BuzzerBeater Forums

Can you tell me the probability that a single rookie has 4-5 star ability and 5 star potential, and the probability of whether the rookie has ATG potential?

If you have the above data, combined with the number of real players and robots in a country, you can use statistics to directly calculate the confidence interval to see if the number of ATGs obtained by the United States is abnormal.

I believe the BB team does not want to be perceived as giving the United States an unfair competitive advantage.

Meanwhile Italy

Thanks Perpete!

Question to Marin and Justin...

Are potentials on draftees defined or related in any way by Gaussian function?

I don't see how this data is relevant to the conversation. What is relevant is the amount of active players generated this season. Players who can be bought, sold and trained. The relevant data is as follows:

The top 9 counties with most active users excluding USA (China, Deutschland, Espana, France, Hellas, Italia, Lietuva, Poland, Turkey) had a total of 2083 active 5-star potential draftees generated. 0.28 5-star draftees per user. 19 of those were ATGs. That makes roughly 1 in a 100 which confirms the popular perception that ATG is visualized for players with 10.98 and 10.99 potential.

USA had 306 active 5-star potential draftees generated (this is absolutely consistent, 0.26 per user). 9 of those were ATGs. The chances of this happening are 0.29% or once in 345 seasons. That's what the facts are, no implications. All I can say is congratz!

EDIT: Knowing now that the Gaussian function is used changes those numbers significantly. It would seem that all 10.9+ potential players are visualized as ATGs. Depending on exact implementation, the chances increase to ~2% or once in 50 seasons.

@Perpete: I posted here before reading all new messages in the thread. Please move this to a relevant thread if you feel like it.

What bothers me much more is that I have read several times now how these recent changes are a great thing for smaller nations.

No, they are not!

What happens now is that big nations obviously get more ATG players each season (almost all relevant, thanks to the 19yo boost), while smaller nations get one in 4-5-6-7 seasons...

And no, the ~0.5 potential level which the ATGs have over the HoFs on average is not a big deal. What is a big deal is that the owners of those players know exactly how good their player can become straight away, as opposed to spending 10+ seasons training a player having zero guarantees whether the player is a 10.01 or a 10.97... This makes a huge difference when a nation has a limited number of active managers, who can train players efficiently and keep them in good GS.

Last edited by mink0ff at 3/26/2024 6:05:43 PM

Although GM-Perpete called us to stop, but this is interesting and I couldn't help, so here are some statistical checks I have done...

TL;DR:

USA's outcomes on 18+19y, 11p is roughly a 1-in-50-season event for an individual country.
It would fail the "traditional 95% confidence interval test", but given that USA had already drafted for 60+ seasons, something like this is not too absurd.

Data: I used the NT-scout to scan all countries with 80+ active managers, and count the number of 18y11p and 19y11p prospects.

There are 38 eligible countries, 12746 managers (as of today) and 46+3 11p prospects.

Assumptions:
1. Utopia dual citizenships are excluded. It would complicate things since we do not know how the flags are distributed. There are only 3 such players, so it should be alright.
2. Numbers of managers are as of today; the more accurate way to do things is to use the numbers in the draft day, but this shouldn't be too bad.

Model assumptions:
1. Assume that the expected number of 11p is proportional to the number of active users. This can be somewhat deviated based on how aware the lowest league managers are about the draft, but let's forget this.
2. Assume that the distributions are Poisson. Just not that if assumption #1 is violated, ODP might be more appropriate than Poisson.
3. From #1 and #2, we assumed that the probability of drawing 11p per manager is a fixed constant P. Justin said potential is Gaussian distributed, but we can ignore that part and try to estimate P.

The result is shown in the link below. We used two ways to estimate P:
1. Maximum Likelihood Estimation (MLE). This is just a fancy way to say P = 46/12746 = 0.3609%.
2. Use a Bayesian approach with uniform prior distribution.

This leads to my claims in the TL;DR part.

(https://imgur.com/a/D7UAQ17)



Last edited by Feizai Passing by at 3/26/2024 6:52:48 PM