BuzzerBeater Forums

Does anyone have any ideas how BB implements ratings/math behind who gets say a rebound?

I remember in Hattrick (back when I used to play a decade ago) there was an issue where as teams got better, the differences between ratings became smaller and random chance played a larger factor. I guess I can illustrate like this:
Rating Awful Rebounding vs Inept Rebounding = 3/7 = ~43% chance + random factor
Rating Wondrous Rebounding vs Marvelous Rebounding = ~48% chance + random factor

In BB it seems to be relatively consistent where a 2 level difference is equal chance whether teams are better or worse. Does anyone know how this is done/what area of math I should look at? I have an idea to do it but it's not particularly elegant and it's a bit of fudging.

Thanks

No idea

So im assuming your talking about game ratings and how that translates to the actual advantage a team gets on the floor.

Im not entirely sure it works in terms of 'game ratings = who gets rebounds" -
I think the game ratings are inversely calculated as a result of game engine results, rather than the game ratings influencing game engine, with the game engine being an combination of individual's rebounding in a given situation plus tactics.
That is - game engine calculates a play - play results in a miss - rebound is calculated based on rebounding skill of players on court plus other stuff.... each play is calculated indivudally like that.

And the end of game, a game rating is calculated based on a sum of the individual plays.

either way, I think what you are trying to work out is actually how much random factor vs how much of the rebounding skill of the individual, is involved in calculating a rebound, whether it be for an individual play, or an entire game.

I don't feel that there is a large amount of random factor. I think defensive team gets a natural bonus. Defender on the play gets a natural bonus and I also feel that certain positions get a bonus. Strangley, I think the order for position bonus's (based on a standard M2M defense) is

C
PF
PG
SF
SG

yes that's right, PG is 3rd. why do I think that? Look at how much salary rebounding adds to a PG AND AND AND - have you ever played with a PG with high rebounding? they smash it! its awesome. best chance of triple double player I feel.

so I think the formula is something like
individual rebounding chance - <is_player_defender?_factor> + <rebounding_skill_vs_opponent_factor> + <is_defender_on_play_factor> + <position_bonus> + <random_factor>

thoughts?

I feel that, if I understand what your saying, that game ratings are more than that. But i dont know, thats for sure!

As a quick 20 sec example, if a team were to smash an opponent in a skill, say passing, that the rating would be way higher? The rating from this game is only 2 sub levels above....
(35937)

Sure, i understand tactics play a part, but we both ran fast offenses, yet we managed 22 more assists and only just rated above them? I also understand we made 3 more shots on the court, not a significant difference.

I don't think I was too clear in my initial question. I was referring to individual's rebounding and how that is worked out for distribution of rebounding.

I guess I'll go with what I was thinking which is absolute difference in values (however they are worked out). So instead of hattrick style (division of values) leading to smaller advantages with larger numbers (level 2 vs level 4 compared to level 14 vs level 16 (2/6 compared to 14/30) I'll essentially do addition/subtraction leading to a max of X difference. This is what you have indicated with your formula anyway, so maybe we're both talking about the same thing.

I just wanted to make sure that I wasn't considering another way of calculating advantage.

yep
so in my formula i suggested a <difference_in_rebounding_skill> - is that linear? not sure.

Yer I reckon it's not linear, maybe sigmoid type logistic like most other things in the game

Maybe try with
L / 1 + e∆-k(x-x0). Minus random factor

Where L = 1 (maximum chance of rebound)
X0 = 4 ( which would the difference in rebounding stats between two players where the graph is mid-point)
e = maybe ln(5) ? I'm guessing here
K = maybe this is just 1 or maybe they use k as the way to favour the defensive player. (Such that defensive player gets k = 1.5 and offensive player gets 0.5)

Thanks, this helps

Cool let me know what u find.

I wonder if u could reverse engineer the formula a bit based on known rebounding data?
Ie: player x vs player y scenarios = 22 vs 34

Hmmm but the formula isn't player exclusive, there really is going to be 10 of those s curves on the same graph. ........hmmmm OK now my heads hurt.

Yes...heads.

Sorry, I wasn't clear when I asked the question - I'm not going to use this to analyse BB data. I'm in the midst of writing my own BB alternative. One that has people running around and reacting to each other/the ball (that you could watch! Better than a shot chart.) - I just wanted to see what alternatives there were for the mathy bits about deciding who is the winner of a particular action.