Page 1 of 4

Elo ranking for BA 1v1 competitive play

Posted: 05 Apr 2012, 18:01
by Griffith
Hi. I propose to make an elo base rankings for BA 1v1 competitive play. Calculation would be based on the 1v1 tournaments results and also on 1v1 played games with some restrictions.
I started with the last 1v1 tournament and here are the results:

Name__________Win___Loss__Elo___Rank
Jazcash_________9____0____1635__1
TeBe___________4____2____1530__2
MightySheep_____4____3____1515__3
Johannes________2____2____1500__4
HappyFace______1____2____1485__5-8
MasterTBC______2____3____1485__5-8
zorro___________4____5____1485__5-8
Flash___________1____2____1485__5-8
Beherith_________0____2____1470__9-12
yakor___________0____2____1470__9-12
halcyon_________0____2____1470__9-12
Norfolk-in-Chance 0____2____1470__9-12

I will start investigate old tournaments in the past years this week end to calculate the elo of old players (but actual elo will be based on the last tournament and the next one).
This is designed only for BA 1v1 competitive play.

Re: Elo ranking for BA 1v1 competitive play

Posted: 05 Apr 2012, 20:59
by Jazcash
Didn't the ladder system Koshi and peoples were working on have this sort of thing?

Re: Elo ranking for BA 1v1 competitive play

Posted: 05 Apr 2012, 21:41
by Griffith
There is elo ranking on some autohost.
But this is for BA only and 1v1 only. I dont think there is already rank for that.

Re: Elo ranking for BA 1v1 competitive play

Posted: 05 Apr 2012, 22:01
by Jazcash
Griffith wrote:There is elo ranking on some autohost.
But this is for BA only and 1v1 only. I dont think there is already rank for that.
Yeah, there was a system for it. I don't think the site's still up, the system isn't running any more anyway though.

http://projects.springlobby.info/projects/springladder

It wasn't really integrated with anything though, but I'm sure it was getting there. Can't remember why it died.

Re: Elo ranking for BA 1v1 competitive play

Posted: 05 Apr 2012, 22:21
by Johannes
I don't remember what exactly was the issue with it, BD worked at some point again to get it back running, but unfortunately didn't get that finished. That must be months and months ago though.

Re: Elo ranking for BA 1v1 competitive play

Posted: 05 Apr 2012, 22:41
by koshi
New engine version came that crapped on our efforts, ladder died. Very few of those people who've cried and cried and cried for a general ladder system were willing to put _any_ sort of effort/help in. The usual around here, you know.

That vented, I'm almost at a point again were its mostly functional. It's entirely built on replay parsing now, meaning any games from anywhere can be collated, besides a bot watching 'live' ofc. It's also got scoring that was actually designed for arbitrarily sized teams, unlike ELO that strictly is 1vs1.

If you hadn't seen it in action Jaz, I did eventually implement your design: http://ladder.springrts.com/

Re: Elo ranking for BA 1v1 competitive play

Posted: 06 Apr 2012, 01:47
by Jazcash
koshi wrote: If you hadn't seen it in action Jaz, I did eventually implement your design: http://ladder.springrts.com/
I saw it a while back, nice stuff :)

If you ever need any help on the design, testing or anything, give me a shout. I'd like to see this up and running again.

On an unrelated, funky note: Remember those cups that TASClient used to display? :D Those were funny. Flop had so many cups :mrgreen:

Re: Elo ranking for BA 1v1 competitive play

Posted: 07 Apr 2012, 16:24
by Griffith
As i promised, ratings based on old tournaments. Here the top 10 players:

Name-------20/12/2010##Nb games##Rank
[PinK]8D-------1675----------21--------------1
Triton ----------1608----------9---------------2
Jazcash--------1602----------21--------------3
Carpenter-----1597----------21--------------4
Wisse----------1575----------11--------------5
Llamadeus----1559---------16---------------6
flopflop---------1548---------25--------------7
SteveZissou---1545----------9-------------8-10
Sleksa----------1545----------7-------------8-10
Masu-----------1545-----------7-------------8-10

Those ratings are based on 6 tournaments. Total of 72 players. Lowest rating : 1456.

