Posted by TulsaG on 12/24/2020 8:12:00 PM (view original):
The salary cap for dynamic pricing must be consistent across the board, but it needs to be looked at as well to check to make sure it's doing what it's intended to do.
There are unfortunately a few inherent flaws in the idea which I will give some examples of.
In an OL, more or less every player is usable, and can be successful. You will see a very wide range of players that will taken. However, there will be some players which appear more than others, which is to be expected. Let's say for arguments sake, 1921 Babe Ruth is the most commonly picked player for open leagues. It would still need to be determined if A: 1921 Babe Ruth is being picked the most because he is under priced, in which case this needs to be adjusted or B: He is just a very popular player because of his name, and is priced correctly.
If we see lots of naturally popular players, especially modern players, on a lot of open league teams, it's quite likely options B. If we see swaths of guys no one has ever heard of, or guys that played before any of us were born, then it could be option A. It would take more data analysis to determine which.
1921 Babe Ruth's price didn't need to go through the roof simply because he was a very popular player (more on that in a minute). The purpose of dynamic prices should be to raise the prices of UNDERVALUED players, not popular players. This cannot be determined simply by looking at who is on the most teams, otherwise you risk bringing massive subjectivity into player pricing.
The second problem, which has been addressed multiple times, is that if you included theme leagues, particularly high cap leagues, it's going to grossly skew the data. In an open league, I can use a very wide range of players that I can expect to be successful. However in higher cap leagues, that pool of players who can compete at that level gets smaller and smaller, thus creating more duplication. In a $255M league, you're probably going to see 24 Ruths, maybe even more if clones are allowed. You'll also see just as many Eckersleys, Gagnes etc. because there are only so many closers that you can expect to compete at that level. In the open league, it would be extremely unlikely to see that much duplication at a position such as the closer since there are countless options. Sure you will probably see more Riveras than anything, but that's likely because of the name on his jersey, not because of his numbers.
Bottom line is: Seeing the exact 3 million dollar pitcher no one has ever heard of from the deadball era on a high number of OL teams suggests something should be adjusted. Seeing lots of Ruths or Riveras does not carry nearly the same meaning. Seeing 2000 Pedro and 1995 Maddux on every 255M league means almost nothing. Which in lies the problem with dynamic prices, even if it's only for $80M caps, the name on the jersey is going to play a significant role in which players gets prices raised, which is not what this is designed to do.
The third issue is, in my opinion, there will always be too small a sample size to make any meaningful conclusions based on number of times drafted.
Think of it this way: If you owned a restaurant with a million different options, if you had 1000 customers a day, even if everyone ordered something different, it would be nearly 3 years before every item was ordered once. Many players are never going to get picked for an OL, not necessarily because they aren't good, but there are just so many players to choose from and not enough managers to ever cover them all. This also factors in to why the high caps skew it so badly, because there are so few options at each position that make sense for high caps.
This is why I suggested that we need experienced managers to have some input here. If we can determined WHY certain players are being drafted so much (undervalued vs popular) we can figure out why certain players or types of players are undervalued, adjust accordingly, and create a more level playing field for everyone without collateral damage, such as taking guys like 2000 Pedro out of play.
I hope WiS can find some usefulness in that.
I don't know that we need to balance out why the player is being chosen since the complaint is largely driven around how often people see the same guys on teams they oppose, not how accurately priced they are. I'd argue the base salary formula is (mostly) on point and the cookies are derived mostly from perceived value or name recognition.
Take '08 Joss for example, he was clearly the most used pitcher before dynamic pricing and was widely considered to be an extreme bargain, but was he really? It was easy to look at his performance and forget how it intersects with the team. Him being a deadball pitcher increased the error rate of the defenders behind him, which also makes his ERA look better because the baserunners scoring are often a result of errors. Which is why when determining a pitchers value in sim, RA/9 should be looked at more than ERA. If you look at zubinsum's old post on
under/over-valued pitchers, Joss was middle-bottom of the list at 12%. Zubinsum pointed out a few posts after that he did not include defensive effects. I've not studied defensive effects thoroughly, but the effect is very clear. Even just at a cursory glance the effect appears to be somewhere between 10-20%. You can see some of this (though the direct reason/measure is different) in the fatigue data for teams that focus on defense versus those that don't in the last
TWISL thread.
From zubinsum's list already, it was clear Joss was taken so often, not because he was the best value, but because he was the relatively the easiest to use AND a recognizable name and because not everyone understands how defense effects overall performance.
From the overvalued list, it was almost exclusively good hitting pitchers and/or modern pitchers that improve defensive performance.
We can measure in-sim performance and value relative to other players and see where popularity is a factor of value being a factor. Keeping the dynamic factor limited to OL/CL leagues keeps a natural limit on how far that effect can impact. Even newer owners are likely to skip a higher priced popular player for a lower priced popular player all else being equal or close to equal. Restating my first comment, "I don't know that we need to balance out why the player is being chosen since the complaint is largely driven around how often people see the same guys on teams they oppose, not how accurately priced they are."