The_Murker
Joined: Jan 30, 2011
|
 Posted:
Oct 18, 2017 - 05:14 |
 
|
So here is a new CR formula. It is simple. It rewards meaningful games with a CR increase. It disregards un-meaningful games almost completely. It is mostly ripped off, but it is deffo not ELO inspired. Picking weak opponents will not alter your CR in a meaningful way. Playing, even defeating, a coach who is out of your league will not alter your CR in a meaningful way. Here goes:
In the beginning, all coaches are born with a rank of 150.00. Their CR fluctuates normally (as a Veteran) for 10 games, but their title will always be Rookie.
There are 6 tiers of coaches, each with a title:
Experienced (CR < 145)
Veteran (CR 145 - 155)
Emerging Star (155.01 – 160)
Star (160.01-165)
Super Star (165.01-170)
Legend (CR > 170.01)
If you win or lose a game against an opponent in your tier your CR changes up or down by 1 point.
If your opponent is in a different tier, the CR adjustment is 1.0xM, where M is the ‘Tier Difference’ modifier.
(M=1. 0, 0.5, 0.15, 0.06, 0.002, 0.001 for tier differences of 0,1,2,3,4 and 5)
If the match is considered ‘lopsided’ due to a difference in the quality of the teams being used, the final CR adjustment made is halved. Period.
A lopsided match is one in which one coach uses a quality of team higher than his opponents. FUMBBL teams have 3 qualities: Poor, Normal, Borked
Poor: Goblins, Halflings, Vampires < 1150 TV
Normal: Everything Else. Only MAJOR differences are considered here
Borked: Amazons < 1200 TV
The formula for the victor: (opposite adjustment for the loser)
CR’=CR+(1.0M/2) for a lopsided match
CR’=CR+(1.0M) for a normal match.
That’s it. You can re-jig the entire tier composition weekly and re-assign the correct amount of coaches into each tier, if you like, if there are too many Legends, etc.
The thinking:
Guys just want to play. A skill assessment needs to acknowledge the fact that low TV games are a crap shoot, and exploits can take down the ‘best of the best’ in any venue, as can dice. We need to realise that only Legends and Super Stars should be knocking down Legends in this wild dice game, where good guys simply want to play as often as possible, with all kinds of teams.
This system crushes noob hunting. This system addresses stunty and zons. This system rewards ‘games played’ with the potential for a Legend to push his CR number very, very high, and establish himself there, but he can only do that with meaningful games against Super Stars and Legends. If a Legend is WINNING that many games… he’s a legend. Right?
But games against coaches of vastly different tiers are reduced to what they should be, a silly, meaningless game. |
_________________
Join the wait-list. Watch the action. Leave the Empire. Come to Bretonnia! |
|
The_Murker
Joined: Jan 30, 2011
|
| Posted:
Oct 18, 2017 - 05:15 |
|
$100 donation to FUMBBL if someone can use the FUMBBl ATI, or whatever, to run this through a year of matches to see what the results look like. |
_________________
Join the wait-list. Watch the action. Leave the Empire. Come to Bretonnia! |
|
The_Murker
Joined: Jan 30, 2011
|
| Posted:
Oct 18, 2017 - 05:36 |
|
Additionally, feedback on whatever CR permutation was in effect when Christer went to bed on the 17th of October.
The following coaches are far too low, and should easily be top 30:
#60 Stonetroll A total weapon who plays all, and fears no one. All around great FUMBBL participant.
#115 KFoged A FUMBBL ninja. Nuff' said.
#80 Purplegoo Top 15.
#40 Wreckage Probablly higher.
#719 PurpleChest I'm pretty sure this is just a joke among friends. I'm just pointing out it is funny. Good stuff.
The following coaches are not top 100, imo.
#36 smallman Wins via exploit, over all CR ranks. Wins vs. legends ill gained seem weighted very heavily. This coach would not stand amongst legends, imo.
#37 The_Murker Not top 100. Not legend. An example of wins vs. Rookies and Veterans weighted too heavily. |
_________________
Join the wait-list. Watch the action. Leave the Empire. Come to Bretonnia! |
|
Purplegoo
Joined: Mar 23, 2006
|
| Posted:
Oct 18, 2017 - 10:33 |
|
Only top 15?! You wound me! 
All kidding aside, it’s quite a narrow numerical range at the moment, and I lost my last game against PainState, someone who isn’t traditionally that highly ranked by the formula. It makes some sense I get knocked back a fair way if an ‘upset’ between our ranks is to be heavily rewarded / punished. I’m sure all of the other examples you give have similarly straight forward explanations. Perhaps everyone has just ended up where they deserve to be?
