Using the "Correlation" slider

What are correlations?

Correlations measure the impact one player’s performance has on another’s. Players can be positively or negatively correlated.

Examples

Batters on the same team in baseball are positively correlated because when one player does well, it creates more at bats for the others and when somebody gets an RBI another gets points for scoring a run.

A quarterback and their opposing defense are negatively correlated, because the QB gains points for throwing touchdowns while the defense loses points for allowing touchdowns.

How do correlations affect my lineups?

The higher you set your correlation weight in SaberSim, the more your lineups will favor positively correlated players, giving you more (and larger) stacks. Setting it lower will give you more players that do well individually, and going negative will give you those players who do well when another in your lineup does poorly.

How does it work?

By simulating every game thousands of times, we’re able to accurately predict the outcome of games but on top of that, these data points show us in detail how a player’s performance can vary and how those variations impact the other players in those games.

Using this unique data, we’re able to measure the actual relationship between two players which we then use to build your lineups.

How should I use it?

GPPs

On bigger slates, I recommend going with a higher correlation weight to make sure you get bigger stacks. The reason for this is that as the number of games increases, so does the likelihood of a single team having a huge night and to maximize your upside to that event you need to make sure you have exposure to larger stacks.

In smaller slates, you still need stacks, but those huge nights are less likely so it becomes more important to have more exposure to the individual players with the most upside for the slate.

Cash

I suggest lowering correlation, but I don’t recommend turning it off entirely or going negative. While we allow those options because there may be some edge cases where it could make sense, but as a rule it’s generally safe to ignore those options

Examples

Here is a build where Correlation was set to “Neutral”

This is similar to what you’d get from a traditional lineup builder where there are a few mixes of multiple players from the same team, but a ton of lineups don’t even have two players from the same team. This is because these lineups were built using the average projection for each player individually and not accounting for the additional upside that comes from correlations.

Here is a different build where Correlation was set to “Very High”

Even when no stack rule was set, you’ll get more stacks and biggers stacks just from setting Correlation high. From here, you’re able to adjust the exposure to get the mix of stacks you’re looking for, but you’re able to get a strong foundation right out of the box which no other lineup builder can do.

Things to keep in mind

First, setting correlation high will automatically build stacks into your lineups, but, especially in a sport like baseball, it can still be useful to include separate stack rules, which you can do by clicking “Stacking” in the “Create New Build” window. On the other side of the coin, just because you set a stack rule, doesn’t mean you should set a low correlation weight.

When you create a stack rule, say for at least 4 players from the same team in every lineup and have a higher correlation, the stack rule guarantees all of your lineups will have at least a 4-stack while the correlation weight will make sure those stacks use combinations of players with the most upside.

Second, if you go very high on correlation weight, mediocre players, like those batting 8th and 9th in MLB, may get enough of a boost to be added to your stacks. This isn’t necessarily bad, those players are correlated with those at the top of the batting order, but it brings additional risk with it that you should be aware of.


How did we do?


Powered by HelpDocs (opens in a new tab)