The Correlation slider controls how strongly SaberSim favors correlated players when it builds lineups. When you build lineups on SaberSim, we build the best possible lineup for a given way the slate may go. The correlation slider encourages the Sim Optimizer to combine players together with positive correlations in those lineups, and avoid players with negative correlations.
Turn it up and lineups lean harder into stacks. Turn it down and lineups use more one-off plays that don't depend on each other.
These correlations already come from the simulations. When a QB and WR score together in the same simulated game, they appear together in the lineup built from it. The slider changes how much SaberSim emphasizes that pull. As an example, imagine a set of simulations where Joe Burrow scores 20 points, Ja’Marr Chase scores 19, and Justin Jefferson scores 21. With no correlation, the Sim Optimizer would build the lineup with Joe Burrow and Justin Jefferson. They were the two best performing players in the sim.
With increasing correlation numbers, the Sim Optimizer becomes more and more likely to use Ja’Marr Chase instead of Justin Jefferson, despite the fact that Jefferson scored more, because Joe Burrow and Ja’Marr Chase are positively correlated on average.
The correlation slider is backtested for what we believe is optimal for a given sport and slate size.