Five Tribes Five Tribes

Forums

Search
Forums » Memoir '44 Online - English » Explanation of the ELO ranking
Show: Today's Posts 
  
AuthorTopic
Yann
-= Crew =-
KS Backer - Skeleton

User Pages
Posts: 2355
Registered:
October 2002
Explanation of the ELO ranking Wed, 16 June 2010 16:16
Recently, it was asked how the ranking would work in Memoir '44 Online, especially in regards to the fact that scenarios are often unbalanced. Who would be willing to play the Allies at Omaha beach, with only 22% of chance to win?...

We came up with a simple solution to this issue.

As usual, the ranking system is based on the classic ELO system that we have been using for years and which works nicely.

If you are not familiar with the ELO system, this is the system used in international chess ranking.
Basically, ELO computes an amount of points that depends on the previous ranking score of each player, and on who won the game. The winner's score in increased by this amount, and the loser's score is decreased by the same amount.
If the winner was already much better ranked than the loser, then he wins only a few points, because the outcome of the game is considered normal. But if he loses, his score is reduced by much more points, as the system is trying to move the winner up the ladder.
In an ideal world, the system converges and stabilizes if everyone plays at his strength and wins (or loses) as expected. Of course, this is never completely true - which makes the whole thing interesting, but let's no digress here Smile

To keep the explanation simple, let's consider two players of equal strength and ranking - so we have a fixed and identical amount of points at stake. How do we handle the Omaha Beach case?

It's very simple: the amount of points given to the winner is compensated by his probability of winning.

Example: say we have 32 points at stake. If the Axis win at Omaha, they will get 32*22% = 7 pts. If the Allies win, they get 32*78% = 25 pts (almost 4 times more!).

In other words, if you have less chance to win, it's compensated by the potential reward.

The ice on the cake: we also take into account the performance of each player, compared to the average score for each side. This is the purpose of the "thumb up" and "thumb down" icons in your game history, and why we show the average values in the AAR screen.
On Omaha Beach, the average score for the Germans is 5.5, and 3.1 for the Allies. The idea is that you get 10% of bonus (or malus) for each medal above (or below) the average. So say you lose 5-6 playing the Allies: you opponent get a 6-5.5=0.5=5% bonus and you get a 5-3.1=1.9=19% bonus, which makes a 14% difference in your favor. As a result, the amount of points he will earn will be reduced by 14%. He won, but not well enough. Twisted Evil

It does not change dramatically the result in the end, but it is still a good incentive to keep playing even if you are in a hopeless situation, and try to sell your life as dearly as possible! It should help resolving the issue of impolite players who quit games that they think they won't win. It also keeps an interest for both parties up to the very end of the game.

Conclusion: if you really want to make lots of points, play against someone much better ranked than you, on the Allies side at Omaha Beach, and win 6-0! Piece of cake Smile

Yann

Note: keep in mind that the ELO ranking is something completely different from the Officer promotions system - which I promise I will explain in a future memo.

[Updated on: Wed, 16 June 2010 16:19]

      
Brummbar44
DoW Content Provider
Artillery Specialist

User Pages
Posts: 1129
Registered:
June 2004
Re:Explanation of the ELO ranking Wed, 16 June 2010 16:49
Thanks Yann.

I'm not sure if it's working right then. LOL!

If I win at Mont Mouchet as the Axis (who win 52%) by a score of 4-1 why do I get a thumbs down? Would I have to win 4-0 to get a thumbs up?

Another example, I won at St. Vith as the Allies (who win only 47%) by a score of 6-2 and also get a thumbs down?

      
Yann
-= Crew =-
KS Backer - Skeleton

User Pages
Posts: 2355
Registered:
October 2002
Re:Explanation of the ELO ranking Wed, 16 June 2010 16:53
Brummbar44 écrit le Wed, 16 June 2010 16:49

Thanks Yann.

I'm not sure if it's working right then. LOL!

If I win at Mont Mouchet as the Axis (who win 52%) by a score of 4-1 why do I get a thumbs down? Would I have to win 4-0 to get a thumbs up?

Another example, I won at St. Vith as the Allies (who win only 47%) by a score of 6-2 and also get a thumbs down?


Sounds like a bug indeed! Crying or Very Sad
Please file a bug report, with all the details about the two games, so that we have take a look.

Thanks!
Yann
      
Brummbar44
DoW Content Provider
Artillery Specialist

User Pages
Posts: 1129
Registered:
June 2004
Re:Explanation of the ELO ranking Wed, 16 June 2010 17:28
Ok, thanks Yann. I won't have the logs, but I'll pass on what I've noted in that regard.

      
yangtze
DoW Content Provider
Major

User Pages
Posts: 1842
Registered:
July 2005
Re:Explanation of the ELO ranking Wed, 16 June 2010 19:33
Ok, BUT...

If I lose 6-5 at Omaha to an equally ranked player I will still lose points, albeit a small number of points, but I'll never gain them. To provide a real incentive to play Omaha as the Allies I'd want to score positive points, albeit a small number of points, if I scored above average medals.
      
