39 coaches online • Server time: 16:38
Index Search Usergroups Profile
Log in
Recent Forum Topics goto Post FUMBBL Dice 2021goto Post Unable to activate t...goto Post Creating a tournamen...
SearchSearch 
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic
Christer



Joined: Aug 02, 2003

Post   Posted: Apr 27, 2021 - 22:54
FUMBBL Staff
Reply with quote Back to top

Inducements are purchased with petty cash. The TV difference is given to the underdog as extra petty cash, so the CR system update does exactly that.

To clarify the 0.62CR in Halfabrain's example:

Home Coach, playing CDs:
CR 152.59 (Emerging Star)
TV 1800k

Away Coach, playing Underworld:
CR 138.73 (Veteran)
TV 1130k + 670k spent (Babe, Glart and Morg) = 1800k

Resulting in a 2-0 victory for the CDs and a 0.62 change in CR.

Quick summary of where the 0.62 comes from:

At its core, it's based on the following: k * (S-p)

k = 1.56 in this case (normally 2, but bracket differences result in a reduction)
S = 1 for a win
p is the expected win probability as follows (slightly simplified because of dTV = 0):

Calculate weighted CR difference W

W = dCR + (dCR*0.275)^3 = -69.23

p = 1 / (1+10^(W/400)) = 0.5983

And finally:

1.56 * (1-0.5983) = 0.62

With equal CRs (and brackets), the CR delta would be 1.0 instead, or 2*(1-0.5).

End result is a slightly different meta with massive underdogs and inducing a lot of stuff won't be beneficial anymore.
XpherAndTheAxes



Joined: Jun 20, 2020

Post   Posted: Apr 27, 2021 - 23:09 Reply with quote Back to top

so just to clarify, if we get a nightmare matchup like that, its better for our cr to just bite the bullet and not induce anything?
MattDakka



Joined: Oct 09, 2007

Post   Posted: Apr 27, 2021 - 23:20 Reply with quote Back to top

Apologies for the dumb questions, but I think that an example could help me to better understand the new CR calculation.

Let's say CR 160 Super Star coach A plays a TV 1100 tier 1 team vs CR 160 Super Star coach B with TV 1000 tier 3 team.
Coach B buys inducements for a value of 100k, so now the teams have a theoric equal value although we know this is not often true because most inducements are deliberately overpriced and tier difference is not a secondary factor.
Coach A wins the game 2-0.

How many CR points are gained by coach A?
And how many CR points would be gained by coach B if he had won the game, instead?

I deliberately picked coaches of equal CR and rank because I'm curious to know how much playing a tier 3 team affects the CR gain/loss.


Last edited by MattDakka on Apr 27, 2021 - 23:29; edited 2 times in total
sebco



Joined: Feb 14, 2005

Post   Posted: Apr 27, 2021 - 23:26 Reply with quote Back to top

Thanks for clarification (and sorry for bad "petty cash" LRB vocabulary use).

In my opinion, this new system will be good with BB 2020 and it's a good thing you already implemented it (but I still agree with coachs saying it's harsch in BB 2016 as, in your example, game should be harder in BB 2016 for 1130k team than for 1800k team even if 1130k team is minmaxed and can chose 670k inducements).

_________________
I like cheese but don't call me skaven !
MrCushtie



Joined: Aug 10, 2018

Post   Posted: Apr 27, 2021 - 23:33 Reply with quote Back to top

Well, from Christer's example, if CR is equal then W = 0.
So p = 1/(1+ 1) = 0.5

CR change is 2 * (1 - 0.5) = 1 (and that would be the same in both directions).

But I'm guessing there's a bit more to cover racial probabilities (otherwise as given there's no difference what tier you're playing, as long as the TVs are equal)

_________________
Image
XpherAndTheAxes



Joined: Jun 20, 2020

Post   Posted: Apr 27, 2021 - 23:35 Reply with quote Back to top

just to put my word in, as its my UW team- they struggle to get games at their own tv because of the way the scheduler works, meaning that they keep getting lumped with the biggest teams to activate that round, making it almost impossible for them to survive, and they can in no way compete with most of the larger teams even with spending every penny of inducements.

now i dont mean to be a whinger, but its really annoying to have your team torn to shreds even with inducements, and lose this much CR in the meanwhile, and if to protect my CR i have to skip inducements then it might be time to stop playing in box
MattDakka



Joined: Oct 09, 2007

Post   Posted: Apr 27, 2021 - 23:40 Reply with quote Back to top

As an aside, I think that bracket difference (Star, Super Star etc.) should not be taken into account for purpose of calculating the CR win/loss.
Only the CR should be considered, because makes the CR system more accurate.
For example, a coach with 159 CR could be a Star, while a coach with CR 160 could be a Super Star, but actually there is not as much difference in terms of coaching skill as their bracket difference suggests.
Taking into account the brackets (especially for a blind MM division) can create odd swings in CR gains and losses.
If I could choose I would never ever play vs an Emerging Star or lower (nothing personal, just business Very Happy ), but since in the Box I can't pick my opponent and his bracket, I should not be penalized for playing vs an Emerging Star like if I had deliberately picked him in Ranked division.
The userbase is too small to split the Box into 2 divisions (Gold for Stars and higher and Silver for Emerging Stars and lower, for example), so a logical alternative could be removing the bracket from the CR formula.


