there's a huge obvious flaw with this one that everyone invariably points out - is having one SC and participating in a draw better than a 16-SC survival?

TheGhostmaker: no, it's your points divided by everyone's points. so no inflation. but then you get ranked lower, not by getting worse, but by not getting better - which is the exact same problem with the current (and inflatable) point system.

we've had millions of suggestions, and debates about this. just leave the stupid point system as it is. (unless we get that Elo thing working)

I would write the Elo-thing for phpdip if I knew how to program it. If there is somebody who does, I would be delighted to work with them.

To attempt to interpret what the equation would be:

Rank Value= (Points owned + points in games in progress)*(3*win count+2*draw count+survival count)

Am I correct in that, if not, please correct me. Else, there definitely is very serious inflation

I think the ELO-esque system that Ghost suggested not long ago is definitely in order.

How many times are we going to have this suggestion?

I think we should have points AND Elo, so that we have a good ranking, but points to keep n00bs out of our games.

of course. that was the original ELO suggestion.

You can use the Elo system to "keeps n00bs" out your game. When you set up a game you could specify a minimum rating required to join. Beginners could also specify a maximum rating if they truly wanted to have a game of beginners. Obviously, that would take a little bit of programming but it is very certainly doable.

Please check the FAQ, these alternative points system ideas always crop up, and there are /always/ big tradeoffs to be made with the current system in all of them. There is no perfect incontrovertible system

Personally I find inflation to be a good thing (as long it stays roughly in check with member numbers), so any system designed to remove it is going to be a step back for some

Adding ELO alongside wouldn't remove inflation of points. It would remove the incorrect representation of ranking.

Rait's points could go to 0 because of multiple large bets. Does that make him a lesser player than me? No.

I don't like the idea of 2 for a draw and only 3 for a win. Thats the problem with hockey overtime, people playing for the tie to get the point.

I'd prefer something like 5 for a win, 2 for a draw and 1 for a survival.

to aoe3rules:

"there's a huge obvious flaw with this one that everyone invariably points out - is having one SC and participating in a draw better than a 16-SC survival?"

this problem is accounted for by incorporating the point values. unless the bets are very different in value, a one SC draw would be less then a 16 SC survival.

to illustrate:

Wager = 100
Pot = Wager * 7 = 700
Winnings = (Player's final SC count/ Total SCs {34}) * Pot

Survival scenario winnings = 16 SC = (16/34) * 700 = 330
Draw scenario winnings = 1 SC = (1/34) * 700 = 20

Rank Value (Survival) = Rank Point * Point Pool = 1 * 330 = 330

Rank Value (Draw) = Rank Point * Point Pool = 2 * 20 = 40

16 SC survival > 1 SC draw

so, unless we compare two very different wagered games, having more SCs is the most significant variable. This reflects the game's objective of holding 18+ SCs.

but that still leaves the unfair fact that there's a different multiplier. maybe every survival could have a different variable... but no, a 1-SC third place would be assigned the same as a 7-SC third. and assigning every single draw possibility would be incredibly overcomplicated.

also, there would be a huge difference from the original game. i'm not one of those "never draw a game or you will die" people, but think - if someone else has 17 centres, you have 12, and another player has 5, it's actually better (to a point) to let that last power keep a few centres than to try to take them (even if that does risk the chance of the first player winning). so, a 16-SC survival is better than a 1-SC draw, but is a 16-SC survival better than a 15-SC draw?

To Kestas:

I disagree that a perfect system is unachieveable. The perfect system may not be obvious, but with some effort and communication, we can discover the perfect system, or at the very least the best system.

hmm... Drawing is really the problem in the equation because it isn't the same as other end game scenarios. All other end game scenarios are precisely that, end game. Draws are the premature cessations of games because in theory, someone would eventually screw up and someone could win. So, the real problem with ranking is figuring out how to address a draw

Aoe3rules:

I see your point with draws and survivals. Setting draws as higher preferred end game results than survival would encourage players to seek draws instead of attempting to win, already considered a problem to some. This could be eliminated by setting the rank values equal to one another. This would encourage each player to win because the value of a draw is the same as a failed attempt at winning. This is just a matter of changing the weighted amounts though

For a genuinely accurate rating system there should be a few things recognised in the algorithm:

1. WTA games and PPSC games must be treated differently, to respect the different objectives.
2. It must reward beating a better player more than beating a weaker player.
3. It must not inflate, as if it does more prolific players get a huge advantage. The same is true for deflation in reverse.
4. Players should not get artificial boosts or rebates
5. All games should be equal. This stops "all or nothing" players, as well as stopping the "meaningless" games we see. For variants, an unrated game option would be good, and would go hand in hand with a proper rating system by not corrupting the results for "pure diplomacy"

Once all of these are done correctly, you have a perfect rating system, unless you consider other things, such as this argument of "gameplay management" where the current system has be argued better because of the way people are encouraged to play by it.