Yann
-= Crew =-
KS Backer - Skeleton

User Pages
Posts: 2355
Registered:
October 2002
Re:Explanation of the ELO ranking Thu, 17 June 2010 07:50
yangtze écrit le Wed, 16 June 2010 19:33

Ok, BUT...

If I lose 6-5 at Omaha to an equally ranked player I will still lose points, albeit a small number of points, but I'll never gain them. To provide a real incentive to play Omaha as the Allies I'd want to score positive points, albeit a small number of points, if I scored above average medals.
No this would not work because you still want the winner to get some points (even if it's a very very small value). And the fundamental principle of ELO is that it is a zero-energy system: if one player gets X points, the other gets -X points. So you cannot have both players making points.

The first system we designed was by entirely based on performances against the average values (king of what you are suggesting). This was resulting in situation where someone winning the battle might lose points - which is definitively wrong and against the spirit of the game. So we changed to the system I described, which is more elegant in my opinion.

Yann
      
LotusArdent
Senior Member
Commandant

User Pages
Posts: 598
Registered:
December 2003
Re:Explanation of the ELO ranking Thu, 17 June 2010 10:22
Yann écrit le Wed, 16 June 2010 16:16


It's very simple: the amount of points given to the winner is compensated by his probability of winning.


Does it mean that the "probality of win" is written in a holy stone ( Traduction osé de ma part de Gravé dans le marbre Very Happy ), or it would be reavaluate each future result of online battles?

Jerome
      
Yann
-= Crew =-
KS Backer - Skeleton

User Pages
Posts: 2355
Registered:
October 2002
Re:Explanation of the ELO ranking Thu, 17 June 2010 13:51
LotusArdent écrit le Thu, 17 June 2010 10:22

Yann écrit le Wed, 16 June 2010 16:16


It's very simple: the amount of points given to the winner is compensated by his probability of winning.


Does it mean that the "probality of win" is written in a holy stone ( Traduction osé de ma part de Gravé dans le marbre Very Happy ), or it would be reavaluate each future result of online battles?

Jerome
We update the stats for a scenario after each game, of course. Therefore, probabilities evolve (albeit very slowly).

Yann

PS: you say "cut in stone" in English Wink
      
bdgza
Senior Member
Captain

User Pages
Posts: 243
Registered:
February 2010
Re:Explanation of the ELO ranking Thu, 17 June 2010 14:26
Yann wrote on Thu, 17 June 2010 13:51

PS: you say "cut in stone" in English Wink

Actually "set in stone", or "carved in stone".

      
henryblake
Junior Member
Major

User Pages
Posts: 14
Registered:
June 2005
Re:Explanation of the ELO ranking Thu, 01 July 2010 11:16
Yann wrote on Thu, 17 June 2010 01:50



The first system we designed was by entirely based on performances against the average values (king of what you are suggesting). This was resulting in situation where someone winning the battle might lose points - which is definitively wrong and against the spirit of the game. So we changed to the system I described, which is more elegant in my opinion.

Yann


The ELO system used online in some ways seems to me contrary to the design of the game for several reasons (though I will of course have to experience it in action to make an informed judgement).

For me an important principle of this game is that victory is only judged after you have played both sides of a scenario. The ELO system seems like it would work well if this continued to be the case online, but it sounds like the ELO will be implemented combined with an online system that allows and encourages single-round pickup games.

I have over 1200 games (not rounds) recorded between me and my playing partner--we always play both sides each game. If we didn't play both sides of a scenario, I would definitely want my score to advance if I lost well on Omaha as Allies, 6-5 (no question I would demand bragging rights, despite my narrow loss). I enjoy trying to win low probability scenarios, and if I can't win I try hard to lose them better than my opponent. The latter case doesn't seem to be adequately represented in this system and it may make it difficult to find willing opposition for "unbalanced" scenarios (of course with the two round system, there is no such thing as an unbalanced scenario).

Another question I have with the ELO is that if it was designed for chess rankings, is it really appropriate for a game with the variability of dice and cards? Is backgammon ranked similarly?

Reading your post, it is interesting for me to see how closely the historical average of Allies in Omaha tracks our own. The top five most frequently played scenarios for us, along with our allied percentages and the (Battle Report allied%) are:

Juno 77% (61%)
Omaha 22% (22%)
Ardennes East 65% (65%)
Carentan Causeway 83% (79%)
Arnhem Bridge 77% (84%)

I think it is interesting that the two that diverge the most, Juno and Arnhem, probably were much closer to the published percentages when we first started playing these scenarios. With continued experience we have modified our play in these substantially and the balances have shifted. I expect our perceptions of all of our well-worn scenarios to change dramatically once we are able to play adjudicated games with random players online--definitely looking forward to it.
      
    
Previous Topic:release date ?
Next Topic:Next beta - nice work!
Goto Forum: