Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[soulnet]

A couple of comments:

A. I think specification of how divisions are divided in non-top level is required. Specifically: Non-promoting non-demoting people will definitely/maybe/randomly stay together for next season? I think this should be:
1. Decide whether existing divisions will mantain themselves and then assign new people (promoters, demoters, newcommers) in some way, or just take the entire pool of players in a given level and redraw the divisions each season.
2. To draw or assign new players, either use some algorithm considering timezones or pure random (anything else seems unnecessarily convoluted).

I would go for not mantaining divisions, because it is so much easier to process people leaving/joining in between seasons.


B. What happens with the bottom divisions? Each get floor or ceiling of k/nr of divisions, where k is the number of people in the bottom level? That seems like it could give some stupid things like divisions with 2 or 3 people in. I would just have less divisions in the bottom level and try to have 4/5/6/7 people in each division, possibly making only #6 demote from some next-to-last divisions and even have no demotion in some (randomly assigned) if necessary.


C.

2. If a good new player joined, we try to insert at a sensible level

I really dislike this. I understand the reasons for it, but I think it is a too sensitive issue to leave at the organization's will. Basically, I am worried that some people can be easily offended. So, I would either specify a deterministic algorithm beforehand or insert people disregarding "good player" considerations other than last performance in the league. Here is a possible algorithm:

Process levels from top to bottom. If there is a hole and there is a returning player who had that level or higher at the time of departing, put them there. If there are still holes, put highest Isotropish ranking newcomer if it is higher than all possible promoters (all #2/3/4 still available for wildcard promotion). If there are still holes, promote extra people from the immediately lower level (use divisions in random order, first all #2s, then all #3s and so on, and producing new holes to be processed when is that level's turn).

This makes returning players to be inserted at their old positions or worse, prefering promoting existing players if a too high spot is available.

I would still prefer Isotropish (or Goko Pro or whatever other external ranking) to be disregarded after season 1 and just place newcomers randomly. They would raise to the top by promotion soon enough anyway, and makes the league independent from external sources and also the placement algorithm way less convoluted.

D. I think a ruling on Salvager usage and specifically VP counter for online games should be done (i.e., what happens if players do not agree on something). Possibly also for identical starting hands, given that it seems like it would be technically possible in the near future. In all cases, I think both player's agreement overrides the ruling.

E. I think a ruling on a couple of players not being able to play a match should be done. Moreover, I would force some schedule, like you have to play at least half the matches before the last week, to avoid people playing everything on the last day, possibly earning an advantage from knowledge of the current status.

F. I think mandatory/optional/forbidden publicness of results should be explicitly stated. There is reason not to publicly discuss results, because the league is asynchronous. However, I think this is really bad because it stops people from discussing matches online, which is a great thing about playing tournaments.

G. Since several new players and some old players believe that score difference is good to take into account, I would explicitly say (even if it is redundant) that the final score does not matter. Also, I would say tied score counts as win for the player who played less turns, if such player exists (because "tie" in the rules could be interpreted as tied score instead of tied game by the rules of Dominion). I would just say "shared victories" in the rules instead of ties, to make it clear that the official rules apply. Technical thing: Can we just award lone victories 2 points and shared victories 1 point? I don't like the halves if we can avoid them.

[Tables]

I think Soulnet raises some very good ideas. It might be worth considering having a 'suggested schedule' for each league, where if you have say players A, B, C, D, E, F then if possible the matches should be played:

Week 1:
A vs. B
C vs. D
E vs. F

Week 2:
A vs. C
B vs. E
D vs. F

Week 3:
A vs. D
B vs. F
C vs. E

Week 4:
A vs. E
B vs. D
C vs. F

Week 5:
A vs. F
B vs. C
D vs. E

I just made that from scratch but with a spreadsheet it should be very easy to generate something like that automatically for each league. With something like that in place it encourages players to play their matches in a fair(er) timing schedule, and if someone has gone multiple weeks out of sync then they it can be investigated.

Timing aside, one other thing I'm slightly less enthused about is the exponental structure of number of leagues in each division. I realise it makes things as easy as possible but it means around half of players sit in the lowest division and the vast majority in the bottom two. Now maybe this won't be too big of an issue if we only have enough people playing for around 3 tiers (19-42 people), but if it reaches 5 tiers (91-186 people) or even 6 tiers (187-378 people) I think it's kinda disappointing for the majority who sit so far down. Sadly I don't have a great suggestion for a better way of organising the leagues, but one possibility is a structure something like 1 - 2 - 2 - 4 - 4 - 8 - 8 - ..., where every second league doesn't increase in size. This mitigates the problem somewhat and makes a somewhat lower proportion sit near the bottom. Most likely you could even have the number of leagues at each place just cap at 8 as well, that'd probably be fine.

[SCSN]

2. If a good new player joined, we try to insert at a sensible level

I really dislike this. I understand the reasons for it, but I think it is a too sensitive issue to leave at the organization's will. Basically, I am worried that some people can be easily offended. So, I would either specify a deterministic algorithm beforehand or insert people disregarding "good player" considerations other than last performance in the league.

