Anyone who thinks that you should simply pick the goalkeeper with the highest penalty save percentage is stupid.
1. Sample size
If Goalkeeper A only has faced 10 penalties in his professional career whereas Goalkeeper B has faced over 50, then it's not a fair comparison. 10 penalties is practically useless as far sample size goes. Hell, even 50 is not that good of a sample size.
2. Improvements over time
Let's say that a goalkeeper starts off his "penalty career" by conceding pretty much everything up until a certain point. Even if he improves drastically, his overall stats will be doomed by his poor start. You could of course only focus on recent history, but then you potentially have biased data. Also, where do you draw the line? If a goalkeeper has saved 5 of the last 10 penalties does that mean that he's a penalty expert now? Is a goalkeeper who has saved 0 in the last 10 automatically bad at penalties?
3. The quality of the penalties faced
It goes without saying that this is not only the most important factor, but it's also completely outside the goalkeepers control. And considering how few penalties the average professional goalkeeper will face in his career, this can have pretty devastating effects on their save percentage. If a goalkeeper faces 50 penalties in his career and 40 of them are "un-saveable", then he's fecked. Even if he saves 50% of the saveable penalties(which is insane), he will only have a 10% penalty save rate. Another goalkeeper could be lucky and only be faced with 10 un-saveable penalties, and thus have a sample size of saveable penalties that are 4 times as high. Even if the second goalkeeper is nearly twice as bad at penalties, his penalty save rate will be more or less the same as the first goalkeeper.
_______________________________
TLDR; I would never use "penalty save percentage" as the only decider for which goalkeeper I pick for a penalty shootout. I'd want my data scientists to analyse the penalties in depth while heavily considering the 3 variables above.