the whole of PVA and xT are done through grids which we get from open-sourced. @EightyFivePoint and @karun1710 are the pioneers here. There will be limitations in using these models and you need to acknowledge that in your conclusions.
It is a good step to thinking about valuing actions and offers a very nice way to see real dangerous events get quantified in actual analytics. Using these models, crosses might really add to the danger which is up to you whether you wanna use or not
Stuff like those are the nuances of such a model. Play around, tinker, ask if you need help, and improve! I regularly work with these two so you can ask me but here are some other people that are (more) knowledgeable than me (should probably ask them than me!):
@HemmenKees @DomC0801 @topimpacat @uapersonal @Wideoverload @pwawrzynow. You don’t always have to do a PVA/xT if you don’t feel comfortable! There are much more pathways in analytics with the data from FBRef and simple sites! Here are some people that do this excellently:
@GoalAnalysis @lgopfelix @ParthAthale @AshwinRaman_ @52_break @NinadB_06 @DrMukherjeeS @nandy_sushruta @DyslexicDdue @Advt_played @AnalysisEvolved @VenkyReddevil and so many more that I probably forgot! Apologies to anyone that I forgot - you all still do amazing work!!
Most common data sources: @fbref, WhoScored, and open-database of @StatsBomb. Other includes stuff like @smarterscout, while more advanced sources include WyScout. Explore, use, tinker, and expand!