I really disagree with this. It will be by far the most fun for everyone if all players are matched up with opponents as close to their skill level as possible. If you insert a high-ranked player into the lowest league that will create local distortions over the course of many seasons, putting players matched up with him at a significant disadvantage relative to their peers in parallel leagues.

I really don't see the potential for offending people. Well, some people will take offense at basically everything, but that's their own problem, not ours.

but I think it is a too sensitive issue to leave at the organization's will.

I trust Stef's judgment in this matter over any algorithm.

[Voltaire]

3. for each game you won, add the total score of that opponent

Rather than a suggestion, I simply have a question: what does this mean? It's unclear to me.

[WanderingWinder]

3. for each game you won, add the total score of that opponent

Rather than a suggestion, I simply have a question: what does this mean? It's unclear to me.

It means that if I won 4 games against player X, and player X won 12 games in the league, I get 48 tiebreaker points from my matchups with player X.
It's a strength of schedule thing.

[EgorK]

3. for each game you won, add the total score of that opponent

Rather than a suggestion, I simply have a question: what does this mean? It's unclear to me.

It means that if I won 4 games against player X, and player X won 12 games in the league, I get 48 tiebreaker points from my matchups with player X.
It's a strength of schedule thing.

Considering everyone has same schedule (well, excluding themselves) it just assumes win against strong player and lose to weak is better then win against weak and lost to strong

[soulnet]

Considering everyone has same schedule (well, excluding themselves) it just assumes win against strong player and lose to weak is better then win against weak and lost to strong

Yes, I was thinking that, is that reasonable? Or at least, more reasonable than the opposite? I would ask for scrapping that tie-breaker, unless someone has a good argument for using it in this all-vs-all setting.


Also, do all 6 matches have to be played on the same day? I guess not, but it would be nice to state it.

[EgorK]

Considering everyone has same schedule (well, excluding themselves) it just assumes win against strong player and lose to weak is better then win against weak and lost to strong

Yes, I was thinking that, is that reasonable? Or at least, more reasonable than the opposite? I would ask for scrapping that tie-breaker, unless someone has a good argument for using it in this all-vs-all setting.

Well, in chess it is used in almost all tournaments, so it should not be completely unreasonable

[soulnet]

Considering everyone has same schedule (well, excluding themselves) it just assumes win against strong player and lose to weak is better then win against weak and lost to strong

Yes, I was thinking that, is that reasonable? Or at least, more reasonable than the opposite? I would ask for scrapping that tie-breaker, unless someone has a good argument for using it in this all-vs-all setting.

Well, chess use it in almost all tournaments, so it should not be completely unreasonable

It is reasonable for swiss system, but seems unreasonable for all-vs-all.

[EgorK]

Considering everyone has same schedule (well, excluding themselves) it just assumes win against strong player and lose to weak is better then win against weak and lost to strong

Yes, I was thinking that, is that reasonable? Or at least, more reasonable than the opposite? I would ask for scrapping that tie-breaker, unless someone has a good argument for using it in this all-vs-all setting.

Well, chess use it in almost all tournaments, so it should not be completely unreasonable

It is reasonable for swiss system, but seems unreasonable for all-vs-all.

I mean like Candidate Tournament use it, which is round robin. Also Solkoff is better for Swiss than Sonneborn-Berger

[Kirian]

Tables, both of your suggestions break the league structure and purpose.  The whole point of a longer season with round robin is to not have to play specific people on specific weeks and allow for flexibility of scheduling.

The whole point of the exponential increases is that, well, it's a king of the hill structure with a literal pyramid.  Doing it without that is basically just a ladder and we already have two ranking systems for that.

[silverspawn]

2. If a good new player joined, we try to insert at a sensible level

I really dislike this. I understand the reasons for it, but I think it is a too sensitive issue to leave at the organization's will. Basically, I am worried that some people can be easily offended. So, I would either specify a deterministic algorithm beforehand or insert people disregarding "good player" considerations other than last performance in the league.

I really disagree with this. It will be by far the most fun for everyone if all players are matched up with opponents as close to their skill level as possible. If you insert a high-ranked player into the lowest league that will create local distortions over the course of many seasons, putting players matched up with him at a significant disadvantage relative to their peers in parallel leagues.

I really don't see the potential for offending people. Well, some people will take offense at basically everything, but that's their own problem, not ours.

but I think it is a too sensitive issue to leave at the organization's will.

I trust Stef's judgment in this matter over any algorithm.

yea... i have to agree, maybe with the cutback that I could understand it if people are offended.

still

I don't like inserting people in the top division out of nowhere, but in the end the League is supposed to be fun and making a top-ranked player join in the bottom division is fun for exactly nobody.

this basically covers it. you just have to set priorities here.

[EgorK]

I think we need to add that whoever have more cards should host the game. Also eternal question of vpon/vpoff should has default answer

[soulnet]

this basically covers it. you just have to set priorities here.

