PRO Chess League Metrics
Last Tuesday's PRO Chess League 2018 debut showcased some of the most exciting online chess that fans have ever witnessed. The live chess coverage ran for over 12 hours, with live commentators and thousands of fans watching. 32 teams of battled it out to see who could start off the 2018 season with a victory. Lots of triumphs and heartbreaks came down to the last few minutes. How could we add excitement to the already thrilling PRO Chess League from last week? You guessed it, live odds!
Back in November, Chess.com introduced new metrics for the Speed Chess Championship. With the support of Danny Rensch and Gerard Le-Marechal (head of fair play and game statistics), I have developed a new system to predict the outcomes of the PRO Chess League matches as they happen live!
After the Speed Chess Championship concluded, I began work on a new system to predict the outcomes of the PRO Chess League matches. Here's an overview of the system, and a worked example from the Minnesota Blizzard vs. St. Louis Arch Bishops match from week one.
Each team has a four player roster, in an all-play-all format. The pairings format can be found at the PRO Chess League Rules page.
Monte Carlo Simulation
To predict the outcome of the match, I have created a Monte Carlo Simulation that runs 1000 random events and tabulates the results of those events. There are a few input factors that go into those simulations. The main three factors are FIDE ratings, color allocation, and draw rates. Heading into the match, here were the predicted odds of each outcome:
In addition to the match odds, we are also able to compute the likelihood of every possible score outcome, ranging from 16.0-0.0 to 0.0-16.0. Here were the most likely results pre-match:
One of the benefits to the Monte Carlo Simulations to determine our match odds is that we can re-compute the odds after every single game. As the match gets down to the wire we will have updated match odds to display for all of the ongoing matches.