Thursday, May 9, 2013

Request for comment: The size of the global highscore list

Jochen and I have been talking about the size of the global highscore list for ADOM Deluxe. How many entries should we store? Globally and per player? Should there just be a global limit (e.g. the global highscore contains a maximum of 100.000 entries)? Or should we "just" define a limit per player (e.g. not more than 20 globally stored highscore entries per player)?

Currently I lean towards one global limit (option 1) instead of the "per player" limit as this seems to be the more common way of doing highscores, but let us know what you think...

31 comments:

  1. Why not both? Store, say, 10 per-player high scores locally, so players who are aren't all that good can still feel accomplished, but store 100,000 (or whatever) global high scores on the server, so if you actually are any good at the game, you can show off!

    ReplyDelete
  2. I think a global limit makes more sense, personally.

    ReplyDelete
  3. 1 limit per player, per class :-) Your best score for each class is listed. Didn't best yourself? Try again.

    ReplyDelete
    Replies
    1. On the server - absolutely, I think this is it. A local offline highscore list with 10 or 20 slots seems like a good counterpoint for folks that just want to keep trying with a favourite class.

      Delete
  4. A global high score limit will quickly be dominated by the best people and become irrelevant for 99% of players. New players especially need a local high score to help them see their ongoing progress.

    To be honest though, high scores are pretty damned irrelevant in ADOM. The best way to get a high score is through scummy and boring methods. Some of the coolest characters score lowly, and Ultras in particular score worse than regular victories at the moment. In a game with an infinite dungeon I don't see the point of a score system after victory - it's just inviting the most patient to claim the top spots.

    The community-run ADOM servers have had joint high scores for a long while and I can assure you that very few people care about them.

    ReplyDelete
    Replies
    1. And just to clarify, here's the ADOM server high score that has been running for about 6 months:

      http://ancardia.us.to/adom_hiscore/hiscore_v111.txt

      You have to scroll down to the 80s to find the first non-winners. Imagine that running for longer with way more users (since only a small percentage of fans use the server).

      Delete
    2. This comment has been removed by the author.

      Delete
    3. I agree that a highscore list alone is not very useful. But we have some plans for it that hopefully enhances its value.

      a) we will allow filtering of the list, so you can compare scores of your favorite race/class combination with other players
      b) we could add some nice charts, e.g. which race/class combination is played the most, etc.
      c) we could sort the list by calculating weighted/average scores (like Vind mentioned below)

      There is probably much more we can do with this data, so I think it would be a loss to ignore it completely.

      Delete
    4. I do like the idea of charts and things. ToME4 does them pretty nicely on its site: http://www.te4.org/game-statistics http://www.te4.org/realtime-statistics-levels It tracks top killers of players and things like that.

      Anyway, I hope the local hiscore table gets retained as a way to track one's own progress regardless of what others have done.

      Delete
  5. What if every registered player could list up to some number (10 or 20) scores? This way, no one would be excluded and everyone could post what they wanted, but they couldn't spam the list.

    It could also be "both", where the top 10,000 or 100,000 could have as many entries per character as were the highest, but then people could also be recorded with up to 10 or 20 results, regardless of score.

    ReplyDelete
  6. Like people have already said, score is only a secondary motivation for most players. Beginners want to win. Winners usually want to repeat it with different race/class combos and experts go for restricted tactics games. Come to think of it, scoring high is a restricted tactic game that has become less interesting than the ones invented by the community. Maybe scoring should be expanded to multiple desperate metrics - change the score list to a playthrough list with limited number of records per player. Imagine if you could sort the list on number of turns, minimum number of kills, etc. I am sure the community will come up with new metrics to attempt maximising.

    ReplyDelete
    Replies
    1. Also a very interesting idea Jochen and I will have to consider :-)

      Delete
  7. I meant separate metrics not 'desperate'.

    ReplyDelete
    Replies
    1. Haha, I wonder what the 'Desperate' metrics would be!

      Delete
  8. A high score list that can be filtered and sorted by race and class would be a good thing.
    Also, displaying best weekly and monthly scores from the list - all-time scores can get quite high quite quick, so the "top 100" may get fixed after a while, and people rarely check scores below the top ones.

    I've just had another idea however, and I think it may be a good one. How about automatically calculating race/class difficulty ratios? You could gather the "race/class: score" information from each playthrough and calculate average score for each race/class combo. This information could be visible somewhere along with the highscores. How many times certain combinations have been played and what is the average score for that combination. What's more, you could set the score multiplayer: total avarage score divided by race/class average score. And then display weighted scores (scores multipled by those ratios) on the global highscore list with the option to sort the list using those values.
    This way there will be an encouragment to play unpopular race/class combinations, and a way to eaisly check which combos are easy and which ones are hard - an useful information for both new players not knowing the game and veterans looking for a challange.