That's why I proposed an algorithm. If someone gets offended with an algorithm, well, I do not care. Judgement calls like "this is a good player and this other is not" or the like are bounded to upset people. Algorithm is also bounded to upset some people, but at least is something established and sitting out, joining late, etc, can be done at own risk.

Also, I like the basic idea algorithm I proposed, though it can be tweaked.

Important points: it should be explicitly stated that staying players will never be demoted to make room for return players or newcomers. It should also be explicitly stated if return players are going to be judged by their placement before leaving or by their ranking at the moment of rejoining (I advocate for the former).

[Simon (DK)]

II. Ordering
4. Coin flip

What if there's 3 players tied? How is a coin flip going to decide that?

Signing up for season 1 does not imply signing up for season 2, you will have to rejoin in due time.

I think having to sign up for each season will be annoying. As long as someone players all his/her matches in a season, I don't see any reason not to automatically put him/her in the next season.
Instead of this I think you should remove players, who has shown inactivity by not playing all their matches.
I also think this will be easier for the organisation, because there will be more players staying in the league than players leaving the league.

[Tables]

Tables, both of your suggestions break the league structure and purpose.  The whole point of a longer season with round robin is to not have to play specific people on specific weeks and allow for flexibility of scheduling.

Hmm, good point. Like I mentioned though, my idea was to have a suggestion for scheduling, as a means to keep things active and make it so generally people won't have played most of their games at once while others haven't started. I guess I could make that clearer. It wasn't meant to be a 'you must play your games at these times' list, and that would have to be made clear. It also might well be better for the suggestion to try and do maybe two games in the first week, to leave one clear week at the end to finish things off. But either way I can see the argument against this.

The whole point of the exponential increases is that, well, it's a king of the hill structure with a literal pyramid.  Doing it without that is basically just a ladder and we already have two ranking systems for that.

I think you're missing the point here I'm afraid. I like the pyramid structure. I like that spots at the top are harder to get at. I dislike that the vast majority of people are going to sit in the bottom two levels, simply due to that being where the vast majority of leagues are. What I proposed isn't making a ladder, that takes too long to climb your way up out of, but a more gradual pyramid. Ideally I'd suggest every tier being slightly smaller than the one below, but that's pretty much a nightmare to organise fair promotions/relegations in.

Regarding new players joining and/or players returning: I'm in favour of an algorithmic approach to replacing them. Probably base it on the Isotropish level of the player and maybe do something like put them in the highest tier available where their Isotropish is better than 2/3rds of players, but never in the top (top 2?) tiers. Details can obviously be calculated. But details aside you'd want something fair, that's more likely to slightly underestimate than slightly overestimate (so they still have to play their way up a division or two, rather than get placed in over their head). Perhaps a clause that prevents people going up more than one division if they are rejoining (I doubt anyone would game the system in that way, and even less that they'd be capable of a dramatic improvement of position, but best be safe).

[Joseph2302]

Just checking, is it fine for me to schedule all my games in the last 2.5-3 weeks of the season then? I have exams until 3rd June (ish), and so am really busy before that, and free almost all the time after that, so will definitely be able to play them all in that time frame.

[-Stef-]

Just checking, is it fine for me to schedule all my games in the last 2.5-3 weeks of the season then? I have exams until 3rd June (ish), and so am really busy before that, and free almost all the time after that, so will definitely be able to play them all in that time frame.

Yes thats perfectly fine. Especially if you explain it in the scheduling thread for your division, once that is available

[soulnet]

BTW, I have a decent argument (maybe edge casey) on why people should always join at the bottom. Suppose a good female player (according to both Isotropish and Goko Pro) joins for season i and due to lack of higher placed empty spots is placed on a 4th tier division. She wins all her matches and promotes to a 3rd tier division. Now a male really good player, but not as good as the one in the previouse sentences, joins for season i+1, and there are empty spots available in a 2nd tier division. Placing him on a 2nd tier division while she has to play 3rd tier for a season seems unfair. However, algorithmic or case-by-case considerations of double-promotions to fill empty spots seem convoluted and even more prone to questioning the system.

This case can be summarized as: If joining in a non-bottom division is allowed, it could be optimal for a player (assuming playing in a higher division is the goal) to leave and rejoin on the same season. If that is disallowed, there are similar cases where sitting out one season can lead to faster ascension than actually playing and promoting. Even if rating is only used the first time a player joins the league and afterwards previous league placement is used, more convoluted examples like the one I laid down above still make the system seem unfair.

One solution would be to go back to "people joins at a bottom division" so rankings are only used in the first season and there no cross-season comparisson of rankings, which lead to swingyness as described.

Another solution would be to consider everyone, not just newcomers, as candidates for filling empty spots with a ranking-based consideration. But this is a pretty bad alternatie: the extreme of this basically makes the league be reseeded every season, which is of course absurd. I think the whole problem arises because of trying to mix two ranking systems (league and Isotropish or Goko Pro) into one decision algorithm (even if its not algorithmic) yields problems regarding the weight of each, which were not addressed in any post so far on this topic (and, my personal opinion is that it is pretty hard to address properly, if at all possible).