Last edited by MattDakka on Apr 27, 2021 - 23:55; edited 2 times in total
MrCushtie



Joined: Aug 10, 2018

Post   Posted: Apr 27, 2021 - 23:53 Reply with quote Back to top

XpherAndTheAxes wrote:
just to put my word in, as its my UW team- they struggle to get games at their own tv because of the way the scheduler works, meaning that they keep getting lumped with the biggest teams to activate that round, making it almost impossible for them to survive, and they can in no way compete with most of the larger teams even with spending every penny of inducements.


Come drink at the Gastro Pub, monoactivating with low-TV Nurgle since 2021...

_________________
Image
MattDakka



Joined: Oct 09, 2007

Post   Posted: Apr 27, 2021 - 23:57 Reply with quote Back to top

MrCushtie wrote:
Well, from Christer's example, if CR is equal then W = 0.
So p = 1/(1+ 1) = 0.5

CR change is 2 * (1 - 0.5) = 1 (and that would be the same in both directions).

But I'm guessing there's a bit more to cover racial probabilities (otherwise as given there's no difference what tier you're playing, as long as the TVs are equal)

Thank you!
Well, if playing an underdog tier 3 gives the same CR points of playing a tier 1, then there is no incentive to play the tier 3 in my book.
Playing a tier 3 should make the coach win more CR points and lose fewer CR points.
There should be a multiplier, so, if playing a tier 1 gives you 1 CR point in case of win (assuming same coaches' skill and TV), you should gain the double or the triple CR amount in case of win with a tier 3 and the CR loss should be halved or divided by 3.
That way would be less penalizing to play a tier 3.
Christer



Joined: Aug 02, 2003

Post   Posted: Apr 28, 2021 - 02:23
FUMBBL Staff
Reply with quote Back to top

If you want to minmax your CR, you shouldn't be deliberately picking a team that's designed to be worse.

As for not using brackets because raw CR would be more accurate, the simple answer is that you're wrong. I don't intend to elaborate further on that point at this time.
koadah



Joined: Mar 30, 2005

Post   Posted: Apr 28, 2021 - 09:46 Reply with quote Back to top

Christer wrote:
If you want to minmax your CR, you shouldn't be deliberately picking a team that's designed to be worse.


I would have thought that it is not just about "min-maxing" CR. It is more whether the CR loss is perceived to be "fair."

Though many people seem to say (and maybe think) that they don't care about CR, I suspect that they do to some extent.
Just not as much as Matt. Wink

_________________
Image
[SL] Rookie Rumble - New teams ONLY - ALWAYS recruiting
MattDakka



Joined: Oct 09, 2007

Post   Posted: Apr 28, 2021 - 12:19 Reply with quote Back to top

Some coaches want to gain as much CR as possible and will not touch a tier 3 with a barge pole no matter the formula, it's true, but some other coaches, even if they don't care as much about CR, don't want to be penalized by a big CR loss either when playing a tier 3 team vs a tier 1.
There are many levels of CR-interested-people, there are not just 2 categories: CR-obsessed coaches and not-giving-a-damn-about-CR coaches.
Some don't get mad about losing CR but still they don't want to lose too much CR when playing tier 3 teams vs tier 1, because the game is harder and the CR win/loss should mirror that difficulty.
It's not a matter of minmaxing CR or wanting to climb to the top rankings by playing tier 3 teams only, it's just a matter of giving an incentive in terms of CR gained/won when playing a tier 3 vs tier 1 or 2.
tussock



Joined: May 29, 2011

Post   Posted: Apr 28, 2021 - 12:45 Reply with quote Back to top

It's really a matter of if it's possible to make a system of rating the teams objectively. Obviously Zons are just better when they refuse to play Dorfs/Chorfs, but also at low TV where other teams have less T/MB.

But there's not enough games to give every one of the 729 different matchups (especially rare ones like flings vs zons) at a bunch of TV brackets.

Then, gotta ask, would linear trends of "chaos are better at high TV, and goblins are worse, in general" you might extrapolate from way less data just open up weird points where a team does unusually well that, say, a team might smallkosp their way into for unlimited CR gain?

Like if it was just 729 race vs race, plus 27 general trend lines for race vs TV, that might give a better veiw of people whose CR varies a whole lot. Could also be used to paint an expected win% on prospective games that lined up a bit better that just plain CR.

--

Like, if there's a problem with CR now, it's that as it responds to me playing endless stunty games, that I'm not great at in the first place, switching to tier 1 teams, well, I don't tank the CR of people who are nice enough to play me along the transition quite so much.

But, yeah, it'll eventually tell you if I'm mostly playing Lizards and Dorfs lately. And then eventually it'll tell you I've gone back to my joke box teams for a bit. Very Happy

--

Should it even try to tell you I'm unusually bad with Chaos Dwarfs? On account of I don't buy anything but Hobgoblins?

