vyeh

SMAC ladder

191 posts in this topic

Solver wants us to feel home here.

 

He wanted to know if we have a desire for anything, like a SMAC ladder.

 

This is a good opportunity to express your interest.

Share this post


Link to post

Sounds like a cool idea !

Questions:

 

Is there enough man-power to CMN such ? I guess it would involve many PBEMs with few players (2-3) ?

 

Rules of the ladder ? Same map, fair balanced conditions are obvious, but how about further details, like faction choices.

 

Point system of the ladder ? Is it based on win/lose only or matters in how many turns the game was won or any other indicator / measure of advantage? How does it count when a player drops out ?

 

I am actively participating in the CivRev MP leaderboard games on PS3 and there are some things I like and dislike on their system, plus a lot of questions I do not understand about the mechanics of how they are scoring the games.

 

So I am interested in both participating and helping to organize a ladder.

Share this post


Link to post

Great. Why don't you organize it?

 

First step, answer your own questions (or tell us what the options are)!

 

Thanks.

Share this post


Link to post

OK, let's start the ball rolling by a short summary of how game-ladders work in general (be it chess, tennis or whatever):

 

There is a ranked list of players, the ladder. A lower ranked player may challenge a higher ranked player for a match. There maybe some restrictions on the distance of the players to prevent everyone challenging the #1 player all the time. There are usually some rules about accepting challenges so that users must accept at least some of them or if they refuse too many then some of them are considered as lost matches. When a lower ranked player beats a higher ranked one, then he/she advances up on the ladder. The ladder maybe administered purely based on rank ordering in which case the advancement is based on the number of positions between the players: if they are consecutive and the lower one wins, then they switch places, otherwise the lower winner advances half the distance between them. Another option is to base the ranking on some scores/point system which may depend on the actual results of the matches. Some point maybe taken away from the loser and some awarded to the winner. The amount of points may depend on the difference of points between the players (some portion or percentage of that). For example, on the CivRev MP board the victory type and game-score also influences the amount of points awarded: you get few points for ancient horse-rush domination victory, more for late-game tank-rush domination and much more for economic or technology victory. However, if the losing player gives up (quits) before the victory condition is reached, then the winner is considered to have won a domination victory.

 

So, after this general introduction, let's consider the special requirements for a SMAC PBEM ladder:

 

  • Games are several months long, so we need to allow players to participate in multiple games parallel
  • Parallel games puts some twists on the ranking, scoring system, we need to figure out a consistent, logical point system
  • We should consider 3 or 4 player matches and how to apply those in the ladder system
  • We need to choose a balanced map based on the number of human players
  • Faction choices: everybody may choose any faction per game, OR always play with same factions against each other (in 2 player matches) to ensure fair comparison, OR associate different scores for playing different factions, so winning with weaker faction gives more points -- admittedly hard to come up with a scoring system for this because what faction is "better" may depend on personal preference and game style.

 

Of course, the whole thing makes sense only if there are enough interested players to participate. So, whoever is interested to participate, please post in this thread to indicate it. And also add your opinions, point of view about the scoring system and above raised questions decisions to make. And I am also looking for volunteers to act as CMNs for the games.

 

--

::Zsozso::

Share this post


Link to post

We are going to be training CMN's. It won't take much training to set up the type of games you're talking about.

 

Thanks for getting the ball rolling.

Share this post


Link to post

i'm in for the ladder and i can also setup some games.

Share this post


Link to post

It sounds interesting, though doing it via PBEM seems like it would take an ungodly long time to do. In addition, due to the fact that these aren't exactly 1v1 matches, we'd need some kind of system for how we rank the various players in a game, whether its winner-take-all, or some sort of gradiant to account for the end scores for the various players at the end, given it is quite possible for multiple players to be still present on the board at the end, or even for multiple players to share victory.

Share this post


Link to post

I'm currently looking into some software that can assist. We may get something that allows for good management of ladder/tournament systems, with automatic point tracking in an ELO-style system.