[EgorK]

I changed my mind about adding new people to non-bottom division and agree with soulnet. What I propose is following algorithm to fill vacations in order of priority:

1. Sit-outs not higher than their division
2. Second places in next division ranked by Isotropish rating
3. Third places and so on

[soulnet]

I changed my mind about adding new people to non-bottom division and agree with soulnet. What I propose is following algorithm to fill vacations in order of priority:

1. Sit-outs not higher than their division
2. Second places in next division ranked by Isotropish rating
3. Third places and so on

I agree. We may consider adding #5s non-demoting to the pool of #2 extra promotions, I am fine with either option being prefered or both being compared by Isotropish.

[Tables]

I think the argument Soulnet puts forward seems like a rare case which, at most is likely going to end up with someone entering at a league only one place lower than they might have been by waiting a season - and to me that seems like firstly a minor issue compared to lower leagues often being distorted by players far too good for them (and worse the possibility two very good players are placed (or promote into and meet) in the same low division, leaving everyone in that league little hope of promotion for two seasons running (assuming the leagues aren't reshuffled between seasons) and leaving one to wait a whole extra season to gain their promotion.

I think that for returning entrants, they should never be placed higher than they were when they left*, and should be slotted into the highest available open slot before new entrants. In the case of multiple people trying to return at the same time and there not being enough open slots in the division they want to be re-entered, probably go by who played most recently, then by who finished higher in their league last time they played, then randomly.

For new entrants, do some kind of algorithm that determines which division is appropriate for them - and as I said before, I think it's better to underestimate slightly than overestimate slightly. Then put them into the highest available division that isn't higher than the calculated one, breaking ties by something (Isotropish ranking probably).

After new/returning slots are filled, I think you'd probably go to filling vacancies with extra promotions based on something like closest runner ups and the like.

Also, I think it might be best if new/returning players aren't added into the top level, even if their skill indicates they should. Those empty spots should, I think, always need to be played for via at least one promotion, since that is the top division.

[soulnet]

I mostly agree with Tables. Maybe shuffling divisions is a good idea, and it could help solve faster some division being overflowed with too many "too good" players.

[shraeye]

I'm willing to go as is, and adjust if people really feel it's necessary.  I think it's just a fact of life that some of the divisions in a lower tier will be better than others.  No way around it.  Relegation/Promotion should help slightly in adjusting that.  But striving for perfect balance across a tier is just impossible.  Variations exist.

I'm just going to assume that people like to play dominion, and they won't leave for a season and re-enter just to game the system.

[soulnet]

I'm just going to assume that people like to play dominion, and they won't leave for a season and re-enter just to game the system.

Just to clarify: My point is that being ungameable is a desirable property. Usually being able to game the system is saying something bad about the reasonability of the system, even if the gaming itself is not a practical problem.

[Tables]

I've just realised I was going to add a note to my previous post and forgot to add it, it's not too important but I'll add it here instead for completeness:

I think that for returning entrants, they should never be placed higher than they were when they left*

*Unless they finished first in the season they left, then they could be treated as having been in the higher division upon leaving, and vice versa for finishing in the relegation places.

[mail-mi]

Question: Can we request bye's if we dont have time to do a match?

[shraeye]

Question: Can we request bye's if we dont have time to do a match?

Half-point splits awarded by mutual agreement?  1.5 points to each side for the match of 6 games? But I think the flexibility of being able to schedule the matches ANY time during the season should help byes be needed very infrequently.

[Joseph2302]

Question: Can we request bye's if we dont have time to do a match?

Half-point splits awarded by mutual agreement?  1.5 points to each side for the match of 6 games? But I think the flexibility of being able to schedule the matches ANY time during the season should help byes be needed very infrequently.

Surely people can find enough slots in 5-6 weeks to play all their games? If not, I think the above idea is good, so as not to get weird results in the week i.e. people agreeing to draw, as it is mutually beneficial.

[mail-mi]

Question: Can we request bye's if we dont have time to do a match?

Half-point splits awarded by mutual agreement?  1.5 points to each side for the match of 6 games? But I think the flexibility of being able to schedule the matches ANY time during the season should help byes be needed very infrequently.

Surely people can find enough slots in 5-6 weeks to play all their games? If not, I think the above idea is good, so as not to get weird results in the week i.e. people agreeing to draw, as it is mutually beneficial.

Okay thanks.

[-Stef-]

Thanks to everyone posting in this thread. All suggestions/comments are still welcome. Some of them made me change the rules.

This one feels like the most important one to address:

this basically covers it. you just have to set priorities here.

That's why I proposed an algorithm. If someone gets offended with an algorithm, well, I do not care. Judgement calls like "this is a good player and this other is not" or the like are bounded to upset people. Algorithm is also bounded to upset some people, but at least is something established and sitting out, joining late, etc, can be done at own risk.

Also, I like the basic idea algorithm I proposed, though it can be tweaked.

