confidently publish lines for all players and very quickly after news breaks, whether it’s the morning of the game or a shootaround ankle roll.

VQ is an engineering, AI, and process story: We built the model to adjust quickly to breaking news and to quickly assess the impact of lineup changes so the news’ impact can be quickly reflected by sportsbooks to their customers in the form of open betting lines.

On this occasion, we quickly republished prop predictions for all players on both teams within minutes of the changes being confirmed. VQ predicted Anthony’s line at a 51.56% probability of being under 10.5 points. He finished with 10 points.

Bryan Shumway is the product manager for Stats Perform VQ, a new advanced player props prediction and probabilities API which uses proprietary AI-driven data.