Re: Elo ranking for BA 1v1 competitive play

Posted: 07 Apr 2012, 17:52
by marciolino
Griffith wrote: Name__________Win___Loss__Elo___Rank
Jazcash_________9____0____1635__1
TeBe___________4____2____1530__2
MightySheep_____4____3____1515__3
Johannes________2____2____1500__4
HappyFace______1____2____1485__5-8
MasterTBC______2____3____1485__5-8
zorro___________4____5____1485__5-8
Flash___________1____2____1485__5-8
Beherith_________0____2____1470__9-12
yakor___________0____2____1470__9-12
halcyon_________0____2____1470__9-12
Norfolk-in-Chance 0____2____1470__9-12
Is it fair?
For example I had the 4th place at the tourney and still get the 8th place on this ranking. Because i lost the 3th place match. :roll:

Re: Elo ranking for BA 1v1 competitive play

Posted: 07 Apr 2012, 17:55
by Jazcash
marciolino wrote: Is it fair?
For example I had the 4th place at the tourney and still get the 8th place on this ranking. Because i lost the 3th place match. :roll:
Personally, I don't like ELO much. Don't get me wrong, I think it's a much more accurate portrayal of skill than ingame time. However, it's still not really that accurate compared to some ranking systems that other games use.

Re: Elo ranking for BA 1v1 competitive play

Posted: 07 Apr 2012, 18:29
by very_bad_soldier
marciolino wrote: Is it fair?
For example I had the 4th place at the tourney and still get the 8th place on this ranking. Because i lost the 3th place match. :roll:
It is just not enough input data. How fair is it that Behe is place 9-12 in that skill rating? It doesnt tell much...

Re: Elo ranking for BA 1v1 competitive play

Posted: 07 Apr 2012, 18:39
by Griffith
marciolino wrote: Is it fair?
For example I had the 4th place at the tourney and still get the 8th place on this ranking. Because i lost the 3th place match. :roll:
You are ranked 5 - 8 same elo as 3 other guys. That means you are technically 5th on the tourney which is not far from your 4th place.
Jazcash wrote: Personally, I don't like ELO much. Don't get me wrong, I think it's a much more accurate portrayal of skill than ingame time. However, it's still not really that accurate compared to some ranking systems that other games use.
What othey ranking systems are you talking about ? I'm interested to know about that. I can think about the points in tennis maybe which is very different from the elo ranking.

I use the same elo system as the FIDE for Chess (which is the international chess ranking). Elo is accurate to compare people between them. It is not the elo by himself that is accurate but the difference between the elo that mean something. For example someone that have 72 elo more than other guy should win 6 games out of 10 against him.
very_bad_soldier wrote:It is just not enough input data. How fair is it that Behe is place 9-12 in that skill rating? It doesnt tell much...
This is right, need more games to be accurate. Behe only played 2 games so his elo doesnt mean a lot.

Re: Elo ranking for BA 1v1 competitive play

Posted: 08 Apr 2012, 02:14
by Griffith
New ratings after the 1v1 tournament of 7/04/2012.

Name--------------Win-----Loss----Elo------Rank
Jazcash------------13--------2------1633------1
rANDY---------------9---------3------1607------2
MightySheep-------9---------5------1559------3
Tebe----------------10--------6------1550------4
eXist-----------------2---------2------1502------5
Johannes-----------3---------4------1485------6-10
HappyFace----------1--------2-------1485------6-10
MasterTBC----------2---------3------1485------6-10
zorro-----------------4---------5------1485------6-10
Flash-----------------1---------2------1485------6-10
Shox-----------------0---------2------1481------11-12
Princess-Grapes---0---------2------1481------11-12
DraonSpawn-------0---------2------1471------13-14
DayWalkeRz -------0---------2------1471------13-14
Sleksa---------------0---------2------1470------15-20
Eximius--------------0---------2------1470------15-20
Beherith-------------0---------2------1470------15-20
yakor----------------0---------2------1470------15-20
halcyon--------------0---------2------1470------15-20
Norfolk-in-Chance 0---------2------1470------15-20

Re: Elo ranking for BA 1v1 competitive play

Posted: 08 Apr 2012, 08:31
by Senna
BA needs definitelly elo ranking, it would be a great idea.

