Quote:
Originally Posted by BPS
Batters, pitchers, and fielders all have clutch abilities.
But, you could simply add a variable "clutch ability" for each batter, C, that affected outcomes only in pressure situations. C would be distributed over batters with a mean zero and a "relatively small" variance.
If you ran a series of seasons where players typically played only a small number of years, and you did this one time, the combination of randomness and small sample size would make it very difficult to determine (as an outsider who didn't know whether C existed or not) whether players had clutch abilities or not.
As I said, it is very simple to add clutch ability and very difficult to determine whether clutch ability existed just looking at the data.
|
This would be discoverable in something like OOTP merely by setting all players other skills to neutral and running a hundred seasons or so, then doing a study like I described. It takes work, but is able to be discovered.
But, yes, by definition if you make the variance you code extremely tiny, then you can implement anything and have it hard to feel in everyday play. This is the scenario by which its impossible to prove something doesn't exist without very large samples. All you can say in real life is that the impact is so small it's hard to find without more sample. The key here is that it is so small it is not affecting the outcome in any perceivable way. Operationally, some people would then say it does not exist, which is fair enough even if it's wrong.
Discounting, perhaps, some more advanced deep-learning algorithms, my view is that if you code it in software, I will eventually be able to find it in results if I care to spend the time looking.
I can, for example, see influences of managers and coaches in OOTP data, though I haven't gone the extra step to quantify them. It's hard work. Tedious and somewhat complex. But easily doable with enough motivation.