Share this post


Link to post

Simple rank-based system (without scores, points) would only work for 2-player games and I think that would not be ideal for SMAC. However, too many players would be too slow, indeed. We also need to come up with a challenge system involving multiple players. So, let me throw in some more concrete ideas for setup (then we can debate and tweak them until it seems good enough to start the games):

 

Simple 2-player rank ladder

  1. Games are between 2 human players and 5 AI
  2. Lower ranked player may challenge a higher ranked one
  3. A challenge must be accepted if the player has less than 3 active games
  4. Standard size map with 2 identical shaped islands for the players, AIs on a larger continent (or more seprate ones)
  5. There is no scoring system, when a game ends, if the winner is higher on the ranked list (at the time of the game completion), then nothing changes. If the winner is lower on the ladder, he advances: if directly below loser, then they switch places, otherwise advance by half the distance between them.

 

3-player ladder with point system

  1. There are 3 human players and 4 AIs in a game
  2. Standard size circular map with 3 identical islands placed along the equator equidistant from each other on both sides (this creates full symmetry even in terms of how far an opponent is for attack), the AI factions are placed on 2 continents one spanning across the north, another one on the south.
  3. A player may challenge a higher ranked player on the ladder -- announcing it on the forum. Then a third player (also lower ranked than the challenged player) has to join the challenge to form a game.
  4. A player loses 1 point if eliminated in a game, but no point is lost if alive when another player reaches some victory condition (other than conquest).
  5. The winner gains at least 2 points, more if a beaten opponent had higher points on the ladder: the difference is added to the 2 points.

 

4-player ladder with scoring system

  1. There are 4 human players and 3 AIs in a game
  2. The game is played on the veterans_04 map: http://www.apolyton.net/forums/showthread.php?p=5191932&highlight=veterans+map#post5191932
  3. A player may challenge a higher ranked player on the ladder -- announcing it on the forum. Then two more players (also lower ranked than the challenged player) have to join the challenge to form a game.
  4. The winner gets a score equal to the turn number the victory was reached (mission year minus 2100).

 

Other scoring ideas:

  • Based on Alpha Centauri score as reported by the game
  • Highest tech level reached in the game
  • Population count

 

Of course, these are only for conversation start, feel free to suggest any tweaks or completely new rules.

Share this post


Link to post
I'm currently looking into some software that can assist. We may get something that allows for good management of ladder/tournament systems, with automatic point tracking in an ELO-style system.

 

Don't claim credit where it isn't due :p

 

Guys, see if this is any good for you:

http://www.weplayciv.com/forums/competitions.php

Share this post


Link to post

I'm not a statistician, and this is only the glimmer of an idea, but it seems a central problem with a rating system for SMAC is that most games have 4 or 5 players, which makes a simple ELO system pretty difficult to execute. So what about the following:

 

Every player starts with a rating of 0, and ratings are zero sum. Each game has 1 point up for grabs, which is given to the winner. The point is subtracted evenly from each losing player. So for instance, in game with players A,B,C and D, if B wins, his rating goes from 0 to 1, and A's, C's and D's go from 0 to -.33. What this fails to account for is differing playing strength among different players opponents, which is a significant negative. On the other hand it would be pretty simple to calculate I think, and it would give a rough idea of how skilled players were relative to others in the same pool.

Share this post


Link to post

I imagine there's a lot of information about the ELO system on the net, and those mathematically inclined ought to look into it.

 

I've alway heard it was just the best for ranking competors at anything. Its biggest problem is a universal one- these things aren't worth much at first, until there are enough games to be statistically signifigant.

 

Justin, it seems to me that your suggestion would leave negative scores to the less-successful players for a long time. I don't think that's the best way to keep them plugging away and getting better at MP...

Share this post


Link to post

Good point. What about starting at 10 instead of 0 then?

Share this post


Link to post

What about 100? I mean, what the heck. Unrated US chess players start at 800 for their base-calculations.

 

I have fallen below that, BTW. Down to 692. I suck.

