Bucking, Balling, and Mathematics · 3:28pm May 12th, 2019
Good to see in Common Ground that the analysis of Buckball statistics I imagined in Buckball Abstract is now canon.
I think it must be time to investigate this branch of applied mathematics further.
Baseball analysist have developed an impressive set of equations to assess the performance of their favourite players and teams and add some mathematical precision to their assessment of who is the battiest batter. The favourite metric appears to be the OPS (on-base plus slugging), which can be calculated as
OPS=\frac{AB\times(H+BB+HBP)+TB\times(AB+BB+SF+HBP)}{AB\times(AB+BB+SF+HPB)}
where H = hits, AB = at-bats, BB = base on balls, SF = sacrifice flies, HBP = times hit by pitch, and TB = total bases.
Got all that?
So, math and pony fans, here is a world-building exercise: derive the optimum equation to judge the performance of a professional buckball player? Of course we would need more buckball data to check it, but what parameters should we measure to start the job?
Wouldn’t you need three different approaches for the three positions?
5057646
Indeed you would.
You'd want to know the opportunity cost of employing a player, i.e., their Revenue Above Replacement (RAR). How many more tickets are sold, how many more jerseys sell, etc., when you employ the player over a hypothetical baseline player.
It may be that the owners of buckball teams have preferences beyond the financial: they may want their teams to win as well as to make money. In which case the metric would be Win-RAR—though it doesn't really sound like the sort of thing ponies would buy into....
5057794
As buckball appears to be an amateur sport just on the point of becoming professional, there is clearly an opportunity for an enterprising pony to step in and make a lot of bits.
vignette.wikia.nocookie.net/mlp/images/e/e9/Daisy_giving_bits_to_Snips_S9E6.png/revision/latest?cb=20190505202418
The nerds beyes for stats?
5058103
Bayesian or Frequentist?
5058184
I dunno. From what little of maths I can understand, running Bayes bigrams over frequency lists gives you a neural net equivalent over ngrams. Except absolutely predefinable. And with self learning properties with the right alterations?
Some other things Ive seen with it give data errors that to me look like various mental problems. Using resource limitation and allocation mistakes.
Supposedly just simple code to try but every time I try doing it myself I get ill.
I'm like Quibble Pants here, I have no idea what this means, because I have no idea about sports. But the idea of calculating that sounds exciting and I would love to do that if I wouldn't suck with numbers so much. But I see a cute picture of Wind Sprint up there, so it's all good.