Jump to content


Photo

Alternative Stratego Rankings (ASR)


  • Please log in to reply
46 replies to this topic

#1 Master Mind

Master Mind

    Major

  • Tournament Manager
  • 1,224 posts
  • Coat of arms
  • Platinum General

Posted 28 March 2018 - 01:06 PM

For the last couple of weeks, I have been working on an alternative ranking system for Stratego, based on the Glicko System. The reasons I started doing this are actually quite simple: I wanted to learn better how Microsoft Excel works, I was interested how ranking systems worked generally and I was curious to see how certain results affected the rankings. 

 

My goal with this project is nothing but to create an additional Stratego Ranking, besides the Kleier ranking. It is not my intention to replace the ASR with the Kleier Ranking in any way. However, I do have a few objections to the Kleier ranking, which I try to improve in the ASR, the main objection being the use of effective games, which, in my opinion, do not make the Kleier Ranking very reliable and up-to-date.

 

In the ASR, only international tournaments will be included. Furthermore, the ranking will only be based on played games. This means the actual tournament results will not always match exactly with the rankings. 

 

In the future, any updates will be posted in this topic, together with a spreadsheet of the standings, the results and the calculations.

 

More detailed information about the Glicko System I am using can be found here: http://www.glicko.ne...icko/glicko.pdf.

I highly suggest reading this document for those interested.

 

If there are any questions, you can ask them in this thread, or send me a PM.

 

Master Mind


  • Don_Homer and TheOptician like this

#2 Master Mind

Master Mind

    Major

  • Tournament Manager
  • 1,224 posts
  • Coat of arms
  • Platinum General

Posted 28 March 2018 - 01:17 PM

Updated: WCO 2017-2018

Date: 28-3-2018

 

Alternative Stratego Rankings:

80kLYoo.png

 

Link Ranking Tablehttps://imgur.com/gallery/glG0n

Link Spreadsheet: https://www.dropbox.... 2018.xlsx?dl=0



#3 astros

astros

    Stratego TM

  • WC Online Team
  • 812 posts
  • Coat of arms
  • Platinum Captain

Posted 28 March 2018 - 01:27 PM

Can you explain why you assume it would take 100 rating periods for a player's RD to return to return to 350? It seems that you are defining a time period as one tournament, so this would imply it will take a minimum of 10 years for RD to reset.


I'm in love with Stacy's mom.

#4 Master Mind

Master Mind

    Major

  • Tournament Manager
  • 1,224 posts
  • Coat of arms
  • Platinum General

Posted 28 March 2018 - 01:30 PM

Can you explain why you assume it would take 100 rating periods for a player's RD to return to return to 350? It seems that you are defining a time period as one tournament, so this would imply it will take a minimum of 10 years for RD to reset.

 

I indeed define one tournament as one rating period. I don't understand how you come to the statement about 100 rating periods. Could you please clarify that?



#5 astros

astros

    Stratego TM

  • WC Online Team
  • 812 posts
  • Coat of arms
  • Platinum Captain

Posted 28 March 2018 - 02:12 PM

I indeed define one tournament as one rating period. I don't understand how you come to the statement about 100 rating periods. Could you please clarify that?

Your spreadsheet is a bit hard to follow. What value are you using for your constant?


I'm in love with Stacy's mom.

#6 Master Mind

Master Mind

    Major

  • Tournament Manager
  • 1,224 posts
  • Coat of arms
  • Platinum General

Posted 28 March 2018 - 02:38 PM

I think you can find most of the information here: http://www.glicko.ne...icko/glicko.pdf I think it is quite self-explanatory.

 

Briefly, I use the term PR for Personal Rating, PRD for Personal Rating Deviation, OR for Opponent's Rating and ORD for Opponent's Rating Deviation. R stands for Result, g(RD) is derived from your opponent's RD, while E stands for the expectancy. d needs to be calculated to calculate r', the exact New Rating. Also, d is used to calculate RD', the exact New Rating Deviation. Factor is something useless. It says how much games a particular player won relatively seen to 10 games. I used this to see whether the Ranking Table was accurate or not.

 

Furthermore, I think you mean the constant 'c' in the calculation of the Rating Deviation for a player who did not play in a rating period / tournament. If a rated player has the same rating deviation as an unrated player (350), that player will be removed from the Ranking Table. As I want to see this happen after approximately a year, we should do that calculation based on 10 tournaments. This gives c = 109,5. However, this formula will only be used in upcoming tournaments.