Share this post


Link to post

In the USCF (US chess federation) rating system there is a lot more ratings dynamism than there would be in a SMAC rating. For instance, in my second chess tournament I lost every game and dropped more than 200 points within 3 days (the combination of a severe migraine and prescription pain killers is not in fact a chess performance enhancer). On the other hand, with the system we are speaking of, you're losing at most 1 point over the course of a two player PBEM game, which would probably take at least a month. So starting at 800 seems a bit much, and if we started at 100, the rating changes would seem so small as to be barely noticeable.

Share this post


Link to post

How about applying a multiplication factor to keep the points integer ? Instead of +1 point to the winner and -0.33 to the losers of a 4-player game, we could give +3 to the winner and -1 to the losers. Then we could start at 30 instead of 10. Come to think of it, why not 42 then ?! ;) That should answer all questions :lol:

Share this post


Link to post

Basically, the winner gets the number of points equal to the number of opponents and the losers all lose one point.

 

So if you play against 1 player, the most you can get is 1 point.

 

If you play against 6 players(!), you can get 6 points.

Share this post


Link to post

Interesting.. in my idea, only one point can be gained in any game regardless of how many opponents there are, but in games with more opponents each individual player has a lower potential loss. For instance, in a two player game the loser loses one point, but in a 7 player game the losers each lose .167 points. So games with many players have less rating risk, but more potential game.

 

In vyeh's proposal, the loss is the constant - one point for each loser. The variable that shifts with the number of players is the rating increase of the winner. There's a lot to be said for that, not the least of which is zsozso's point of integers being preferable. Additionally, while I'm not sure, I think that the higher potential gain might be a greater motivator to join games with more players than reduced potential loss would be. And it does kind of feel like you ought to get a bigger reward for beating 6 humans than one. I like it :).

 

As to where to set the starting rating, I'm not sure. I think 10 might still be reasonable, as it would take 11 consecutive losses to drop below 0. I think that this is not a bad system, though it does have the major drawback of potentially having weak players who play against weaker yet players have much higher ratings than good players who play against better players. As far as I can tell though, PBEM games tend to be pretty mixed, so that may not be a deal breaker. zsozso, what do you think of this type of rating system?

Share this post


Link to post

Also note, the more players the longer the game. So maybe the winner should get more points.

 

Maybe we don't allow negative points. So if someone has 0 points and loses, they remain at 0. It is no longer a zero sum game.

 

And it would be odd for someone to have a lower rating than someone who never played.

Share this post


Link to post

i have a proposal that ca work with any kind of points system:

 

every player has to play with everyone of the factions allowed in the tournament (let's say the original 7).

otherwise, based on the maps used in the tournament, we can have games with 4 Morgans or 4 Universities... and that can kill the pleasure of playing in the long term. also this will encourage players to start multiple games.

 

the player must try to check all 7 factions in his first 7 games in order to be displayed in the ladder. let's call this a 'full score' (seven results from games played with seven different factions)

i know it can take lots of time but we can reduce the number of factions involved (a builder, a momentum and a hybrid faction) or we can use the average of the results to build a ladder and have some standings in the mean time.

 

if the player has to re-play one of the factions that he already played then he will loose the points gained playing the first time that faction (like ATP standings, i think) so the payer will have the 'full score' at any time after he completes it.

 

can i have some feed-back?

Share this post


Link to post

Also we can get Darsnan to design some scenarios that won't take a lot of time.

Share this post


Link to post
Take a look at a similar discussion in CtP:

http://www.weplayciv.com/forums/showthread.php?t=268

Why reinvent the wheel?

 

As far as I see they were tossing around some ideas without conclusion until they got the same message as us from Locutus:

 

Guys, see if this is any good for you:

http://www.weplayciv.com/forums/competitions.php

 

We should definitely check out, test that ladder system, seems to be based on Elo-ranking with a start score of 1000. Since the system is already installed by a site-admin, we should definitely try to use it and tweak if it does not fit our needs.

One fear I have about the Elo-ranking is that SMAC PBEM may be too slow for it, i.e. it may take a very long time to finish enough games for the Elo rating to actually mean something.

 

Locutus, is the source code available for us to tweak in case we want to ? It seems to be php, i.e. script so technically it should be, the question pertains more to the license, legal angle.

Share this post


Link to post

We could consider designing scenarios that could be completed in a lot less time.

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now