I read your concerns, I understand them, and yet I'm not going to write down an algorithm. I think this would result in too much loss in flexibility, and we would end up with players in divisions where everyone would say "yeah, it is according to the rules, so we can't possibly be against it, but it doesn't feel fair at all and it definitely isn't fun for the players in that division". Also, if you write down an algorithm, you sort of invite people to game that system.

If the league would be 100 times bigger it would be impossible, but for now I really feel we're best off with a couple of guidelines. And off course I'll post every decision there is to post about.

Important points: it should be explicitly stated that staying players will never be demoted to make room for return players or newcomers. It should also be explicitly stated if return players are going to be judged by their placement before leaving or by their ranking at the moment of rejoining (I advocate for the former).

I agree with your first statement. If I ever *really* want to make room for someone, this can only be done if the extra demotion is announced before the season even starts. You will never demote 'out of the blue'. The only exception here may be if an entire new level is created (more on that below).

There is a problem with looking at 'placement before leaving'... maybe the league is now a lot stronger/weaker then it used to be.

Yes, I was thinking that, is that reasonable? Or at least, more reasonable than the opposite? I would ask for scrapping that tie-breaker, unless someone has a good argument for using it in this all-vs-all setting.

Yes, it's arbitrary, but that's why two more sensible criteria are in front of it. It's exactly as good as the opposite would be, I just prefer it to coin flipping.

Also, I think it might be best if new/returning players aren't added into the top level, even if their skill indicates they should. Those empty spots should, I think, always need to be played for via at least one promotion, since that is the top division.

agreed.

Question: Can we request bye's if we dont have time to do a match?

As a general rule of thumb: no, you can not.
Please don't postpone your matches to the last possible date, and you should be fine. 5 weeks is pretty much.
If something really bad happens in real life, you can always explain it a bit and we'll probably figure something out.


about the possibility of restructuring.
For now it looks like we'll start with a lot of very good players. So no problems whatsoever in the top 2 (3?) levels.
However, the lowest level may have a huge diversity in players strength. If it stays that way, that's no fun, and we may have to change the nature of the pyramid a bit.
The most likely change now would be that in stead of 8 level-4 divisions, we'd go with 4 level-4 divisions and 4 level-5 divisions.
Season 1 will not change, and if I want to change season 2 that will only happen if I announce it before season 1 starts.

[-Stef-]

about the possibility of restructuring.
For now it looks like we'll start with a lot of very good players. So no problems whatsoever in the top 2 (3?) levels.
However, the lowest level may have a huge diversity in players strength. If it stays that way, that's no fun, and we may have to change the nature of the pyramid a bit.
The most likely change now would be that in stead of 8 level-4 divisions, we'd go with 4 level-4 divisions and 4 level-5 divisions.
Season 1 will not change, and if I want to change season 2 that will only happen if I announce it before season 1 starts.

We're approaching the end of signups and I want to make a decision on this matter soon.
The number of participants looks really nice. Currently it's 6+12+24+35.
Maybe a few more in the last days but I don't expect to reach the awkward 6+12+24+48+1.

If we look at the levels in the different divisions, it will be approximately 47-54, 39-47, 30-39, 0-30
At this moment I'm leaning towards 'no split-up of D division unless the league grows to 90-100 people'.
Argumentation: even though 0-30 looks like a huge range, most of those players will be 25-35 in reality now or really soon.
Quite some of them will be stronger then their rank, either due to inactivity or due to being really new.
If they did sign up despite being completely new to online dominion, they're probably looking for a challenge / learning experience.
Basically I expect a lot more fun competitive matches in D division then easy 6-0 victories, which would be the only real argumentation to split it up.

If you have a strong opinion about this though, please post it here and I'll reconsider.

And another point: The number of players in division D is probably not dividable by 6, and then I'd create some 5-player pools and some 6-player pools.
(rather then 7-player pools, or 6-player pools and one pool with 3) Any thoughts on this?

[Hugovj]

Well, I was wondering: will you always hold on to 8 groups in D, or is it possible there'll be 6? With 35 participants in this level, you'll get 3 groups of 5 and 5 of 4 if you'd go for 8, if you'd go for 6 leagues you will have one with 5 and 5 with six people.

If you go for 6, you can do something like advancing the two best numbers 2 to match the difference in promotion/degradation.

[-Stef-]

Well, I was wondering: will you always hold on to 8 groups in D, or is it possible there'll be 6? With 35 participants in this level, you'll get 3 groups of 5 and 5 of 4 if you'd go for 8, if you'd go for 6 leagues you will have one with 5 and 5 with six people.

If you go for 6, you can do something like advancing the two best numbers 2 to match the difference in promotion/degradation.

I will not always hold to 8 groups in D. It depends a bit on the number of participants, but I prefer to have 6-player groups.
If less then 8 groups are created, the best #2 will promote indeed, although it's not that trivial to determine who's the best #2.
I don't like "The one with the most points", because they play against different people entirely.

