Planet M.U.L.E.

I dont know why I am in the 9° position today.
Prez

I can't say for sure, but I think it's loosely based on % of games won, with a bit of overall score and number of games played thrown in...  Dunno the exact calculation tho sorry :-(

Ok,

may be someone have posted the "rules" of hi-score inside the forum.

Thanks anyway,

currently, the ranks are calculated each 24 hours upon the following data:

games won, ratio, games played, hi-score.

If anyone has a good idea how to make the ranking more fairly balanced or interesting for new players (like a weighted system between games won, ratio and highest score and probably colony score), we like to hear it!

This ^^^^  is meant for High-Score in Tournament?

Because if i take a look at the High-Score of Training games i cant see such an order... ?

Personally I think that using average score (total score / games played) would be a much better way to rank players than using highest score. Actual score numbers are very vulnerable to factors outside a player's control (random events, willingness of other players to cooperate), so taking all of a player's scores into consideration gives a more balanced view than just looking at their highest. Also, it provides an incentive for players to stick with a game through the end even if they're not winning in order to maximize their score, instead of dropping out or playing the spoiler card.

Without getting into the mathematical details of it, I would recommend an ELO type rating system similar to the kind used on games.asobrain.com for their Xplorers game.  It eliminates a lot of possibilities of cheating the ranks.  For example, they don't allow ranking against computer opponents.  Also, they reduce points gained for playing against repeat opponents.  Players from the same IP address are not allowed in ranked games.  Etc. etc.  More details about their ranking system is on their site if you wish to take a look.  Their Xplorers game is a Settlers of Catan clone, if anyone is interested in that.

As a reference, ELO ranking makes a calculation of ranking points based on (some of) the the following parameters:

Diffence in current ranking score
How thick you have beat your opponent
How many times you already played against your opponent

However, IMHO, competition in MULE is really unique to each individual game played.  I'm not sure there is any fair way to calculate who the "best" player is.

ELO is a very good system but it only works for 2 player contests. It doesn't even really make sense in multiplayer games.


Is the formula a secret?

On the contrary, it works quite well for Settlers of Catan, which is a 3-4 player game.  They use it at http://games.asobrain.com and people seem happy with it.  They also do not allow multiple users from the same IP to play ranked games. FYI.

I've played xplorers a lot, but I don't think their ranking system works very well and as such I don't use it. I'd like to play xplorers with some mule people though!

No matter what ranking system you use, there will be people who like it and people who don't.  However, if you don't put certain safeguards in place then ranking has no meaning anyway.  For example, allowing multiple players from the same IP to be in ranked games together, or allowing someone to gain rank by playing the same opponents repeatedly, or allowing players to drop from a game without penalty.

I've played too much to reach (at least temporarily) the no1 ranking.

I can't continue at the same pace, so for me to have any hope to have a shot for no1 in the future, there should be more ranking points available from games with higher-ranked players.

Also, playing with human players should give more ranking points, and there should be some way to tell if for example two players start a game with two A.I.'s and another player drops out immediately when the game starts. Then its effectively a single-player game.

On the other hand, people drop out due to technical problems, so to be considered a four-human game there could be a threshold round, like round 6, that would determine how many human players were participating. Or then a strict limit, all human players need to finish. This may lead to some arguments and one may only want to play with players he knows so that there wouldn't be any last-round disconnections...

All in all, very difficult subject. Maybe the ranking system could be based on some sports like tennis. Then old games would drop from your score, in tennis they stay for a year but here the cycle could be shorter, like a month.

Just some of my thoughts, I know I may wake up already tomorrow and see my ranking dropped again with the current system...

Asobrain also has a mechanism to deal with players who drop out of games. The requirement is that the player participate in at least 75% of the total game in order to be safe from penalty.  I think that is fair.

i've never tried asobrain rankings and just play the free 1 time games of settlers

i think the 1 out of 1 people should at least be taken out of the ranks.  Perhaps have people ranked once they play automatically once they play at least 5 games for example or 10.  This way if someone does happen to win their first 10 tournaments on here, they will be up in the ranks.  I have absolutely no idea how to rank these people fairly though like if one person won 20 out of 100000 games and another won 19 out of 20 games, but right now it would rank the 19 out of 20 lower than the other.

I'm definitely going to be impressed with whatever magic formula you all come up with!  I was trying to think how i'd code it and came up empty-idead

That shouldn't be the case, because Rank is ordered by:
games won, ratio, games played, hi-score

so only for games won=0, the ratio is also 0%, and the number of games apply, so someone who played more would be ranked higher.

if games won is >0, the ratio comes before total games, so someone who played less and still won the same amount will be ranked higher.

So all piete needs to become number one is one more game won than the second place.

We know this is not near perfect score, but we could finetune this much more with the suggested abandon rate adjustement (taking into account round number where the drop out occured), etc.

penalties should be enforced for closing games you start prior to the end as well as abandoning rather than being kicked... that would at least ensure ranking would be appropriately diminished for people doing this to keep their stats up

I don't know what exact algorithm ELO uses, but I like what you're describing about it, Mole. I think 2nd place should receive a slight reward if four human players are playing, but receive a penalty if two human players are playing. i also think fourth place should get a greater penalty than third place. Does the ELO algorithm fit these criteria as well?

