BuzzerBeater Forums

i dont know of a way to gradually go between what we have now and what we want to have...

what we have now is that getting a rebound is picked from a distribution proportional to the ratings.... sounds reasonable right?

but there is a big difference between picking between a level 1 and 3 and a level 11 and 13. In the first case, the level 3 gets a 3x time advantage in rebounding.. in the second its almost a toss up and so who gets more rebounds is actaully just a function of a) who had more defensive rebound opportunities and b) luck, which is how you are getting the "buggy" rebounding results you are getting now.

on the other hand, what i would propose to do is look at the mean rebounding on the court.. subtract that off and then have the weighting be a sigmodial function of the difference from the mean... so there is some maximum advantage you can have over rebounding... a sizeable but finite advantage... (so one team can never get all the boards).

EDIT: SOME OF THESE STATEMENTS MIGHT BE MISLEADING.. PLEASE READ (66639.86)

Last edited by BB-Forrest at 1/7/2009 9:49:46 PM

i dont know of a way to gradually go between what we have now and what we want to have...

use weight and both algorithms, one who use absolute differences and one who use relativ difference.

For example:
absolute difference Team A gets Rebound 55%
relative difference Team A get rebound 51%

weighted relatives difference/absolute -> 3/1
result -> 52%

so you get an slithly change to the way you like it, with a pretty basic way of implementation.

This depends on how the GE calculates the percentage probabilities. It is easy to find the midpoint ex post, but I am not so sure this can happen in the process of game simulation.

you don't have to change one algorithm, you only have to extract the rand function out of it ;)

and start it after running the new algorithm, and rebalance the "fields" with the weigths. And i don't know a way to implement that another way, you had to store one help variabel, which shouldn't be a problem but it needs more time to process which could be a problem.

fair enough.. i could do something like that... though i will say it is more complicated than you laid out.. given how its really a 10 way choice, and we would have to normalize both distributions which we didnt have to do before, so its gonna be computationally more intense....id have to think harder/experiment to find out if its prohibitively so...
EDIT: SOME OF THESE STATEMENTS MIGHT BE MISLEADING.. PLEASE READ (66639.86)

Last edited by BB-Forrest at 1/7/2009 9:50:33 PM

I can say that I am continuing amazed at the ability of our users to detect bugs with disturbing accuracy. There was indeed a bug in the GE with respect to how OD team ratings were presented, that had to do with the SG OD contribution... (his jumpshot was being added to the rating and not the OD). Now before everyone freaks out and demands their games rerun, this is simply a presentational error.. as teamratings are not the things we principly use when calculating results for things in the GE.

I fixed it on Tuesday, so things should be presented better going forward.

perhaps this is the OD error ned is winking about...

I just started reading this thread, so I don't know if anyone has raised my point already, but I'll give it a go. I really think something like this should be mentioned on the News page.

I don't often read the threads in Global and, with this change, I would be awfully confused by the change in my own team's ratings as well as those of upcoming opponents that I'm scouting. Might also be worth keeping some sort of public changelog.

no let me be clear... the OD team rating is a weighted average over time of all the positions OD ratings (modulated by energy etc). So there are 5 values that go into the calcuation at each moment in time.... 1 of those values.. the SG was his JS and not his OD.. a typo. So the ratings was mostly OD and a little bit SG JS... ratings wont be drastically different, just a little bit different... and there shouldn't be any noticeable effect on the outcome of games, just on the rating.

EDIT: SOME OF THESE STATEMENTS MIGHT BE MISLEADING.. PLEASE READ (66639.86)

Last edited by BB-Forrest at 1/7/2009 9:50:52 PM

perhaps the best we can do is announce the change now and implement it for scrimmages for the rest of this season and then make the change next season.
note... this is not an announcement, i am only speaking for myself here.

This will be more than great!

Coming back to inside vs outside, I also think that the game is well balanced but maybe there is still a little bit advantage playing inside. I think BBs can see everything from the headquarter, stats and other things that we don't know.
There are two other reasons that I didn't write before; I think OD training is a very slow training, maybe slower than ID, if you see the market is difficult to find players with more than lvl 11-12 in Od, second, who plays inside has 2 different tactics to use against outside attack, 3-2 and 1-3-1, who plays outside has only 2-3, an implementation of an extreme defense against inside could be a good and funny solution.

About GE.
Yesterday I had an easy game in cup and I've selected Cinotti on the field and Zanenga in backup. Well, Zanenga started since the beginning on the field and he didn't quit neither in the garbage time playing in total 47 minutes. Zanenga is 10 times stronger than Cinotti, but the coach should "respect" my choices, don't you think?