Maybe I should copy the leaderboard once more halfway during season 1, and then promote the highest #2 on that leaderboard?
This is off course assuming we won't have 8 groups (Currently 39 people in D, which would be 4 6-player groups and 3 5-player groups)

[Hugovj]

Well, I see three possibilities for choosing who to promote (in case of less than 8 groups):

1. You can play play-offs
2. You can choose League prestation as tiebreaker.
3. You can choose a ranking (Iso) as tiebreaker

I'd be a fan of play-offs, but this was voted off, so I won't suggest we should introduce them again. If you promote the highest ranked, you mix up the League and something different, and highly disadvantage newer players. If you are good at the League, you should be promoted accordingly I think. The second one I like best. 'The one with the most points' is maybe not the best, but Solkoff-Scores (or something different) sound reasonable to me. Obviously you can't level out all the groups, but you have to have something, and I think it's more fair than basing it on Iso-ranking.

[SCSN]

Something that sounds fair to me is to compare the iso ranking of all #1s and pick the one with the highest ranking. Then the best #2 is he who finished 2nd in the highest ranked #1's group.

Alternatively you could pick as best #2 he who won the most games against the #1 of his group, with perhaps using the #1s iso-ranking as a tie-breaker.

[serakfalcon]

Something that sounds fair to me is to compare the iso ranking of all #1s and pick the one with the highest ranking. Then the best #2 is he who finished 2nd in the highest ranked #1's group.

Alternatively you could pick as best #2 he who won the most games against the #1 of his group, with perhaps using the #1s iso-ranking as a tie-breaker.

I like this in theory, but it has the potentially undesirable quality of having rank in the league partially determined by non-league games.

[SCSN]

Well, you need some way to compare players across different groups. Since there is no league data to go by, the only way to not use an external measure is to just pick someone randomly or use a meaningless metric like total matches won. Once multiple seasons are played you could potentially run the isotropish algorithm over past league matches only, but that sounds like quite a bit of work for very little benefit, and I suspect it would take a long time for this thing to be morally superior to the true isotropish rankings.

[soulnet]

morally superior

Wow, I would have never suspected you were a religious guy.

Anyhow, I think total league points is still better than ranking (Isotropish or any other), even though the counterpoints against it are valid. Rankings are not really accurate (I have already argued this). We are already using pseudorandom metrics (see discussion above regarding tiebreakers) and this seems like a decent one (if the groups are random, opponent's should be reasonably similar in strength, and there is luck on getting an easy group, but there is luck in that to get 1st place too, and there is luck in Dominion in general). Competitions with a lot more at stake use this kind of comparison (like European World Cup qualifiers if I am not mistaken), so, why shouldn't we?

[SCSN]

morally superior

Wow, I would have never suspected you were a religious guy.

I'm perfectly indifferent to religion, though I wouldn't mind having a few beers with some of the good old Greek and Roman Gods. I quite enjoyed many of their tales I had to translate in high school, and their Judeo-Christian successor always stroke me as a dreadful bore.

Growing up in a bible-belt town I learned at an early age that any positive correlation between religiosity and moral behavior exists exclusively in a staunch believer's head, and what I mean to connote by "morally superior" is just something like "better aligned with the intended purpose".

[GeoLib]

morally superior

Wow, I would have never suspected you were a religious guy.

Why do considerations of morality have to be rooted in religion?

[-Stef-]

On a side note, this discussion may very well be pointless (not just the religious one). We now have enough people in D division to make 8 groups. So the outcome of this discussion will only matter if we get enough people for an E division. That in fact starts to look like a realistic possibility now, and even if it doesn't it could happen next season.

I don't think the Isotropish leaderboard is 'not really accurate'. It's the most accurate thing existing that measures peoples skills. I also don't think it's unfair to newer players. If I would start a new account now, I think I'd be in the top 5 within 2 days if I played a lot. However, using an 'old leaderboard' is unfair to new learning players, like what we have now with seeding for season 1.

I like this in theory, but it has the potentially undesirable quality of having rank in the league partially determined by non-league games.

I agree it's a disadvantage, but it's still (a lot) better then something close to entirely random.

I don't want to use a leaderboard that's too old, but I also don't want the promotions to be decided by a still changing leaderboard after the last match in the league. That would lead to highly undesirable scenario's where people could want to postpone their last match in order to outrank someone first. Balancing between those two, my current suggestion is to copy the leaderboard 3 weeks into the season = 2 weeks before it ends.

Something that sounds fair to me is to compare the iso ranking of all #1s and pick the one with the highest ranking. Then the best #2 is he who finished 2nd in the highest ranked #1's group.

I like this one. Either looking at #1's rank, or #2's rank, or their combined rank... would all be fine with me.

Alternatively you could pick as best #2 he who won the most games against the #1 of his group, with perhaps using the #1s iso-ranking as a tie-breaker.

I strongly dislike this one. Suppose I would place a new hypothetical Chuck Norris player in D division who wins all his matches 6-0. I would really want to promote the #2 in that group, even though he lost his match against Chuck Norris 0-6.


Btw: 90 people will fit in A+B+C+D. If we end up with something between 92 and 110, I intend to keep ABC the way it is (6 player groups), but play with 5-player groups in D division. 91 and  I'd just cheat by creating a single 7-player group, hoping nobody notices. I'm not closing signups at 90 people - "everybody is welcome" is very important to me.

