clock menu more-arrow no yes

Filed under:

Swarming MLB: AI Picks for August 29th

UNU, MLB

As the curtain falls on August in the 2017 Major League Baseball season, the weekly swarm of baseball fans convened by Unanimous A. I. can take pride in another successful month. Even with a sour note on Friday, the swarm’s results continue to show strong amplification of individual intelligence related to picking baseball game winners. Even though baseball can produce highly unpredictable results, the swarm continues to out-perform the individual users who make up this weekly group. The swarm’s picks for the season so far are also almost 60 games over .500.

For this week, we’d like to go into some more detail on how the Unanimous team establishes the swarm’s confidence in the various games it predicts. First, let’s look at the chart for tonight’s picks.

We typically list on the chart two pieces of confidence information: brainpower and the swarm’s “bet”. The “brainpower” is a measurement of the swarm’s cohesiveness. How quickly did it form a decision? How much dissent was there? In general, a more cohesive swarm that arrives at a decision quickly is a more confident swarm. So, when the group is swarming a prediction for Pittsburgh at the Chicago Cubs, there’s not much dissent going on, and it’s a confident, fairly boldly stated pick. Compare that to the swarm’s prediction for the Marlins and Nationals. Even though eventually the magnets get to a consensus, at the start the swarm seems unsure, and there are plenty of magnets pulling towards Miami. That tells our researchers that the swarm’s prediction is built on a shakier level of confidence.

The two replays above help us illustrate two other ways we measure confidence. Here, we’ve got two games with predictions expressed at 81% brainpower. The swarm doesn’t give us much help either on its bet confidence either, because both have the same $40 rating. So why is one 81% game our second-highest pick in the middle grouping, while the other is buried in the “high risk” zone? Looking at those swarm replays, we get some clues.

First, consider the game pitting Texas and Houston tonight in Tampa. The brainpower of 81% suggests some leaky cohesion, but looking at the swarm itself, we see what’s really happening. The swarm is torn on picking the Astros by one or two runs…but in much more general agreement that they think Houston will win. Our researchers can quantify that left-right momentum and assign a value to it that helps us to order the list. The other big apparent difference is the run differential noted here. The UNU Baseball Swarm thinks the Astros will win by two runs, while Cincinnati wins their game by just one. The Unanimous researchers have discovered that the expressed run differential is a very important expression of the swarm’s confidence. That 2+ run prediction and the momentum of the magnets towards Houston helps us to order it within the upper half of our predictions for tonight.

Unanimous AI runs predictive swarms every week for MLB, the NFL and more. Want those picks delivered to your inbox? Drop by UNU.ai.