Have you sent it to D1baseball suggesting they use it?
I did t intend for my post to come across as arrogant, if it did. I legitimately see there is an error in their coding (specifically on the pitching side of things).
Their model has bugs in pitching code and doesn't adjust the hitting WAR enough for the run environment and park factors. It's entirely possible they're regressing things differently than I do, which is fine. Lots of ways to build one.
That said, there is no way they intended for Sam Cozart to be a negative WAR player. Their own site named him a freshman All-American. Their model uses ERA rather than FIP like mine does, but his ERA is bonkers, so that should actually help in their model. I have him supplying over 3.2 WAR. This is but one easy example.
View attachment 366801
They have a calibration issue as well, where hitting WAR is over 80% of the wins and their implied win totals have a team made up of entirely replacement level players winning 37% of their games, which is, objectively, far too high. Typically, you would tune your model to where 78-80% of available wins are spoken for (making a team of replacement players winning around 20-22% of their games).
Their pitching model doesn't adjust for leverage index, meaning they apply the same weight for impacting wins to a guy pitching in a 13-1 game as they do a closer protecting a 3-2 lead. One impacts winning to prevent a run far more than the other and should have a leverage index applied accordingly.
I could go on and on with critiques for what their outsourced model has come up with, but it feels harsh for a site doing a ton to grow the game.