Actually balance its based on ingame time, does about wins, loses will balance the games better?.

Anyway, The best to balance games even more than elo or about winning and loses balance system its still human:)

Re: Elo ranking for BA 1v1 competitive play

Posted: 08 Apr 2012, 11:27
by PRO_rANDY
dunno if you have seen this thread griffin

http://springrts.com/phpbb/viewtopic.php?f=16&t=15325

Re: Elo ranking for BA 1v1 competitive play

Posted: 08 Apr 2012, 15:03
by Griffith
PRO_rANDY wrote:dunno if you have seen this thread griffin

http://springrts.com/phpbb/viewtopic.php?f=16&t=15325
Yeah i saw this thread Randy, but i can make elo calculation only on clear results. That is why i used only 6 tournaments that were containing enough information for me. Actual rankings are only based on the 2 last tournaments.
And congrats for your victory :-)

Re: Elo ranking for BA 1v1 competitive play

Posted: 08 Apr 2012, 18:05
by Jazcash
The problem with Elo is that it doesn't take accomplishment of victory into account. For example, if a pro player beats a new player 10 to 0, he gets the same amount of points if it were the other way around and the new player beat the pro player 10 - 0.

Also, as mentioned in the past, it's easy to cheat an Elo system by playing a friend and having him let you lose or just having two computers and doing it yourself.

Ofc, there are ways around this. Hmm. I think an Elo system would be nice, it has to be thought out a bit more though imo.

I like Halo's ranking system quite a lot. http://halowiki.net/p/ranks

Re: Elo ranking for BA 1v1 competitive play

Posted: 08 Apr 2012, 19:10
by luckywaldo7
x_x
The number of points won or lost in a contest depends on the difference in the ratings of the players, so a player will gain more points by beating a higher-rated player than by beating a lower-rated player.
(from http://en.wikipedia.org/wiki/Elo_rating_system)

Actually, Halo uses TrueSkill iirc, which is just a modified Elo ranking system.

LoL also uses a modified Elo ranking system, but everyone just refers to it as Elo.

Re: Elo ranking for BA 1v1 competitive play

Posted: 08 Apr 2012, 19:12
by Jazcash
luckywaldo7 wrote:x_x
The number of points won or lost in a contest depends on the difference in the ratings of the players, so a player will gain more points by beating a higher-rated player than by beating a lower-rated player.
(from http://en.wikipedia.org/wiki/Elo_rating_system)

Actually, Halo uses TrueSkill iirc, which is just a modified Elo ranking system.

LoL also uses a modified Elo ranking system, but everyone just refers to it as Elo.
Yeah, I guess a modified Elo system is what I should've just said off the bat :P

Re: Elo ranking for BA 1v1 competitive play

Posted: 08 Apr 2012, 19:13
by Griffith
Jazcash wrote:The problem with Elo is that it doesn't take accomplishment of victory into account. For example, if a pro player beats a new player 10 to 0, he gets the same amount of points if it were the other way around and the new player beat the pro player 10 - 0.
Wrong. Elo calcul is based on the rating of the two opponents. For the first tournament, no one had a rating so i put everyone on 1500. That is why you thought it was the same beating a noob or a pro.

To see what would happen now to the elo :
You are rated 1633, you fight a noob (default rating 1500). You win 10 to 0.
Your new elo will be 1729 and the noob will be rated 1404.
Now you fight someone strong but not rated. You loose 10 to 0.
Your new elo will be 1429 and your opponent will jump from 1500 to 1704.

That means you will win 96 elo and your opponent will loose 96 elo if you win 10-0. But if you loose 10-0 you will loose 204 elo and your opponent will gain those 204 elo !
Jazcash wrote: Also, as mentioned in the past, it's easy to cheat an Elo system by playing a friend and having him let you lose or just having two computers and doing it yourself.
That is why i calculate only with the tourney results and not with random 1v1. It could be possible to take into consideration random 1v1 but that would need to have a supervisor to avoid the possible cheat you were talking about.