I am afraid that Zarathustra's system doesn't seem to check any of the boxes as far as I can see, so probably isn't that good.


A word such as perfection needs defining. Kestas' is right that a truly perfect rating system is impossible- you don't look at one game and say "He played at such and such a standard in that game". What you do do is say "He beat such and such a player, so must have played better than him" Here we enter the realms of statistics, where nothing is exact, and balances must be made. But we can have the perfect balance. Deciding on the perfect balance was the one very hard part of designing an Elo-system.

I challenge any of you to find an unfixable, inherent flaw in the Elo-rating that Cannot be fixed without removing the fundamentals of the system. I know I can do that with the current system.

you have still not stated what ELO is.

@Zara: just read the wikipedia article on it.

Ah, there was a lot of discussion about this system earlier, and I assumed that everyone saw it (quite wrongly)

Players start with a ranking of 100.
You calculate an Expected result for each player as (chance of victory essentially) as Rating of player/Sum of all players ratings
The actual result in PPSC is number of SCs/34
The actual result in WTA is 1 for a win 1/x for a draw with x players and 0 for a loss (be it survival or defeat)

New Rating= Old rating +V*(Result-Expected Result)

V is a multiplier which relates to the balance I was talking about. You adjust it to adjust the balance between high accurracy with relatively large numbers of games necessary to reach the correct ranking, and lower accuracy but a better indication sooner.

@ aoe, since diplomacy is a seven player game, the explanation I have given is probably more helpful.

TheGhostmaker: that leaves the (easily solvable) huge problem of new players being able to play with pros. you could just make a system where people who want to join have to be approved by the player who created the game.

i just thought i'd point this out since you didn't mention it.

hmm, we should probably remove V. assuming that eventually it will work, anyway.

As far as I can tell TheGhostmaker is correct, this system would lead to huge point inflation. I also don't see how it is any improvement to current system. It would reward quantity of games over quality of games.

In fact this system doesn’t seem very different than the current system. The number of points that a player has is related to the outcome of the games that they were in (win/draw/survive/eliminated). The proposal is to multiple the number of points that a player has by rank points. However the number of rank points that a player has is related, in almost exactly the same way as the number of points the player has, to the outcome of the games that they were in (win/draw/survive/eliminated). Essentially what you’re doing is multiplying two numbers together that are different ways measuring the same the thing. This isn’t giving you any more information than if you just use one of the measures (i.e. rank points or just plain old normal points). The ranking order created by this system would be different than the current system, but my guess is that they would be very very similar.

aoe, if V=1 then you need to play a few hundred games for the rating to be intelligable. I am at the moment thinking a level of 10 might be best. If I get a sign that we might get this implemented in some way, I shall spare my time/processor power to find what level is best.

Also, as you say, it is easily solvable: You make it possible to have games that limit the ratings or the game counts of players who join. The beauty of it is that then you can do this without skewing the rating system.

actually, i was thinking of manually specifying whether a player on the applicants list can join. but that works too.

it would also prevent inflation, and best of all, stop multiaccounters and metagamers!

I do not believe my system is perfect and agree it needs more work. It was not meant to be perfect, but instead to provide a mathematical aspect to the discussion so that we can start to create a better system that can be implemented instead of simply discussing our complaints with the current system. I personally would be fine with just the current point system with the slight addition of a 'point pool' value to show how many points a player is playing with.

No-one can agree on what parts of the current system need to be changed, so I really doubt we're going to see any agreement on an entirely new system.

The fact that every new points system thread has only ever ended up with unresolved debates and disputes makes them really tiresome and repetitive. All these threads are the same, the same issues, the same arguments, the same end result

The only compelling reason I can think of to change the points system is that if it gets intricate and convoluted enough no-one will be able to understand it, so no-one will complain (if we implement ELO I doubt anyone will be complaining about the unfairness of logistic curve derivatives or whatever)

I think this thread may perhaps be even more futile than the thread where Sicarius is trying to convert a pastor, but maybe I'm just in a cynical mood