[soulnet]

...

Why do considerations of morality have to be rooted in religion?

I was joking. However, morality is usually tied to religion for religious people. And I have never heard the terms "morally superior" in a literal way from a non-religious. My usual non-religious friend believe morality to be subjective.

Anyway, to the matter at hand, I don't think it is worth it to keep fighting using Isotropish rankings, but I think even randomness could be more accurate, because Iso-ranking is biased to "people who take all their pro games seriously" and I think the league should be biased to "people who take league games seriously". I  will not do the former because I am not willing to play seriously all the time, but I will certainly do the latter (same as GokoDom or any other tournament). And it seems to me that there is a significant number of players that play this way.

[-Stef-]

And it seems to me that there is a significant number of players that play this way.

Yes I play that way. I suppose almost everyone plays that way.

I'll try to play better in league games, I tried to play better during Gokodom.
Yet I made some terrible mistakes during Gokodom for which I'd love to claim "I'd never make them under normal circumstances".
I also want to claim I'd never make the silly mistakes I make in everyday dominion during tournaments.
Both claims, however, are false.

In the end my level of play during tournaments is probably a little bit better, but not as much as I'd like to believe.
There is a strong correlation between the two regardless. I'm sorry but I can't take "it's about as good as random" very seriously.

[silverspawn]

Anyway, to the matter at hand, I don't think it is worth it to keep fighting using Isotropish rankings, but I think even randomness could be more accurate, because Iso-ranking is biased to "people who take all their pro games seriously" and I think the league should be biased to "people who take league games seriously".

this is slightly off topic, but isn't the intention of having pro games precisely to take them seriously? I get that you don't to play serious all the time, but why don't you play casual/unranked whenever you just want to have fun?

[Holger]

morally superior

Wow, I would have never suspected you were a religious guy.

Anyhow, I think total league points is still better than ranking (Isotropish or any other), even though the counterpoints against it are valid. Rankings are not really accurate (I have already argued this). We are already using pseudorandom metrics (see discussion above regarding tiebreakers) and this seems like a decent one (if the groups are random, opponent's should be reasonably similar in strength, and there is luck on getting an easy group, but there is luck in that to get 1st place too, and there is luck in Dominion in general). Competitions with a lot more at stake use this kind of comparison (like European World Cup qualifiers if I am not mistaken), so, why shouldn't we?

Not only the qualifiers; the next European Football Cup and the World Cups 1986-1994 with 24 participants use(d) this method.
It seems best to me as well to use an "internal" comparison like this. It's true that the 2nd-placed players had different opponents, but the same is true for the 1st-placed player, yet they promote without having their ranking checked. And I suppose Stef will try to create more or less equally strong groups within a division.

[AdamH]

Anyway, to the matter at hand, I don't think it is worth it to keep fighting using Isotropish rankings, but I think even randomness could be more accurate, because Iso-ranking is biased to "people who take all their pro games seriously" and I think the league should be biased to "people who take league games seriously".

this is slightly off topic, but isn't the intention of having pro games precisely to take them seriously? I get that you don't to play serious all the time, but why don't you play casual/unranked whenever you just want to have fun?

FWIW, I play pro games so that I can have full-random kingdoms that people don't get to see before the game starts. I play tournament matches to focus entirely on winning the game.

[pst]

FWIW, I play pro games so that I can have full-random kingdoms that people don't get to see before the game starts. I play tournament matches to focus entirely on winning the game.

Exactly. Isn't the designation "pro" just because Goko originally thought that the way to play the game is with pre-made boards (with the emphasis on features like saving boards)?  It still has the unfortunate effect that many casual players don't play the normal game because they think that "oh, that's called 'pro', so it's obviously not for me!".

[AdamH]

I mean, some people see Pro mode that way and I don't think they're wrong, but I was mostly pointing out that a lot of people view these things differently and I think nobody is wrong.

Personally I think Isotropish measuring Pro mode the way it does is the best thing out there, but that doesn't mean it's entirely accurate. And all of this doesn't mean we shouldn't use it, because I think we should. If people have any issues at all with their ranking they'll have the opportunity to prove the leaderboard wrong as I plan to do

It's a personal goal for me to make it into the A league, and I'm starting in C. I think I can get there if I try hard enough.

[Awaclus]

I think nobody is wrong.

I think you're wrong.

What now?

[AdamH]

I think nobody is wrong.

I think you're wrong.

What now?

You're not wrong either.

Maybe you aren't right. I dunno. Can't we all just hug it out?

[shraeye]

/in for hugs

[EgorK]

I think with up to 98 we would be better of with 7 player groups. It doesn't really matter wether there is one such group or all

[Monsieur X]

Stef i saw you created a topic for results, and you think to create an other one to put current leaderboards of all groups
I have an idea. Why not asking one player in each group to post a message on top of his group's topic dicussion with results and current leaderboard?
Other group players would verify it and it would take less time for you?

[2.71828.....]

