Unofficial statistics site

Discussion in 'Guilds' started by Farbs, Jan 2, 2014.

  1. Farbs

    Farbs Blue Manchu Staff Member

    Whoa. This was down for over a month and nobody noticed! In that time a new principal took over Braves of Glory.
    The scraper caught up today, so that's official now. BoG has a new leader.
     
  2. ParodyKnaveBob

    ParodyKnaveBob Thaumaturge

    I know I usually catch and report these (since I still like the stats even without the pizza), but it's hard to notice when I can't play because my laptop's down. $:^ , Thanks for staying ontop it, Farbs!
     
    Sir Veza and fatcat__25 like this.
  3. Lol someone took advantage of the fact it appeared to the scraper that the principal had not played for 10 days or more. Someone cleverly pulled out the little-used Coup Ceremony. I would be mad if I was in the BoG, but then it was a pretty clever exploit so I am laughing about it.
     
    Sasoo8, Sir Veza and fatcat__25 like this.
  4. FcxHiro

    FcxHiro War Monkey

    @Farbs, Hi there, if u take a look at my log, u can see it's impossible for me going unstable.
    and its not 1 Month, as u can see the changes delayed, active per 13 Mar-2020, It's kind of silly if you tell like that, meta not update that make it happen.
    so you can help the issue or not, sasso and Gargamael please response regarding this.

    Thank you,
     
    gurel2, Sir Veza and ParodyKnaveBob like this.
  5. ParodyKnaveBob

    ParodyKnaveBob Thaumaturge

    Yeah, big difference between a coup because a principal is inactive and an exploit in the META site's logging mechanism.
     
    Sir Veza likes this.
  6. Shanomo

    Shanomo Guild Leader

    Unfortunately, the guild Sucre and I started up some years ago has been the subject of an unplanned coup. Please return leadership of the guild to FcxHiro S.Bussi, I'm really saddened that this has happened, and I hope that you would realise they are just impeding peoples enjoyment of the game. Maybe it's funny to you, but all you're doing is forcing our guild to relocate and waste the effort we spent building up from a small start, this months game points, and putting a black stain on the history of our guild. The braves will be relocating to the new guild Fcx created "Elite of Fun" until such time as you change your mind, sorry to all the braves guild members affected by this.
     
    ParodyKnaveBob and Sir Veza like this.
  7. tq4t

    tq4t Kobold

    @Farbs The Meta site appears to be down again. Just letting you know!
     
  8. Auburn

    Auburn Guild Leader

    @Farbs I think that if we restored the old point awards system, it could encourage people to return to Cardhunter. Could we try it?
     
    Stexe likes this.
  9. Sir Veza

    Sir Veza Farming Deity

    Farbs doesn't work here anymore.
     
  10. Auburn

    Auburn Guild Leader

    Didn't realize that. But still, he might check the forums and decide to change the meta game back for the heck of it.
     
  11. Sir Veza

    Sir Veza Farming Deity

    A happy thought, that.
     
  12. Auburn

    Auburn Guild Leader

    After the meta game was torpedoed, I was told that it was totally fine to use the API to recreate the old meta game or any different meta game desired.
    Unfortunately, I know nothing about API, but if someone does know and is interested, speak up! The Cardhunter API site is here
     
    Stexe and Sir Veza like this.
  13. doublequartz

    doublequartz Lizardman Priest

    For a long time I had basically ignored [META]. I was only recently informed by @Auburn that there was an exodus of players following the change in the guild rating system. I am very much feeling the pain in the form of slow and poorly balanced matchmaking.

    The thread about the change shows a bit of controversy. The way I understand it, there were problems with some players' behavior, which was very real. But there is no easy cure for this. Some moderator activity is and will always be required. Scoring system can only do so much.

    It's ironic that people complained about being "sniped". Now there are so few players it's hard to not know who you'll be matched with.

    The "old" scoring was mostly the same as the classic ELO rating, except it was shared guild-wide. With that system, a player who plays a lot of games could singlehandedly determine his guild's rating, either getting to near the top of the ranking or completely ruining the guild's rating. I think this was not a good characteristic for a system that's supposed to make people cooperate.

    The "new" scoring simply counts wins over all guild members. This has led to a "quantity over quality" situation where guilds' ranking no longer corresponds to skill. This turned out to be even more problematic as ranking high as a guild was no longer challenging or honorable. As far as I know, it did not matter that people could have some pizzas. It was never the point of guilds.

    I have concluded that a better scoring system should have the following characteristics:

    • Having players in a guild should be strictly beneficial. Weak players or players experimenting with weaker builds should not act to bring down a guild's rating.
    • A single player's success should have limited effect on a guild's rating. A guild should be required multiple players to reach a high rating.
    • Large number of weaker players should have diminishing returns on a guild's rating. A guild should be required strong players to reach a high rating.

    With the given rationales in mind, I hereby propose a yet-newer scoring system.

    • The [META] site shall maintain a personal rating for each player. This rating is similar to the in-game rating, with two differences:
      • It will not be transferred between members from the same guild.
      • It will be reset at the start of each season.
    • Define two constants, starting rating and converging speed. Change these values to tune the number of players needed to reach a high guild rating.
    • At any point, a guild's rating shall be calculated as follows:
      • Define a virtual player for the purpose of calculation. Set their rating at starting rating.
      • Sort all guild members by their personal rating, from lowest to highest.
      • For each guild member, increase virtual player's rating as if they beat the guild member converging speed times. Do not decrease guild member's rating.
      • Take the virtual player's rating as the guild rating.
    With enough effort, I might be able to implement something like this myself. Still, I'd much rather have @Farbs do this on his site as that would save us tons of trouble. Thanks to everyone who's staying in the Multiplayer lobby these days. I hope one day the lights will be brighter and so will the playerbase.
     
  14. Auburn

    Auburn Guild Leader

    @doublequartz
    I was told several years ago that it was perfectly okay to use cardhunter's API to make other meta games. I've messed around
    with it, but I am not a coder. I'd like to know what the Knights of Unity's take on the meta game is, but it seems like it would
    be fine to implement your system. I'm interested in seeing what it would be like.
     
  15. doublequartz

    doublequartz Lizardman Priest

    @Auburn
    There are problems with making it myself that I'd like to avoid if possible.
    • Farbs' site is well-known in the Card Hunter community, but If I created a new site it wouldn't be. It'd take considerable effort to get everyone(which would be in the order of hundreds if the old guild players were to return) to migrate to a new site.
    • Farbs' site can be accessed directly from the game by clicking the player's name. It'd require action by Knights of the Unity to make a new site accessible like that.
    • If I created a new site purely based on the Card Hunter API, It would not know who was in which guild on Farbs' site. Everyone would have to do their ceremonies all over again.
      • This is a bigger problem than the first bullet point. To register that someone is in a guild, the guild principal and guild member need to be present at the same time, requiring more coordination. In the case of bigger guilds, the principal will have to go through a lot of ceremonies that will take a lot of time.
      • I could circumvent this problem by scraping(parsing the HTML of) Farbs' site(what an API-ception it would be!), but then I'm not using stable APIs, so my site could break the second Farbs decides to resume the work on his site.
      • Alternately, I guess I could use my admin powers over my site to manually type in every existing guild-player association(and any titles these players may have), but should I decide to do so there are so many players for me to work through.
    • Last but not least, it'd take a lot of time and effort to reimplement everything Farbs' site has in place, unless I can obtain existing code from him. Design in particular will be difficult to get right, even if I manage to clone the functionality.
    I think the next step will be to contact Farbs via email. And if it turns out that he's not going to fix it, I guess I have to try making it instead.
     
    Last edited: Dec 24, 2022
    Stexe likes this.
  16. Stexe

    Stexe #2 in Spring PvP Season

    I think Farbs has moved on from Card Hunter. Maybe contact him and see if there's a way to open source and/or allow for the community to take over the [META] stuff? Not sure but that is definitely your best bet.
     
    Sir Veza likes this.
  17. Auburn

    Auburn Guild Leader

    Good thinking Stexe. I'll see if I can track him down.
     
    Sir Veza likes this.
  18. Stexe

    Stexe #2 in Spring PvP Season

    I mean he follows me on Twitter and Steam. I could ping him the next time I see him online. But he hasn't been on in a while so maybe just emailing him would be best. I forget his email but I'm pretty sure it is on his website. Might want to wait until after the New Year though...
     
    doublequartz and Sir Veza like this.
  19. First of, Happy New Year to you all!

    Second, I'm not an active PvP-Player, I have very little stakes in the Meta game and by no means am I interested in taking over the responsibility for the Meta-site (given that I already maintain the utils). That said, if someone takes over the Meta-site and needs a second opinion, shoot me a dm and I will see what I can do.

    But the reason I'm posting here is that I wanted to provide my two cents on the topic of a rankng scheme. I appreciate that you try to have a discussion first before you force a new rating scheme upon the players. As you also already hinted at, this ideally should happen with the approval of the Knights (are there still pizza-prices to win? If not, the Knights might not even care that much). But I always enjoy coming up with such schemes and weighting up their pros and cons.

    I completely agree with the second and third criterion, I'm just not sure how the first one should play out in practive:
    Yes, having more players is a good thing. Though, I'd argue that it would be beneficial to apply a "diminishing returns"-rule here as well. This would (a) create a much more active competition between guilds (otherwise you'd have a few very large ones and would always have the same season winners over and over) and (b) not penalize smaller/newer guilds too much.
    What I'm not sure about at all, is whether you actually propose that weak players should NEVER be able to lower one guild's rating. That seems wrong to me.