That more requires a TR for teams, what's a squad really worth at 1400 TV if they're all hobs? Answer for real is it depends what you hit in box, because like minmax Humans or Lizards just rip them up, but like young elfs go the other way because of the stack of skills that aren't Guard or MB.

And that is probably beyond any system to figure out.

_________________
ImageImage
Halfabrain



Joined: Jan 20, 2018

Post   Posted: Apr 28, 2021 - 13:22 Reply with quote Back to top

Tbh I'm not that worried about finding a system that perfectly reflects a coaches ability, as MattDakka has often said even a perfect system is rendered imperfect when it doesn't discriminate between R and B.
Pickers gonna pick but BB is not without its problems and the new CR system seems exacerbate them for stunty coaches. There is a significant minority of guys on fumbbl with a massive stunty addiction and these coaches(I count myself among them) while generally not overly concerned with CR, nevertheless resent seeing coaches they don't necessarily see themselves as especially inferior to lording it over them and grabbing big CR bonuses for what are, inducements or not, hopeless mismatches.
This affects stunty coaches more than any other as they are the teams that generally have difficulty getting above 1.5ish in TV and no amount of inducements can come close to making 1.4 Goblins any kind of equal to 2m Nurgle.

All this is probably moot anyway as I imagine the system was actually designed for bb2020.
Christer



Joined: Aug 02, 2003

Post   Posted: Apr 28, 2021 - 14:34
FUMBBL Staff
Reply with quote Back to top

The problem here is that you're not considering the actual problem and are looking for band-aid solutions to cure the effect rather than getting to the root cause.

The CR system in its current form is about as good as it can get. The tier of a race does NOT accurately reflect the team, nor does the race itself. While the Shambling Undead are "tier 1", if you build it with only zombies you're not actually playing a competitive team. Thus, adding a band-aid for tiers or races isn't going to fix the issue.

A less extreme example of this is team building in general. There is a large difference between a tomb kings team with 4 TGs taking pass block and a tomb kings team with 4 TGs with guard. The two coaches with those teams, playing against the same set of opponents and the same results will imply that the coach with pass block TGs is "better".

In the end, it comes down to a fairly simple point:

Should "meta" choices affect the ranking of a coach? Is a coach picking Pass Block for their TGs worse than someone who picks guard? Obviously, that's an extreme example, but the same thing goes for everything in terms of team selection and building. Should you have a Troll on a starting Orc team? Is it beneficial further down in the progression? Of course, these things are supremely hard to judge.

The way the CR system on the site is set up, it is very very important to understand that team building is absolutely a part of the number. The CR system can't tell the difference between someone deliberately building "bad" teams for the challenge, and someone who genuinely doesn't know better.

Ultimately, the CR system is very very good the way it is now, and the real problem is that TV isn't even close to being an accurate estimate of how good a team is. Just blindly looking at two TV numbers, you need a very high spread before you can accurately tell which team is better than the other.

An elf team with 8 TRRs and dodge/catch/sure hands on all players would be incredibly high TV without being particularly good. Again, an extreme example, but I make it to be blindingly obvious that the core problem here is TV and not the CR system as such, and how applying a band-aid by giving bonuses to lower tier teams (or races) simply isn't the correct solution. You'd just be moving the goal post somewhat without making a significant impact.

As for how the scheduling system affects the ranking system, that's a contradiction in terms of what you're saying. If a system was perfectly able to estimate the win probability (which is the core of the CR system) of any given game, the way you schedule has zero bearing on your outcome. It wouldn't matter at all if the game was assigned to you, or if you chose it. Conceptually, the idea with any kind of ranking system is that as long as you do better than expected, you gain points and if you do worse, you drop. Your actual expected win chance does not matter at all.

Without trying to be rude or disrespectful, these basic concepts for ranking systems in general and our particular environment specifically are critical to understand before you approach me with suggestions on how to fix or improve the system. Also, a basic premise of "I should be higher" isn't great because you need to be super objective about these things, and who should or should not be at the top.

As an example, in the last major review of the CR system I did, I was approached with a statement that "cherry farming" was too strong, and coaches X and Y have a too high CR because of it. I investigated the issue and concluded that there was some element of truth in this and after a number of iterations came up with practically what we have now. The end result was that coach X ended up with a lower CR while Y remained roughly where they were before. I defended coach Y in this case, because even though the types of games they were playing and the type of team they were cultivating weren't popular, the coach in question appears to be fairly strong. CR isn't a popularity system, which is another important thing to consider. Just because you don't like someone's way of playing doesn't mean they're not good at it.

Anyway, this is a long enough post. The TL;DR is that I don't believe tier or race should be considered in the formula. CR is intended to reflect your ability as a coach, which includes the rosters you pick, and how you build your teams. Yes, things change around a bit occasionally with adjustments but for some reason the truly good coaches end up at the top no matter what because they're actually good players. It's an indication that the CR system works well for the most part as it stands, and I remain happy with the current state of it.
Display posts from previous:     
 Jump to:   
All times are GMT + 1 Hour
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic