Divisions, Levels or Sub-leagues?

Please discuss issues concerning the Ducati League here. This is the liaison between league players and the league council.
Post Reply
User avatar
CannonBallGuy
Private First Class
Private First Class
Posts: 2083
Joined: Wed Apr 12, 2006 1:31 am
Contact:

Divisions, Levels or Sub-leagues?

Post by CannonBallGuy »

The Issue:
Chestal brought up an interesting point about veteran teams fighting new teams because they'll be easy to beat but have a ranking of 1200 when they start, meaning the vets get points easily(see here) and we also have the issue of poor teams always getting a thrashing and thus not enjoying the game and quickly leaving.
Some may say that the strong teams get bored of always beating the new teams, but I'm not so sure on that.

My Idea:
Anyway, I figrued we can beat this problem pretty easily, the way it is beaten with sports.
In the UK, we play football (sissy sport, no contact, *yawn*) where there are different divisions. More detail on this can be seen here.

(Note, for all data/examples, I will use GU league which is more active - how we hope the ducati league will be, right? No point using the unactive league for examples of how an active league could be structured imho.)

Basically, I propose all teams start with 1200 points, as they do now, but this puts them in Division/Level/Sub-League #2 with other teams ranked between... lets say 1150 and 1250 - meaning they have about 6 or 7 other teams to match. They match these teams, and only these teams and points are awarded however points will be awarded.

If they do well, they will get over 1250 points for their rank. This results in them moving up to Division/Level/Sub-League #1 and perhaps they can all recieve congratulation BZMails, etc. Now they fight with all teams ranked over 1250, if they continue to do well, then they can become top of the league and whatever...

If they don't do to well, they drop back down into Division/Level/Sub-League #2 which is no problem because here they fight teams of their own level and have more fun than being thrashed all the time by veteran teams, right?

If they did badly in the first place they would have not gone up to Division/Level/Sub-League #1 but rather might have gone down to Division/Level/Sub-League #3 - also not a problem as I suspect most new teams will do this and again they fight other teams at their own level, no thrashing, good fun and they will improve until one day they go up to Division/Level/Sub-League #2, woohoo.

And the process continues, teams changing Division/Level/Sub-League until they find the right one for them...


Possible Problems:
For a start, a team could be stuck between too Divisions/Levels/Sub-Leagues - in other words, they could have a rank of about 1250, everytime they win a match, they move up, then they lose to a strong team and go back down, etc etc. This would be a complete pain in the rear.
Next, there is the issue of new teams being stuck fighting other new teams and not learning any thing good from the veteran teams who also only fight veteran teams, which could also just slow the league down, make the new players/teams get bored, leave, etc.
It's also hard to pick the places where one Division/Level/Sub-League ends and the next one begins. Really, it might be nice if the ratings were spread out more, so that you dont have most of the league with between 1150 and 1250 points... Might be nice to see a range of ranks from 600 to 1800 and then it would be easier to find a place to make the spilt(s), I guess.


Please discuss,
- CBG :)
Image

Merry Christmas!

"Look, if I don't buy booze for the kids, I don't get any incriminating pictures to show to their parents, my business goes down the sink, my girlfriend leaves me and the baby goes on ebay. So help me search..."

"go Play With Toys urself in a dark alley u donkey ******" - Lt-Kirby2007
User avatar
tokimi
Private First Class
Private First Class
Posts: 120
Joined: Sat Nov 06, 2004 7:45 am
Location: Storrs CT
Contact:

Post by tokimi »

Thats a good idea CBG, but I see this being a problem because this qould need lots of active teams. Not sure if its implimentable yet, maybe when we get 30 active teams we can do it :)
The very young do not always do as they are told, and the very old never listen any way.
User avatar
CannonBallGuy
Private First Class
Private First Class
Posts: 2083
Joined: Wed Apr 12, 2006 1:31 am
Contact:

Post by CannonBallGuy »

Yes, that's also an important issue.
You would want, say, 4 pretty active teams in each subdivision - teams that play at least 5 matches a week, so would want at least 12 teams with a consistent activity rating of atleast 0.5 before the idea could work...
Image

Merry Christmas!

"Look, if I don't buy booze for the kids, I don't get any incriminating pictures to show to their parents, my business goes down the sink, my girlfriend leaves me and the baby goes on ebay. So help me search..."

"go Play With Toys urself in a dark alley u donkey ******" - Lt-Kirby2007
mistake
Private First Class
Private First Class
Posts: 124
Joined: Wed Jun 15, 2005 12:12 am

Post by mistake »

A completely different aproach with similar ideas is the following: Challenges Ladder

Given a ladder with teams, a team can challenge only one of those teams that are a maximum of lets say 3 positions up the ladder. If the challenging team looses, nothing changes. If the challenging team wins, it takes over the position of the challenged team resulting in that the challenged team as well as any other team between the challengers position and the challenged position move down one position.
Example:
a) If a team on position 6 challenges the team on position 3 and wins, teams on position 3, 4 and 5 move down a step (thus becoming 4, 5 and 6) and team on position 6 is now on position 3.
b) If a team on position 6 challenges the team on position 5 and wins, they simply swap positions

What it shares with the idea of sub leagues is the rating locality, meaning that you match only teams who's rating is not that far away from yours (same division).

The difference is that the locality is not restricted to certain rating levels (the divisions) but is continously across the ladder.
Another advantage it has over the divsions idea is that it works with less teams.

However there are many other reasons why the challenges ladder idea would not work in the ducati league. Mainly because
1. it was expressed that challenges may be refused which is incompatible with the above Challenges Ladder Idea
2. a team can only find matches when any of the 3 teames ranked above him are present or when it is challenged by a team maximum 3 positions below, which is difficult seeing the current activity. But on the other hand, there are 6 matching possibilities, 3 to challenge and 3 to be challenged by.

So I don't think the above idea works for the league, I posted it anyway so maybe it inspires somebody to come up with an adapted idea that could work for the ducati league.
mistake
User avatar
menotume
Major General
Major General
Posts: 232
Joined: Tue Jul 01, 2003 7:48 pm
Location: SE Pennsylvania, USA
Contact:

Post by menotume »

How about a change to the ELO rating (as suggested by Pimpinella for the 1v1 league) ....

ELO calculation would be simlar to what we have now, but instead of just considering winner/loser, we consider the score difference as well.

If a new team (1200) loses to a better team (1600), but the score was 10-9, the new team would be rewarded.

Likewise, for well established teams... if two 1200 teams match, and the score was 10-1, the winner would gain points, instead of 0 points as it is now.

This also would be easy to code :) All we need is a factor for how much the score difference affects ELO.
User avatar
CannonBallGuy
Private First Class
Private First Class
Posts: 2083
Joined: Wed Apr 12, 2006 1:31 am
Contact:

Post by CannonBallGuy »

menotume wrote:Likewise, for well established teams... if two 1200 teams match, and the score was 10-1, the winner would gain points, instead of 0 points as it is now.
Do I smell a typo? :)
If two 1200 ranked teams match now, the winner would get 25 points, iirc, not 0.
The only time 0 points is given would be if those two teams draw, or possibly if a team with rank of 2000 beat a team with rank of 0, for example.
Image

Merry Christmas!

"Look, if I don't buy booze for the kids, I don't get any incriminating pictures to show to their parents, my business goes down the sink, my girlfriend leaves me and the baby goes on ebay. So help me search..."

"go Play With Toys urself in a dark alley u donkey ******" - Lt-Kirby2007
User avatar
Pimpinella
Private First Class
Private First Class
Posts: 96
Joined: Wed Nov 19, 2003 4:50 pm
Location: Jever, NDS, Germany

Post by Pimpinella »

menotume wrote:How about a change to the ELO rating (as suggested by Pimpinella for the 1v1 league) ....
The document where i originally discussed this can be found at the link below. As you can see it´s about the ZELO-rating used for 1vs1, however the idea behind it can be used for ducati-league qiute as well.

http://balin.pimpi.org/newzelo.txt

Pimpi
Admirarch
Private First Class
Private First Class
Posts: 33
Joined: Sat Sep 11, 2004 9:06 am
Location: Seeking lost whimsy

Post by Admirarch »

This idea would make a cap limit an absolute necessity or it would reward brutality when strong teams play weaker ones. This isn't a downside, I just felt like stating the obvious a bit.
User avatar
zongo
Private First Class
Private First Class
Posts: 106
Joined: Wed Oct 01, 2003 7:26 am
Location: Basel, Switzerland, Europe
Contact:

Post by zongo »

For those who dont know, ZELO rating formula as I implemented 2 years ago in the 1vs1 league does exactly what menotume suggests...

The only problem I see in it is that a player might lose zelo even if he wins (in some rare cases) I made this on purpouse in order to even up scores, it can be easily corrected by simplifying the actual ZELO formula (which is complicated exactly because of these rare cases). This simplification amounts to Pimpi's zelo modification.

For those interested in the original ZELO formula its here

http://1vs1.bzleague.com/rules.php

under the point "What is ZELO?"
visit http://1vs1.bzleague.com
and dub's site http://dub.bzleague.com
join #dub channel on irc!
User avatar
Pimpinella
Private First Class
Private First Class
Posts: 96
Joined: Wed Nov 19, 2003 4:50 pm
Location: Jever, NDS, Germany

Post by Pimpinella »

zongo wrote:For those who dont know, ZELO rating formula as I implemented 2 years ago in the 1vs1 league does exactly what menotume suggests...
I guess you misunderstood that. With ZELO a team losing a match will still lose points, even if they played far better than suggested by the difference of the teams scores.
What meno suggested is, that the winner team will lose points and the loser team win points if the match was less dominated by the winner team than expected.. Please take a look at the link i posted yesterday (and here again) it´s explained much more deeply there.

http://balin.pimpi.org/newzelo.txt

Pimpi
User avatar
Xell
Private First Class
Private First Class
Posts: 45
Joined: Sun May 15, 2005 1:55 pm

Post by Xell »

menotume wrote:How about a change to the ELO rating (as suggested by Pimpinella for the 1v1 league) ....

ELO calculation would be simlar to what we have now, but instead of just considering winner/loser, we consider the score difference as well.

If a new team (1200) loses to a better team (1600), but the score was 10-9, the new team would be rewarded.

Likewise, for well established teams... if two 1200 teams match, and the score was 10-1, the winner would gain points, instead of 0 points as it is now.

This also would be easy to code :) All we need is a factor for how much the score difference affects ELO.
But, on the other hand, that would reward scores like 20-0 again. If an old team with a score of 1600 plays versus a team with a score of 1000, they would be trying to win as high as possible, so that they can gain more than just one or two points.

EDIT: whoops, sorry, Admirarch said that already. But in combination with a cap limit of 10 caps, i really start to like the idea. F.e. if dub plays versus bz-inc: the ratings are 1649 : 1133, which means dub has to win at least 10-7 or 9-6 or 3-2..., to gain points. Sounds like the fairest system i've heard so far.
User avatar
zongo
Private First Class
Private First Class
Posts: 106
Joined: Wed Oct 01, 2003 7:26 am
Location: Basel, Switzerland, Europe
Contact:

Post by zongo »

Pimpinella wrote: I guess you misunderstood that. With ZELO a team losing a match will still lose points, even if they played far better than suggested by the difference of the teams scores.
No, no, no and no!

U misunderstood zelo then!
Players who lose do get zelo points if they played a small difference score with a better player!!!! Gilly for example last week lost 10-9 vs me and increased his zelo score. For "worse" player they can still win points even if they lose 10-6 or so!

That's exactly what i wanted for ZELO to be and thats how it always worked!!!
And again i refer to the 1vs1 rules page for the original zelo formula...:)
visit http://1vs1.bzleague.com
and dub's site http://dub.bzleague.com
join #dub channel on irc!
User avatar
zongo
Private First Class
Private First Class
Posts: 106
Joined: Wed Oct 01, 2003 7:26 am
Location: Basel, Switzerland, Europe
Contact:

Post by zongo »

and i certainly do not agree that the winner might lose points if he doesnt score enough, this is certainly a bad idea, like punishing inactivity.
Thats the only thing I would change in the actual zelo rating (though these are really extreme cases)
It's better to reward the weaker team if they play well vs a great opponent and to give small (or no) points to the winner if the opponent was really weaker.
visit http://1vs1.bzleague.com
and dub's site http://dub.bzleague.com
join #dub channel on irc!
mistake
Private First Class
Private First Class
Posts: 124
Joined: Wed Jun 15, 2005 12:12 am

Post by mistake »

Some questions for my understanding:

1. what does count? the number of caps difference or the caps ratio? So would 2-1 be the same as 4-2, or would 2-1 count the same as 4-3?

2. would you know what your score should be to increase you're rating? Lets say you lead 4-2 vs another team, and still 10 minutes to go. How do you know the lead is enough or not?

3. how would it work with shorter matches. Lets say you have to win with a lead of 8 points to win. That's easier in a 30 min match than a 15 minute match if you are better than your opponent.

4. With a team zelo rating (so where not the zelo rating of the individual players is taken), Given a team with a high rating plays vs a team with low rating, and given that the better team would have to win with 5 caps difference. Given also that the team members in each team are not homogeniously strong. The weaker players of the better team will hardly have a chance to make those 5 caps difference. So a good team would never take on new players since each time they play, they only loose points., even if they do win. How could that be made to work?

5. (edit) What happens with matches added or deleted in the passed? Could it lead to change of preconditions for later matches? Meaning that when you thought a 4-2 win was enough, it wasn't really because some matches were not entered yet.

Also some players like me and menotume would have to dramatically change the playing style and not start to shoot each other in the last 10 minutes of a match.

I didn't know that the 1vs1 worked that way, i never tried to win with a maximum lead as possible.

edit Ah, Zongo just answered maybe some points

---

I looked on the 1vs1 site, and it said:

# the winner gets 3 points while the loser gets 1 point
Where is the zelo thingy?

Another point: I am not sure if loosing zelo points for a new team is so bad when they loose, not so sure if they mind, i understood that loosing 10-0 continously is what is annoing them.
Correct me if I am wrong.
mistake
User avatar
zongo
Private First Class
Private First Class
Posts: 106
Joined: Wed Oct 01, 2003 7:26 am
Location: Basel, Switzerland, Europe
Contact:

Post by zongo »

just a quick reply here (i'll answer a bit more prolly later)

In one of my previous posts i said that rules must be as simple as possible.
That's why i said somewhere that zelo system shouldn't be so important.
Mistake's questions arise some of the problems that such a rating would imply. In the 1vs1 league i tried to keep obscure the ZELO formula and to emphasize points ladder in order to make it simple.

for mistake: the zelo formula is explained in the 1vs1 rules page linked above, under the point 7.
visit http://1vs1.bzleague.com
and dub's site http://dub.bzleague.com
join #dub channel on irc!
User avatar
Pimpinella
Private First Class
Private First Class
Posts: 96
Joined: Wed Nov 19, 2003 4:50 pm
Location: Jever, NDS, Germany

Post by Pimpinella »

All answers refer to my paper, Zongo should probably answer for Zelo as there are some constants in his formula that i don´t actualy understand...
mistake wrote:Some questions for my understanding:

1. what does count? the number of caps difference or the caps ratio? So would 2-1 be the same as 4-2, or would 2-1 count the same as 4-3?
I thought the rating should represent an obvious defference in strength, so for every 100 points the higher rated player would be expected to get 1 cap/kill difference, so 2-1 would be the same as 4-3. However i thought about 1vs1 when i wrote it, and matches always go to 10 there. Using the cap ratio might make more sense for timed games. Probably in a way, that 200 points diffenence means the higher rated team is expected to win with a 2-1 cap ratio. There would need to be found a solution for games where the losing team doesn´t get at least one cap. Also if games with defnesive playing teams will lead to higher ratios.
After all i´d still prefer using the difference in number of caps, it´s easier to understand and less problematic.
2. would you know what your score should be to increase you're rating? Lets say you lead 4-2 vs another team, and still 10 minutes to go. How do you know the lead is enough or not?
Yes. As the difference in ratings represents a defined difference in strength you can predict the score you need pretty easily.
3. how would it work with shorter matches. Lets say you have to win with a lead of 8 points to win. That's easier in a 30 min match than a 15 minute match if you are better than your opponent.
Thats no problem if cap ratios are used, using absolute score difference the result might have to be doubled.
4. With a team zelo rating (so where not the zelo rating of the individual players is taken), Given a team with a high rating plays vs a team with low rating, and given that the better team would have to win with 5 caps difference. Given also that the team members in each team are not homogeniously strong. The weaker players of the better team will hardly have a chance to make those 5 caps difference. So a good team would never take on new players since each time they play, they only loose points., even if they do win. How could that be made to work?
I don´t second that. HoW often took new players and they were always strong because they did good teamwork. However they could of course get a bonus for having a player less than ? weeks or something similar. I don´t like the idea though.
5. (edit) What happens with matches added or deleted in the passed? Could it lead to change of preconditions for later matches? Meaning that when you thought a 4-2 win was enough, it wasn't really because some matches were not entered yet.
Yes, that could be problematic.

Pimpi
User avatar
Pimpinella
Private First Class
Private First Class
Posts: 96
Joined: Wed Nov 19, 2003 4:50 pm
Location: Jever, NDS, Germany

Post by Pimpinella »

zongo wrote:No, no, no and no!

U misunderstood zelo then!
Yes, probably (or did you change it at some point?) There´s an C and S in that formula i don´t really get.
and i certainly do not agree that the winner might lose points if he doesnt score enough, this is certainly a bad idea, like punishing inactivity.
Yes, that might need further explanation. If the willer must score enough in order to keep his current score every match is a real challenge, and even more important every match beares the same risk. We´ve seen teams in the past who played against weaker teams, not against stronger though, and always asked for the players that would play for the opposing team and denying if they thought they might loose.
This way they take points from below, but never risk to lose them again. TAA was an example for this.
If you lose points to an opponent even if you´ve won the match because you didn´t win high enough you always go the same risk in every match, wether you´re play the leagues No1 or No15 - you just have to be as good as your rating suggests to keep your rating.
New teams are now motivated to play strong teams while the strong teams still have the kick of playing a hard match even if they are superior.

Pimpi
User avatar
Xell
Private First Class
Private First Class
Posts: 45
Joined: Sun May 15, 2005 1:55 pm

Post by Xell »

Pimpinella wrote:
zongo wrote:No, no, no and no!

U misunderstood zelo then!
Yes, probably (or did you change it at some point?) There´s an C and S in that formula i don´t really get.
and i certainly do not agree that the winner might lose points if he doesnt score enough, this is certainly a bad idea, like punishing inactivity.
Yes, that might need further explanation. If the willer must score enough in order to keep his current score every match is a real challenge, and even more important every match beares the same risk. We´ve seen teams in the past who played against weaker teams, not against stronger though, and always asked for the players that would play for the opposing team and denying if they thought they might loose.
This way they take points from below, but never risk to lose them again. TAA was an example for this.
If you lose points to an opponent even if you´ve won the match because you didn´t win high enough you always go the same risk in every match, wether you´re play the leagues No1 or No15 - you just have to be as good as your rating suggests to keep your rating.
New teams are now motivated to play strong teams while the strong teams still have the kick of playing a hard match even if they are superior.

Pimpi
One thing which made me wonder abou this, is the fact that you always have to be better than before. To win points, you always have to play better than your rating shows you are - thats a bit weird. Does this calculation imply eternal improvement?
User avatar
Pimpinella
Private First Class
Private First Class
Posts: 96
Joined: Wed Nov 19, 2003 4:50 pm
Location: Jever, NDS, Germany

Post by Pimpinella »

Xell wrote:One thing which made me wonder abou this, is the fact that you always have to be better than before. To win points, you always have to play better than your rating shows you are - thats a bit weird. Does this calculation imply eternal improvement?
No, you don´t have to be always better. there´s an expected result, represented by the modifier which is calculated from the difference of ratings. I suggested :
int((rating stronger - rating weaker)/100)
for this, so it´s an integer, not a float (there might be a better way to calculate this though...).
if the modified match result id a tie, the difference in strength between the teams is properly represented by the ratings and neiter the winner nor the loser team´s rating changes.