Additionally, I think dropout players should automatically record a loss, and replacement players (when they're supported) should not have their rank change - with an obvious exception where you're replacing yourself - then you can receive a fair ranking.

I believe the [modified] ELO system they use at Asobrain would work best for MULE.  It does not allow a ranked game to begin with BOTS in it.  The game must be started with all human players in order to be ranked.  However, if someone drops from the game, they are replaced by a BOT.  Unfortunately, this directly affects the possible increase or decrease in ranks for the rest of the remaining players, unless the BOT only plays 25% of the game (see below).


Asobrain also has a mechanism to deal with this as well.  What they do is require the player to participate at least 75% of the total game time.  In other words, if their connection dies or some legitimate reason they have to leave the game, they can still be ranked as long as they played the majority of the game on their own.  If more than 25% of the game was played by a replacement AI, then they would not get ranked at all (and, unfortunately, the other players ranks could increase/decrease less than they would have if the human player had played at least 75% of their game). 


For more information about Asobrains ranking system, check the FAQs on their web site http://games.asobrain.com.  If you like Settlers of Catan, this is a great place to play it (they call it Xplorers).  If you've never played Settlers, I highly recommend you try it.  It's a really fantastic board game.

I'm quite familiar with Settlers, I think it's OK. I've played about two, three dozen games of it. I prefer my games without dice :) but i got nothin against settlers

Why not keep multiple (ok, two or three) individual rankings?  I am in favor of a simple [score] / [games played].  I also would like to see a ranking on average placement (1st, 2nd, 3rd, 4th versus 1st and not 1st).  A third ranking could be a simple average colony score for all games that a player finished.  The overall ranking could be figured out by weighing all three of these together.

As far as rankings for user accounts from the same IP or games with computer players, I agree there should be some type way to adjust for these.  As to what adjustment exactly, I will leave that up to others.

Is there a database that can be queried to work out ranking scenarios based on existing game data?  It would be a very useful tool for developing the best ranking.  I would even settle for an XML dump of all the game data.  ;)

@rommager:

I just asked myself the same question in the other thread: Sportsmanship
as the fight over first rank may be too harsh for some players  ;)

i think going for first place makes sense... this also prevents collusion

Technically, collusion is not in the game yet.  All players have every opportunity to jump into the fray, so long as they got de cash!   ;D

right now, no one gets credit for the victory! heh :)  perhaps this will be fixed in the next version

The game's won issue has been fixed and all the games you have played and won now show up in your score.

@ machinus


No, I play League of Legends which is a 5v5 game and they use ELO

I am curious... is there a reason why the "Games Played" statistic appears to actually only show "Games Finished"?  It appears one's "Ratio" can be artificially high by simply abandoning any game you aren't likely to win.

For example, I've seen one profile that has Played: 24, Won: 14, Abandoned: 20, for a ratio of 58%.  From those numbers I would think Played would actually be (24+20)= 40, for a ratio of 35%.

Its certainly not a huge issue, but if this is in fact the criteria expected, re-labeling the field to "Games Finished" would make it more clear, I think.

Again not a big deal, just curious.  :)

It used to be the case where abandoned games were not counted in the win/total ratio.  The 1.2.3 update changed this:

    * The store starts with 8 smithore again.
    * A game must have two human players with different IP addresses to be ranked. This means that you can’t have two accounts and play against yourself to gain higher score. If a player abandons the game it will still be ranked if it started with two human players with different IP addresses.
    * The game displays a warning if two players have the same IP address.
    * The speed of trades in the auction increases faster and has a higher maximum speed.
    * Moving behind the buyer or seller line in the auction doesn’t slow down the timer.
    * A bug which sometimes made it impossible to sell to or buy from the store has been fixed.
    * Some issues with the Linux game client have been fixed.

Ahh, I hadn't noticed the update on the main page.  Thank you Govt Mule, I appreciate being so enlightened.  :)

I have been analyzing the data of the games and have been making some interesting observations.  Stay tuned - I plan to post a new topic tonight about what I recommend for rankings.  It will take a little time to write, so I may not get it posted tonight, but it is on the way. :)

I basically will providing the new metrics for ranking using percentiles (so you are ranked relative to all other players), and so what it comes down to is how to weigh the factors.  These factors will be discussed in the new topic for ease of use.

Stay tuned!  ;D

I like the current ranking system.  If it was all based on number of wins, it would encourage dishonest gameplay

What do you consider dishonest?

Hello, Brief note on the use of ELO.  I do believe that it is only 2 player. However, the one game I am familiar with that uses it is Scrabble on Facebook for up to four players.  In this case they consider a four player game a series of two ways versus the winner.  So winner gets three wins and each 'loser' loses one game.  ELO is good because you get credit for beating a really good player.
Cheers,
I.O.W.A.