It’s a tough to impossible gig, ranking BB accurately. How does one weight a game where racial and team build factors are so important, let alone a roll of the dice? I think it’s cool you’ve offered an alternative, because picking holes in what exists is always easier than suggesting builds, but I’m more of the opinion that I’ll give it a cursory glance when the music stops and get on with life, rather than worrying about it. After all, if the perfect BB CR formula existed, there would be nothing anyone could do to improve their number other than be a better player! |
|
|
Sp00keh
Joined: Dec 06, 2011
|
| Posted:
Oct 18, 2017 - 10:59 |
|
If someone wins over all coaches, they should be rated highly
It's not a skill ranking, a quality ranking, or tournament success ranking, just an estimate of win chance in the games they play
If there's exploits, they are out of scope of CR discussion |
|
|
Balle2000
Joined: Sep 25, 2008
|
| Posted:
Oct 18, 2017 - 11:17 |
|
CR should account for races. If a 'Legend' coach loses with flings to an 'Experienced' coach, CR should hardly be affected at all. |
|
|
Sp00keh
Joined: Dec 06, 2011
|
| Posted:
Oct 18, 2017 - 11:30 |
|
A few other points (generally, I like this suggestion)
Rather than CR tiers, go with CR difference
There is a massive massive difference in a game between 160.1 coach Vs 169.9 coach, and a game between 154.9 coach Vs 155.1 coach
You would rank these 2 games equally
I also would say 1 point of CR difference means more at high ranking than at lower ranking, which is why Christer's got that exponential factor
The racial thing is a bit simplistic
If elves win rate against dwarves is low, they should get punished less for losing and rewarded more for winning. This would encourage elves to play against dwarves, but currently you'd rank them equal, so elves would avoid dwarves |
|
|
Balle2000
Joined: Sep 25, 2008
|
| Posted:
Oct 18, 2017 - 11:40 |
|
| Sp00keh wrote: | | This would encourage elves to play against dwarves, but currently you'd rank them equal, so elves would avoid dwarves |
Play Box. |
|
|
Sp00keh
Joined: Dec 06, 2011
|
| Posted:
Oct 18, 2017 - 11:43 |
|
@Balle2000 I do. The point is more general- if you don't scale by race then it punishes any non tier-1 teams, so would limit diversity |
|
|
Seany18
Joined: Mar 12, 2016
|
| Posted:
Oct 18, 2017 - 12:43 |
|
| The_Murker wrote: | Additionally, feedback on whatever CR permutation was in effect when Christer went to bed on the 17th of October.
The following coaches are far too low, and should easily be top 30:
#60 Stonetroll A total weapon who plays all, and fears no one. All around great FUMBBL participant.
#115 KFoged A FUMBBL ninja. Nuff' said.
#80 Purplegoo Top 15.
#40 Wreckage Probablly higher.
#719 PurpleChest I'm pretty sure this is just a joke among friends. I'm just pointing out it is funny. Good stuff.
The following coaches are not top 100, imo.
#36 smallman Wins via exploit, over all CR ranks. Wins vs. legends ill gained seem weighted very heavily. This coach would not stand amongst legends, imo.
#37 The_Murker Not top 100. Not legend. An example of wins vs. Rookies and Veterans weighted too heavily. |
Im rightly above all of these for no tangible reason as a glorious fumbbl nobody  |
|
|
koadah
Joined: Mar 30, 2005
|
| Posted:
Oct 18, 2017 - 13:38 |
|
|
thoralf
Joined: Mar 06, 2008
|
| Posted:
Oct 18, 2017 - 14:17 |
|
| Sp00keh wrote: | | The point is more general- if you don't scale by race then it punishes any non tier-1 teams, so would limit diversity |
Why stop there: we could scale by each team state.
Stat freaks.
Threats against which you got no response.
Number of blodgers vs number of tacklers.
The sky's the limit to prevent someone from being punished.
| Sp00keh wrote: | | It's not a skill ranking, a quality ranking, or tournament success ranking, just an estimate of win chance in the games they play |
A rating is more than an estimate of win chance, we wouldn't use it for rankings if it was. It indicates something like performance. That people infer skill or strength from performances is quite natural.
That they infer too much from it too. |
_________________
There is always Sneaky Git.
Last edited by thoralf on Oct 18, 2017; edited 1 time in total |
|
koadah
Joined: Mar 30, 2005
|
| Posted:
Oct 18, 2017 - 14:25 |
|
| thoralf wrote: | | Sp00keh wrote: | | The point is more general- if you don't scale by race then it punishes any non tier-1 teams, so would limit diversity |
Why stop there: we could scale by each team state.
Stat freaks.
Threats against which you got no response.
Number of blodgers vs number of tacklers.
The sky's the limit to prevent someone from being punished.
A rating is more than an estimate of win chance. It indicates something like performance. That people infer skill or strength from performances is quite natural. |
A good team strength algorithm would be excellent. If anyone has the time to do it.
Then they could move on to relative team strength. e.g. How many dodgers vs how many tacklers.
But I think that most people would rather just have another beer. |
_________________
20 Year Anniversary! - 2016 Nostalgia Bowl - Weds 23rd April! -- All Star Bowl - Always recruiting! |
|
Christer
Joined: Aug 02, 2003
|
| The_Murker wrote: | | $100 donation to FUMBBL if someone can use the FUMBBl ATI, or whatever, to run this through a year of matches to see what the results look like. |
I'll give it a go. A couple of questions for your spec:
Qualities: For "Poor", are all goblins considered poor, or only goblins <1150 TV? Ie. is the <1150 specifically for vampires, or stunties as well?
Normal: Define a MAJOR difference.
Borked: I'm guessing this is better than normal? Are amazons at 1150 considered better quality than any other group? for example amazons at 2150tv?
Formula, you're saying CR+(1.0M); is this M supposed to designate a decimal 1.0, or something else?
In the end, this is easy enough to implement and should run quickly given that I don't need to persist the result.
Is a tab separated txt result good enough? Coach, tier, CR? I'll run it over everything 2016 and later most likely. |
|
|
thoralf
Joined: Mar 06, 2008
|
| Posted:
Oct 18, 2017 - 17:02 |
|
| koadah wrote: | | But I think that most people would rather just have another beer. |
I too, if by beer you mean coffee.
It'd be hard to estimate "functional ratings" for teams, as the threats and the counters a team has vary from one pairing to the next. For better or for worse, we have TV for pairing purposes.
In any event, it's important to recall that coaches' performances is related to the teams they have. |
_________________
There is always Sneaky Git. |
|
|
|
| |
All Star Bowl!!!
Search
