Ladder Site Online...

[milwac]

@anoel : Sorry, I did notice your question but forgot to answer, RD = Rating Deviation. It's similar to the deviation/SD you are talking about.

[nagafono]

hi all
as i wrote to eyerouge, i and some guys (some of us're getting or already got math education and so on) from our local forum can research new model of ladder or some kind of modernisation of current version. also any technical ideas are welcome, i also free for realise it, but cann't say that i'm very good php/mysql programmer - my direction is c++ some ides from my side will be later

Regards,
Nikolai aka nagafono

[chains]

Adding code to the ladder is not hard. Several good programmers worked to make sure it was modular, easy to understand, and easy to change. Creating a new ladder from scratch would be a waste of time. The current ladder can be turned into any system and we are happy to add you to the project commit list.

If you want Glicko or Trueskill, all you need to do is code the function in PHP, and drop it into the project. All it requires is a programmer to contribute to the project.

The ladder project was meant only to prove that competitive game play was wanted by the Wesnoth players. The project has shown that competitive play is popular, fun, and civil. In fact, the ladder community has been the most civil gaming circle I've ever been involved with and the replays are so popular that it consumes a huge chunk of the ladder server.

The next phase of evolution for the ladder is to have it built into the Wesnoth interface so that games can be automatically matched. This auto matching solves the problem of aliases and rank manipulation because it prevents players from picking easy matches to inflate their ranks. The Elo or Glicko systems do not work well when players can cherry pick easy games to fluff their rank.

Also, while we are on the subject, I do hope the admin is taking a backup of the database from time to time.

[nagafono]

The next phase of evolution for the ladder is to have it built into the Wesnoth interface so that games can be automatically matched. This auto matching solves the problem of aliases and rank manipulation because it prevents players from picking easy matches to inflate their ranks. The Elo or Glicko systems do not work well when players can cherry pick easy games to fluff their rank.

as i remember there were a lot of suggestions about some kind of automtic mechanism for ladder, but wenoth's developers don't want to integrate it at all (details of this you can find somewhere on forum).

so i cann't find any simple way to deal with it. for making such thing we need an application, which can scan server's games and catch needful games... but there are so many questions without an answer about such kind of solution...

[TheMAze]

I have read 1 post I interested in.
http://forums.wesnoth.org/viewtopic.php ... &start=825
I think there is a problem with multiple-account users. Well not only one. Thanks to Multiple-acc users the honest players dont want to play with me. Because I have a [censored]-low points in ladder. And especially not too comfortable to play ladder games vs "top scammers" that have 1400 points...
They are just hunting for the points, and loves the easy wins.

Example: if any top player who have over 2000points win agaisnt me will get +5 points.
But if the "player" challanges me with one of his multiple-account He have no risk. He can lose without any risk. If he win he will get +15 or +20 points.
Look at those scammers! Some example:

1.hay207
2 games he played in august.

Fortunately I Have found the replay in saves directory and I Uploaded. Just have a look at the statistics and watch 1 replay!! Are the beginners playing like this??
http://ladder.subversiva.org/profile.php?name=hay207

2011-08-09 08:30:23 zara hay207 1686 (11) 1428 (-22) 68 0 64 125 - / 3 L 4Download (4)
2011-08-08 23:50:13 hay207 TheMAze 1450p (34) 1436 (-34) 0 0 0 126 1 / 3 L 2Download (2)
2010-02-18 02:04:52 Aca hay207 1586 (13) 1416p (-26) 111 0 107 0 - / - -1No Details
2010-02-11 13:02:44 Goldilocks hay207 2176 (2) 1442p (-7) 5 0 5 0 - / 5 -1No Details
2010-02-11 00:11:29 hay207 Jelly_the_Mad_Doggy 1449p (48) 1777 (-6) 0 64 0 65 - / - -1No Details
2010-02-10 10:54:35 hay207 Jelly_the_Mad_Doggy 1401p (50) 1789 (-6) 0 61 0 62 - / - L -1No

As you can see "jelly the mad dog" pumped up "hay" with +98points . "Hay" tried to steal points from Goldilocks.
And "Hay" was making +34points on me. There are no replays.. "no evidence"

2. zxa
He easily beats a player that have over 2000points and of course there is no replay...
http://ladder.subversiva.org/profile.php?name=zxa

Ladder is not playable anyway. Till the cheaters have tons of accounts.

Well Hope I can find a list of point-scammers. If anyone knows a "Black-list" please write it into this topic. I think all honest ladder players must be interested in!
Thanks for reading.

[soul_steven]

theMaze I do not want you taking this the wrong way but maybe its just because you are low rank right now that these players can beat you and look like such pros? If someone is 1400 elo it means they lose a majority of there matches, that being said if someone is a really good player they would not continue to use a 1400 elo account at least in my opinion but maybe im wrong? To me the more logical thing would be to use your "good" account to get the for sure points against a low ranker? I see no benefit in using a 1400 account to snipe low rankers...

Furthermore there are many different elements that can control the outcome of a game, RNG matchup ex: that are out of a players control and may be the reason why they win or lose. If a top player like Dauntless was to play a low ranker 100 times im sure he would win 99% of the time but there is always that 1% where things don't go exactly right and cost him the game. I myself have lost games to low rankers before because of rng ex: so don't just assume that because one time a low ranker wins a game it means he is a multiple alias account. I hope I am explaining things properly. That being said is there a issue with the ladder site right now? Heard quite a few say they have had issues reporting games and getting a unkown error?

[Death]

the ladder needs it's own forum. if not here, then elsewhere.

unless you really like a single thread with a hundred different topics spun into one giant run-on sentence, that's cool too.

anyway, if there was a ladder forum, that's also the first place i'd look for ladder-related tournaments, since it never really made sense that all tournaments are run out of the "multiplayer development" forum.

you could even put it under Bfw- miscellaneous, right next to the random-ass forums that no one reads (except for, like two moderators).


i'm just saying.

also, hi.



also, i came by to weigh in on the serious issue of some players using multiple accounts registered in our nationally-recognized standard-setting shining beacon of absolute pure competition otherwise known as the ladder:

who cares.

[soul_steven]

is there a reason the ladder site isnt working? unable to report games (I think everyone is having this issue with how i hear it) Is the site being worked on or something? fill us in! thx

[Gallifax]

@TheMaze:

You never asked me yet.

We can play , I allways played anyone, you can see that if you check my gamhistory regarding wins and losses.

I do have the same issue with some players as you, cause as much as I can loose to anyone I can also beat anyone if I am in the mood for serious games and some people seem to remember that:P

[milwac]

I was waiting to see if somebody would come to the support of ladder2. Guess that's not going to happen.

Anyway since the admins, developers and most of the top tier players are in support of glicko (read : better rating system without flaws) I shall try and explain a bit more about the stuff we will see if the system changed and what all is necessary.

If you want Glicko or Trueskill, all you need to do is code the function in PHP, and drop it into the project. All it requires is a programmer to contribute to the project.

Regarding this, the current database schema ('players') doesn't have any spare columns which we can use as Glicko parameters. We'd need one additional column for now, viz for the RD. If later somebody complained that glicko sucks compared to glicko2 we'd have to add a 'sigma' as well (Not necesaary IMO), and sometime even later if we'd want to have different rankings for a 2v2 ladder, we'd need additional columns. Please let me know who the current developer(s) in charge of the database schema/migration etc (chains, eyerouge?) and we could discuss this over PMs.

Now for some more facts. Trueskill was invented to match up two random players in Microsoft's Xbox Live community so that they could have a challenging and fun game for both. Elo doesn't allow this. Since it is always possible to play a very weak player and gain atleast *some* points, two or three whatever. OTOH Glicko won't add any points to your rating if your RD is low and you win against someone quite weaker than you. (If you ask how weak or strong opponents can you play, it would be roughly like this : for a 1700 player with a RD if 100, she can play anyone between 1500-1900 to see any significant rating change) In this way Glicko handles 'point harvesting' accounts and leads to interesting games at all levels. It is to be remembered that Glicko/Trueskill converges to a players' true strength at present.

Another good thing about Glicko that I've realised is the discouragement of multiple aliases. A multiple alias in Glicko only leads to a bad convergence on RD in any of the alias accounts and leads to a worse ranking of the original account. I could share some results but you can take it from me that it is largely true

I can for sure code up the system in PHP and it can be reviewed by someone else. For the database schema changes we need to have a discussion. However some important thoughts on migration have to be considered.

1. Migrate all past games to Glicko : This is IMO the best thing to do. But won't be popular with the community. Since the original intention of most would be to beat the Elo system and gain some points to move up the ladder (intentionally or otherwise), a lot of rankings will go haywire and a lot of people will end up with negative points, and small RDs. Making it difficult for them to move up, or even get decent games to play.

2. Keep Elo ratings and assign a default RD to everyone initially : IMO this system looks very bad, because there is no chance of calculating rankings from scratch in the future. The rating graphs for everyone would look weird. Also with a mixture of Elo and Glicko rankings existing at the same time, the ladder would seem messed up and wrong. OTOH, players would hardly notice any change initially, but subsequently observe significant changes in the first 10-20 games they play, the rankings rising/falling like crazy. Even though this looks theoretically bad, it's the best thing to do in our situation and is easier as well on the migration front (actually no migration).

Also, if the first scheme is chosen, multiple account holders could have an option to choose one of their accounts and merge all their games to that one (Not very important, but just thought I'd mention)

So please post your thoughts on this, or likes/dislikes if any. If all is well, we could replace Elo with Glicko immediately and make the ladder more competitive and sporty!

Cheers,
milwac

PS: A sample of how glicko graphs look like (http://xntrick.comuf.com/wesnoth/comparison_graphs.html)

[soul_steven]

I will ask again since my post must have been overlooked, Is someone fixing the current ladder or working on it in someway? what is wrong with it why cant we report games? There seems to be some unexpected error we are all getting everytime we try reporting games. Let us know what is going on. Thx

[soul_steven]

Milwac, just for fun I looked at the glicko system for the current top 20 on ladder (this is not top 20 overall since there is inactive players ofc) and take a look at how barbaric some of these rankings become on Glicko

ELO Glicko

1. Cremember 57
2. Dauntless 29
3. Kolbur 233
4. Demogorgon 13
5. The_Black_Sword 21
6. Semler 88
7. plk2 613
8. neki 2380
9. Rigor 59
10. elido 259
11. Shachi 8
12. Oook 160
13. Kira1 207
14. TheFish 3021
15. Dreadnaugh 85
16. Horus 475
17. Undertaker 2498
18. Drakeking 38
19. nelson 925
20. Coriolan 25

I can't even begin to comment on how mad this "new" rating system is.. Good luck trying to get this one approved, give me a break

[milwac]

soul_steven : Haha, yeah indeed this is funny~! I didn't look all that up myself, just compared my Elo results with the ones on the ladder and they matched quite a bit, and considering my implementations ain't flawed this just shows that Elo and Glicko are hugely different in how people are rated. Also I didn't consider factors like increase of RD with inactivity etc, this shall alter the ratings a little bit. Of course we all know that playing against a weak player can sometimes result in a loss due to bad luck etc (Glicko unfortunately punishes such cases severely, it says : did you expect to gain easy points? suits you well, here you go, have -150!) As I mentioned in the previous post, the intention is to match up two random players of equal skill (as chains suggested as well) so that the game is fun and entertaining for both. Of course whether such a system will work for our community with so few players at any given time is questionable.

I am not trying to get anything approved as such! But I would've liked a new ladder with glicko rating system and everyone starting afresh (hard to do, I know), so that all are wary of how the system works and know who to pick fights with As mentioned earlier, someone rated 1700 with RD of 100 is said to be around 1500-1900 in playing strength with a 95% probability. Now in those 5% of cases, if this said player plays someone below 1500 or above 1900 and something unexpected happens, the circumstances that led to this matchup are flawed and not the system. I am happy with the current ladder as well, with the multiple aliases and all (Since so many people are fine with it). And now since this issue is dead anyway, and considering the furore that it created, I don't think people are playing with multiple aliases anymore. Hence any change is unwarranted unless of course absolutely necessary. I am well aware of this

And considering what is better, regardless of the results on our ladder, glicko (and trueskill which borrows a lot of its ideas from glicko) is tried and tested on different platforms and is working fine! Even the graphs that I came up with for the 4 players, seem to converge nicely. Yeah glicko says that Goldilocks is a better player than Dauntless. I cannot answer that question as my wesnoth skills are far from genius.

If Elo is here to stay, so it will!

[soul_steven]

I think what a lot of players are wanting (myself included) is a more in depth ladder that keeps track of other personalized stats. For example (when im drakes I win X amount of games) These things of course are not a neccessity but would be fun to keep track of and to look at. I think the true skill rating is a good system for something big like Xbox but for our small community it would not work out so well. Most the time it is very hard to play games with only good players not because of lack of interest but because of lack of time. This is one of the great things about the current system is you can log on host a ladder game and play instantly.

[milwac]

I agree. I was just interested in the math part of it And am ready to code it up if needed. And regarding your suggestion for improved stats, it was one of the first ideas that I had. But maybe then we need to make updating replays mandatory, or do it automatically (extract from the replay archive?), Quetzalcoatl had some good implementation ideas on this. I guess he is too busy these days to reply. Anyways trying to do this in the current codebase will be challenging with lots of database schema changes, code refactoring etc coming into the picture. Won't happen tomorrow, but is possible.