fix footnote

reframing-post-shot-xg.md

@@ -40,7 +40,7 @@ Randomly sampling with `n=10`, the player with the worse accuracy (`σ = 0.7 met
 
 The best way to combat this overfitting is the allow off-target shots to retain some xG value; in turn crediting narrow misses with more value than wide misses. This isn't an entirely new concept: I collaborated with some very smart people last year to take a stab at the concept[^5]. In other words, we need to extract more meaning out of every sample. However, it's not completely intuitive how this makes any conceptual sense.
 
-[^5]: (Ethan Baron: Predictive Value of Off-Target Shots in Soccer, NESSIS 2021)[https://www.youtube.com/watch?v=zQCl1cL-JxA]
+[^5]: Ethan Baron: Predictive Value of Off-Target Shots in Soccer, NESSIS 2021: https://www.youtube.com/watch?v=zQCl1cL-JxA
 
 To demonstrate, we train a goal expectation model that estimates xG based on the shot trajectory error, as opposed to just using the shot destination. This makes sense if we're interested in rewarding process instead of results. Ultimately we want to credit a player who strikes the ball accurately, no matter which side of the post it happens to end up.