    Let me make a proposal on my own. Note that I have little experience with ELO systems in general and I am aware of some particular flaws in my proposal as well.
    • First, players that were inactive the whole season should be regarded as not being member of this guild for the group's rating. Being a large but inactive guild should not be beneficial for the guild in itself. One might want to discuss whether there should be a minimum number of games (other than 1) for a player to take part in, in order to be factored in.
    • Second, to limit the effects one player's rating can have, what about we use some function on each players' seasonal rating that puts large emphasis on the initial changes (near zero) but diminishing returns on whether a player managed to increase (or decrease) their rating by larger values. Both in the positive and negative scale. The easiest mathematical function would be the qubic root. To keep the values closer to the original ratings, you could also use x^(3/5) (the fifth root of the qubed rating value). That assumes that the personal seasonal rating is just new rating minus old rating, hence that it can just as easily be negative. If that was a false assumption, my bad, not an expert in this seasonal ELO system.
    • Given these ratings for all considered guild members, sort them from highest to lowerst.
    • The highest of these values factors in to 100%, the second highest value to some slightly lower percentage, and so on, until the lowest values (whether they are low positives or even negative) will have a very small effect but will still be considered.
    Some interesting properties:
    • This way having a single very strong or very weak players (large negative score) isn't impacting the rating as much as having more but not as strong/weak players. One very good player could easily be substituted with two less strong players while still resulting in the same guild rating.
    • Having dozens of players is not inherently superior to some guild with only three members. But larger groups have an advantage at getting a higher overall rating.
    • If some players have a phase of losses (for example due to experimenting with the deck) but the guild still has some strong players, the guild score won't be impacted much.
    Some drawbacks that I'm aware of:
    • The guild rating is hardly any ELO like system.
    • The exact calculation isn't as easy to understand as any classical system. That said, a classical ELO system is only considered stable if players don't change much in their strength and if almost everyone already played against everyone else. (But also an argument why I would discourage from using a virtual player to derive the guild's rating. ELO systems afaik can behave a bit funky in such a constructed situation or will just return the highest member score, regardless of any other team members)
    • The effect that one game can have to a guild's rating is hard to predict. As after every game the described algorithm runs anew. A new match doesn't add or subtract some fixed portion, but re-evaluates the score completely. Still, a loss will lower the group score every time and a win will strictly increase it.
    • Negative player ratings are (somewhat) by design not considered as much as wins. That might not be desired. In a 2-player guild where one player improved their score by some amount and the other player lost the exact same amount of points, this will still result in an overall (slightly) positive group score. Is that a good property of the scheme? I can't decide. You could change this by alternating positive and negative scores. That way the larges negative value would be considered almost as much as the highest positive score (not changing the score of a 2-player-guild, but for larger groups).
    • Players that have already maxed their ELO score will not be able to contribute much to the score regardless of the number of wins they have. Unless, thes lose a bunch in one season to win back their rank in the next.
    So, you might as well take some aspects of this idea to create a new scheme but I guess this half-baked one is just as bad as any other. But the current system (as far as I understand it) really has the wrong incentives. Almost every win gives a point to the guild, but only losses to players from different guilds lose you points. What an imbalance.
     
    doublequartz and Sir Veza like this.
  20. doublequartz

    doublequartz Lizardman Priest

    Thanks for the idea. As you have pointed out, designing the guild rating system is quite challenging, and we haven't arrived at a perfect one so far. I'm happy to be able to discuss this and explore more possibilities.

    Players being able to lower their guild's rating is problematic because it would make the guild benefit from throwing these players out. While logical for guilds as competitive teams, it's downright terrible for guilds as social constructs where people belong. I fear that such systems would result in a toxic environment where half the players are not welcomed anywhere. Guilds should be about improving players, not excluding them. Still, I agree balance between bigger guilds and smaller guilds could use some work.

    Tangentially, maybe it shouldn't be allowed to join or leave guilds in the middle of a season. If we delay the effect of these ceremonies to the end of each season, that could save us a lot of trouble with betrayals and stuff.

    Activity requirement makes sense to me. Players should not be allowed to contribute to guilds at their initial rating by not playing the game. However, many players don't have time or aren't inclined to play many enough games for their rating to stabilize, so I think this rule can't be enforced effectively.

    You're getting at something by suggesting a different algorithm to aggregate personal ratings into a guild rating. These algorithms are easily the most important part of any guild rating system, and they deserve a lot of thought put into them.

    I think personal ratings should be independent of previous seasons. It does not make sense for success in one season to harm your rating in future seasons. I understand there is a motivation to keep the metagame constantly changing, but this is ultimately not the rating system's job. The metagame is defined by what build people play, not by fluctuations in people's ratings. I suppose though, it could help newer players get up to speed if there were bonus points for breaking one's all-time personal record.
     
    Sir Veza likes this.

Share This Page