Burrow's interception issue over his career

[Stewy]

(05-17-2022, 02:10 PM)KillerGoose Wrote: It's an advanced metric, standing for "Expected Points Added". It is essentially using the current game state to figure out what a teams expected points are, which can be calculated by having data such as field position, stadium type, time remaining in half and a few others. Expected points is exactly as it sounds - under the input conditions, what is the team's current expectation for scoring points?

The added part is just figuring out what kind of difference the play produced in this expected points total. So, if a team has the ball at their own 10 yard line on first down, they may have an expected points value of one. Let's say Burrow throws a 15 yard pass and they now have first down on their own 25. Their expected points value is now two. Burrow's pass has an EPA of one, as his pass added one expected point. It's the best metric for evaluating QBs due to the correlation it has to team success, which is very strong within the context of football. Passer rating is second. You can also use it to evaluate running backs, and offenses/defenses as whole. 

So with this context, can you explain why your graph is cumulative and why lower is better?

Thanks.