I never understood the third tie-break until I actually looked and figured out what it was to see why Stef beat me this season.  I also found it very vaguely worded in the rules, which is probably why I never really understood it.  I think that it would be beneficial to clarify this in the rules. 

As is, the rules state:

II. Ordering
1. The number of games you won; a tie counts as half a win.
    --> We do not count matchpoints
2. Number of games won counting only games between the players that are now tied
3. for each game you won, add the total score of that opponent
4. Coin flip

Since results are posted under "standings" I propose that this bullet is changed to something like this:

II.  Standings
1. Points are awarded for each game won.  1 point for a win; 0.5 points for a tie.
    a. Points may be awarded for incomplete games based on the circumstances of that game.  The specific ruling for this can be found here.
    b. No points are awarded for winning a match. 
2. At the end of the season, if players are tied, the following tiebreaks will be used
    a. Number of games won counting only games between the players that are now tied
    b. Season Point Value^TM
        1. For each opponent, multiply the number of points earned against that opponent by that opponent's total points.  Your season point value is the sum of your score against each player.
    c. Coin flip

*I just randomly created the term "Season Point Value" because it seemed to fit best with what I understand the third tie-break to be.  It can obviously be changed.  (if you even adopt this proposed clarification to the rules.)

[assemble_me]

I'd rather call the 2nd tie breaker something like "weighted opponent's score" (because that's actually what it is ).
So yeah, I think that term is rather self-explaining.

[mpsprs]

I think that it would be beneficial to clarify this in the rules. 

Thanks for this.  As it happens, we (moderators of the league) are in the process of trying to clarify the rules, especially as it pertains to helping new players.  So if you see anything else that is confusing, we'd love to hear about it (especially if you also have suggestions for a more clear wording!).

[mith]

Another thing that might be worth revisiting is tiebreak procedures for more than two players. It's common (though probably not universal) in sports leagues to determine the next team in the standings and then reset to the beginning of the tiebreak procedure with the remaining teams. For example, in the hypothetical Season 8 standings:

If he wins exactly 4 games ...

A (100%)

rank	name	average	points	#played	2nd	3rd	iso	season9?
1.	Mic Qsenoch	4.0	20	5	0.0	279.5	52.6	Yes
2.	SheCantSayNo	2.9	14.5	5	0.0	219.5	47.4	?
3.	2.71828.....	2.8	14	5	7.0	195.5	38.9	?
4.	Stef	2.8	14	5	6.0	209.5	54.4	?
5.	dudeabides	2.8	14	5	5.0	213.0	34.2	?
6.	Tao Chen	2.7	13.5	5	0.0	202.5	41.8	?

Mic Qsenoch - SheCantSayNo: 3 - 3
Mic Qsenoch - 2.71828.....: 6 - 0
Mic Qsenoch - Stef: 4 - 2
Mic Qsenoch - dudeabides: 3 - 3
Mic Qsenoch - Tao Chen: 4 - 2
SheCantSayNo - 2.71828.....: 3 - 3
SheCantSayNo - Stef: 1.5 - 4.5
SheCantSayNo - dudeabides: 4 - 2
SheCantSayNo - Tao Chen: 3 - 3
2.71828..... - Stef: 3 - 3
2.71828..... - dudeabides: 4 - 2
2.71828..... - Tao Chen: 4 - 2
Stef - dudeabides: 3 - 3
Stef - Tao Chen: 1.5 - 4.5
dudeabides - Tao Chen: 4 - 2

... then dudeabides and Tao Chen demote.

e wins the first tiebreaker over both Stef and dudeabides. As the rules are currently stated (and as illustrated in the above table), Stef also wins the first tiebreaker over dudeabides, because the results vs. e are still included. However, e is no longer a part of the tie, so arguably his results against the two should not have any more bearing on the result than their matches with the other players. If we restart the tiebreak considering just Stef and dudeabides, they were 3-3 head-to-head, so we go to the second tiebreaker, which dudeabides wins (by virtue of the extra win against Mic Qsenoch). This seems to be more in the spirit of the tiebreaker rules, though admittedly it is a bit trickier to process.

[-Stef-]

e wins the first tiebreaker over both Stef and dudeabides. As the rules are currently stated (and as illustrated in the above table), Stef also wins the first tiebreaker over dudeabides, because the results vs. e are still included. However, e is no longer a part of the tie, so arguably his results against the two should not have any more bearing on the result than their matches with the other players. If we restart the tiebreak considering just Stef and dudeabides, they were 3-3 head-to-head, so we go to the second tiebreaker, which dudeabides wins (by virtue of the extra win against Mic Qsenoch). This seems to be more in the spirit of the tiebreaker rules, though admittedly it is a bit trickier to process.

I'm actually kind of happy with this rule. It's relatively simple given that it covers all cases (even the freaky 5-way ties).
And on top of that it hasn't been relevant even once yet (only in hypothetical standings), so I'm not very inclined to change this rule.

[EgorK]

I think first tb should be recalculated only if after first iteration there is still tie. Otherwise why not conclude in the example provided that dudeabides is clearly 5th and then Stef wins over e