Now, it´s only a draft of course, a lot of fine tuning might be neccessary.

Pimpi
User avatar
zongo
Private First Class
Private First Class
Posts: 106
Joined: Wed Oct 01, 2003 7:26 am
Location: Basel, Switzerland, Europe
Contact:

Post by zongo »

a few more things about zelo:

- it depends only on caps/kills difference not ratio

- i tested different formulas before the 1vs1 league even started, with a friend we tested on a sample of about 5000 randomly generated games/results.

- the zelo formula never changed once adopted.

- The C in the zelo formula is not a constant...it depends on the score difference
visit http://1vs1.bzleague.com
and dub's site http://dub.bzleague.com
join #dub channel on irc!
User avatar
SportChick
Private First Class
Private First Class
Posts: 300
Joined: Wed Jan 21, 2004 2:57 am
Location: California
Contact:

Post by SportChick »

menotume wrote:ELO calculation would be simlar to what we have now, but instead of just considering winner/loser, we consider the score difference as well.
I've been sifting through the posts and must say that I have serious reservations about moving to a point differential ELO. Even if we included a cap limit, strong teams would be motivated to rush to hit the cap limit (thereby ending the match with a maximum point difference), rather than play at a more normal pace. Without a cap limit, scores of 20-0 would/could become quite commonplace. While I agree with looking at all alternatives, this one seems to aggravate some of the other issues that have been raised, rather than help solve them in concert.
Image
User avatar
tokimi
Private First Class
Private First Class
Posts: 120
Joined: Sat Nov 06, 2004 7:45 am
Location: Storrs CT
Contact:

Post by tokimi »

Teams may run to the cap limit any way, if thats a sure fire way of ending the
match quickly. Though I am not sure if counting the actually number of caps
in a match in the ELO calculation is good, because that makes the difference
between a 15 minute and a 30 minute match much more noticable. Unless a
proper way of scaling between the two match times, counting the cap score
seems to be a problem waiting to emerge.
The very young do not always do as they are told, and the very old never listen any way.
User avatar
zongo
Private First Class
Private First Class
Posts: 106
Joined: Wed Oct 01, 2003 7:26 am
Location: Basel, Switzerland, Europe
Contact:

Post by zongo »

my opinion:

don't get into trouble with a complicated zelo thingy (like i did ;) )
Emphasize points ladder and keep running the actual elo rating (or a slightly modded version but keep it simple)
visit http://1vs1.bzleague.com
and dub's site http://dub.bzleague.com
join #dub channel on irc!
User avatar
t2m
Private First Class
Private First Class
Posts: 6
Joined: Fri Oct 21, 2005 9:08 pm

Post by t2m »

Admittedly I have not read the majority of this post and I apologize in advance if this is similar to any other posted, however, for simplicity sake, why not: Win = 1, Tie = 0, Loss = -1 (points)?
I understand different skill levels should be rewarded differently, but for the sake of simplicity, could this not work?

Unskilled teams may be more willing to play any team knowing they can only lose 1 point, skilled teams are more inclined to play any team as they can only gain 1 point. Not exciting I know, but simple.

(Great ideas from the few posts I have read!)


t2m
User avatar
CannonBallGuy
Private First Class
Private First Class
Posts: 2083
Joined: Wed Apr 12, 2006 1:31 am
Contact:

Post by CannonBallGuy »

t2m, the idea of giving 4 for a win, 2 for tie and 1 for a loss encourages matches in general, even if a team will lose, they get a point, so it's still worth it. With your method, teams could still be less-keen to match if they might lose points.
Image

Merry Christmas!

"Look, if I don't buy booze for the kids, I don't get any incriminating pictures to show to their parents, my business goes down the sink, my girlfriend leaves me and the baby goes on ebay. So help me search..."

"go Play With Toys urself in a dark alley u donkey ******" - Lt-Kirby2007
Post Reply