#7 Napoleon 1er

Napoleon 1er

    General

  • Honorary members
  • PipPipPipPipPipPipPipPipPipPipPip
  • 2,631 posts
  • Coat of arms
  • Platinum General

Posted 28 March 2018 - 04:06 PM

One weakness of kleier is that once a ranked player stops playing then his ranking remains more or less there where it is with the exception of buch effect. But a player can stop playing for 2 years he is still keeping his position in the ranking. One shall be that any games played more than 12 months ago shall be removed from the calculation, only the games of last 12 ( or let' say last 24) months shall count. Same principle as tennis ATP ranking system. So you really get a ranking which is up to date with active players only.
If you don't know where you go ... you have a lot of chance to arrive elsewhere ...

#8 TheOptician

TheOptician

    Marshal

  • Tournament Manager
  • 3,117 posts
  • Coat of arms
  • Platinum Marshal

Posted 28 March 2018 - 05:15 PM

It is too arbitrary to say that games played 12 months ago count, but games played 13 months ago do not. It is also too arbitrary to say that games played 25 matches ago count, and games played 26 matches ago do not. Ideally they should count gradually less over time - I'm not sure what the rate of decay should be however.

Those players in the Kleier ranking that played eg 3 games 5 years ago seem only to gain ranking points, when it should be the reverse.

I think that the time it should take for a game to count for absolutely nothing however - should be quite long. A series of wins 6 years ago should still count for something if the player hasn't played since.
  • Don_Homer likes this

#9 Master Mind

Master Mind

    Major

  • Tournament Manager
  • 1,224 posts
  • Coat of arms
  • Platinum General

Posted 28 March 2018 - 05:51 PM

The ASR does not make use of effective games, because one tournament is seen as one rating period. After about 12 months, a player who did not participate in any international tournament will not be shown in the Ranking Table anymore. However, if that particular player plays in a tournament again, his last rating will be used, although his RD will be 350.

 

By implementing this, someone's rating is a fixed number. Within the Kleier rating, people can increase and/or decrease in rating without even playing because of these effective games. This is clearly seen in for example NDJ, increasing 147 points in rating without even playing and SEKERTZISS decreasing 87 points without even playing. In my opinion, this is a flaw in the Kleier system. 


  • TheOptician likes this

#10 Don_Homer

Don_Homer

    Captain

  • Members
  • PipPipPipPipPipPipPipPip
  • 751 posts
  • Coat of arms
  • Platinum Scout

Posted 28 March 2018 - 08:55 PM

Nice efforts Master Mind! The quality is for me still hard to judge at the moment though. Im looking forward to more data to compare with Kleier :)

 

In my opinion Kleier is quite good but as mentioned has a few flaws. I agree with the ones TheO mentioned and would like to add that there are sometimes players at the top of the ranking that were way too high ranked compared to their level. They had one passable or even outstanding tournament and suddenly they were at the top 10 (and even first place). The first player had mediocre results in the past (years ago) and the second had just very few tournaments played (which made the tournaments significance very high. Too high if you ask me. 

 

Lastly I would like to add that I am afraid no ranking system is perfect. There will Always be players that are ranked too high and players that are too low in ranking compared to their actual level. 


Molto Bene, Thats a nica Donut !


#11 Napoleon 1er

Napoleon 1er

    General

  • Honorary members
  • PipPipPipPipPipPipPipPipPipPipPip
  • 2,631 posts
  • Coat of arms
  • Platinum General

Posted 29 March 2018 - 07:04 AM

The ASR does not make use of effective games, because one tournament is seen as one rating period. After about 12 months, a player who did not participate in any international tournament will not be shown in the Ranking Table anymore. However, if that particular player plays in a tournament again, his last rating will be used, although his RD will be 350.
 
By implementing this, someone's rating is a fixed number. Within the Kleier rating, people can increase and/or decrease in rating without even playing because of these effective games. This is clearly seen in for example NDJ, increasing 147 points in rating without even playing and SEKERTZISS decreasing 87 points without even playing. In my opinion, this is a flaw in the Kleier system.


Yes this is good ... keep the games for the calculation of kleier but when publishing the ranking only show ranking of the players who have played minimum 1 game in last 12 months. Seing for example NDJ in that ranking is non sense.
  • Master Mind likes this
If you don't know where you go ... you have a lot of chance to arrive elsewhere ...

#12 Master Mind

Master Mind

    Major

  • Tournament Manager
  • 1,224 posts
  • Coat of arms
  • Platinum General

Posted 09 April 2018 - 04:21 PM

I'd like to give some more details and information about some things concerning the ASR.

 

First, about the rating periods I will use. Because the ASR is purely based on played games, not on tournaments, games must be put in chronologically. This means games from overlapping tournaments must be put into the same rating period. I decided I will use the following rating periods:

 

eQwCD1H.png

 

This gives 4 updates on the ASR in one year.