    ReplyDelete
  9. How about the global 10 000 and limit players with being only able to have 10 best scores on the global hi-score list. And with some filters one could find scores with all race/class combinations. And add filters for challenges etc.

    ReplyDelete
  10. Also, for sake of tracking clallenge (ironman etc.) game highscores it would make sense to store the achievements the played game was entitled to. With filtering and some easy to link page of filtered results this would help the community to keep track of completed and/omost succesfull responds to a challenge.

    One more thing to consider. As sure as sun rises from the east, there will be cheated submissions to global high score lis. Any thoughts onhow to fight this?

    ReplyDelete
    Replies
    1. Actually the cheating topic concerns us quite a bit. So far the (to us but probably not to the community) best idea seemed to be to only allow write access to the global lists by registered players. Thus cheaters could be easily sanctioned...

      Delete
    2. If a game (RNG) is deterministic, you could record all player moves and verify that it is a winning game on the server side. Might be too expensive though and still permit cheating from save-scumming, etc.

      Delete
  11. I vote for limiting only per player or per player+class, not globally. That way, every player will be able to access the high scores (even if it's on position 361,273) and compare himself to others, seeing their percentile, etc. With a global limit, players not "good enough" to make the top 100,000 wouldn't be able to do that.

    Also, if ADOM gets accepted on Steam, you should prepare for providing friend highscore lists. Some Steam games have that feature and it's great, I know in my group of friends it makes us play more :) If you're not familiar with this, it's just that you can see a highscore list where you only see your Steam friends.

    A cool feature I have also seen in some games (e.g. Trackmania) is highscore lists per country and per region. Being among the best in the world was only for a selected few, but being among the best in your province or your city was easier an very fun to try.

    Of course, both friend highscores and regional highscores intrinsically require not having a global limit (only a per-user limit)... under that assumption, you can implement those highscores and many others as filters; a globally-limited highscore is much less flexible. So I think it's very important not to have a global limit.

    ReplyDelete
  12. The Angband ladder ( http://angband.oook.cz/ladder-browse.php?v=Sil ) seems like a useful reference. You upload a full character dump (a new dump for the same character replaces the old dump), and the dump is processed to extract race, class, winner status, score, etc.

    Ladder entries can be commented on, so that people can discuss strategies or ask for advice, and the ladder can be filtered by variant, race, class, and player (so everyone who submits has their own individual ladder), and sorted by various criteria other than score. Sorting winners by turncount is a good way to get scummers and grinders off the top of the pile (unfortunately not implemented in the Angband ladder).

    Limiting the ladder to 10 or 20 entries per player would compromise its usefulness as an archive of achievements and advice given - you could achieve the desired effect by adding an option to filter out all but the best 1, 5, or 10 entries from each player.

    The Angband ladder has a total of around 10,000 entries, mostly because players have to submit dumps via a web interface (and many players don't do that if a character dies early). With an in-game submission system you'd be more likely to reach 100,000.

    ReplyDelete
  13. as said by some people above, different statistics and ability to sort by them would greatly appreciated. as the current score system is easily gambled it is not not very interesting to strive for when competing on a global list. but to use it at least as some source of all kinds of interesting stats would be very enjoyable for me.

    ReplyDelete
  14. I think you should have two high score lists. One list that takes the single highest score of each player, and one that displays all scores for everyone.

    ReplyDelete
  15. Perfect highscore system imo already exists in DCSS (http://crawl.akrasiac.org/scoring/overview.html )

    You have best race/class scores,best real time/turn scores,best total scores,recent wins etc. You have easy access to any player games and achievements. If the adom future highscore will be any close to this im fine with that.

    ReplyDelete
  16. @flashdoom thanks, though I'm not sure if we need this kind of complexity in the first iteration :)

    I've just added the "game log" to the highscore, so you can access it from the ADOM server.

    ReplyDelete
    Replies
    1. Ye i forgot this topic is about the size,not the design. Anyway first step sounds promising, keep it up :)

      Delete
  17. I'd propose to store all the scores and later just slice through that data using different filters:
    - 1 per user
    - 1 per user per class
    - best this week/month/year/all time
    As a developer you'll get valuable source of information:
    - which classes people play
    - which classes they don't
    - which classes are resulting in better score and why, etc.
    For steam these stats can result in achievments

    ReplyDelete
  18. defiantly a local and global list. divided/sortable by race/class and week/month/year like Alex stated.

    ReplyDelete
  19. I'd say go for 1 best score per player in global list, and as many as HDD can fit local high scores.

    ReplyDelete