 

Secondly, the value I use for in the calculation of the new Rating Deviation. As I already said earlier, a player not participating in one year (4 rating periods) will not be shown on the ranking list anymore. Hence, after 4 rating periods your RD should be back to a value of 350 (used for non-ranked players). Therefore, I calculate using 4 as the amount of rating periods for the RD to shoot back to its original value.

 

By this, I should also rectify my statement made few posts above, saying ' If a rated player has the same rating deviation as an unrated player (350), that player will be removed from the Ranking Table. As I want to see this happen after approximately a year, we should do that calculation based on 10 tournaments. This gives c = 109,5. However, this formula will only be used in upcoming tournaments.' Actually, the value of will be different from each starting RD. Also, I will use 4 rating periods in my calculations instead of the '10 tournaments' mentioned above. 

 

I will update everything in the first post of this topic.



#13 astros

astros

    Stratego TM

  • WC Online Team
  • 812 posts
  • Coat of arms
  • Platinum Captain

Posted 09 April 2018 - 04:33 PM

The problem with using the Kleier Ranking for seeding is that the sample size of tournament games is too small to produce reliable results. Do you feel that this is a problem for your ranking, and if so would you recommend that the ASR is used for any form for seeding purposes? Or is this just akin to a tournament index, similar to the HOPPI, that will never be adopted?

Edited by astros, 09 April 2018 - 04:37 PM.

I'm in love with Stacy's mom.

#14 TheOptician

TheOptician

    Marshal

  • Tournament Manager
  • 3,117 posts
  • Coat of arms
  • Platinum Marshal

Posted 09 April 2018 - 05:25 PM

astros - How many games annually would you consider an adequate sample size?



#15 astros

astros

    Stratego TM

  • WC Online Team
  • 812 posts
  • Coat of arms
  • Platinum Captain

Posted 09 April 2018 - 05:29 PM

astros - How many games annually would you consider an adequate sample size?

A minimum of 30 for each player, each player's opponent and each player's opponent's opponents.


I'm in love with Stacy's mom.

#16 Master Mind

Master Mind

    Major

  • Tournament Manager
  • 1,224 posts
  • Coat of arms
  • Platinum General

Posted 09 April 2018 - 05:34 PM

>The problem with using the Kleier Ranking for seeding is that the sample size of tournament games is too small to produce reliable results

 

In my opinion the sample size of tournament games will stay the same, whatever formula or ranking you will use, because there is a pretty fixed amounts of tournaments and so a limited amount of games a year. I'm not sure what sample size you would need to have to come close to getting reliable results into a ranking. 

 

>Do you feel that this is a problem for your ranking, and if so would you recommend the ASR is used for any form for seeding purposes?

 

To be honest, I don't think the amount of tournament games is a great deal for the ASR. Moghedien, for example, already played 26 games only within WCO, Winter Tournament 2018 and PRO Ladder 2018. I think such an amount of games has already the ability in the ASR to reflect his abilities pretty well. However, the ratings of his opponents might only be based on 2 - 5 games, which decreases the reliability of his rating for now. If the ASR is well-established, with quite some results put in, I think it indeed can potentially be used for seeding purposes. 

 

>Or is this just akin to a tournament index, similar to the HOPPI, that will never be adopted?

 

I started up this whole project because I like to do it, and I learn quite a lot from it, for example how Microsoft Excel generally works. I try hard to improve the flaws I see in the Kleier rankings, giving me quite some information about how rankings work, and how certain changes to formulas change the whole outcome of the system. Like I said in my first post, the ASR is not meant to take over Kleier in any way. However, I think it can be pretty helpful to improve Kleier.

 

>A minimum of 30 for each player, each player's opponent and each player's opponent's opponents

 

Which amount of time would that 30 games need to be in?



#17 Master Mind

Master Mind

    Major

  • Tournament Manager
  • 1,224 posts
  • Coat of arms
  • Platinum General

Posted 09 April 2018 - 05:37 PM

Updated: Rating Period 1 (Winter Tournament 2018, PRO Ladder 2018)

Date: 9-4-2018

 

Alternative Stratego Rankings:

GsbFp9q.png

 

Link Ranking Table: https://imgur.com/GsbFp9q

Link Spreadsheet Rating Period 1: https://www.dropbox.... 2018.xlsx?dl=0



#18 astros

astros

    Stratego TM

  • WC Online Team
  • 812 posts
  • Coat of arms
  • Platinum Captain

Posted 09 April 2018 - 05:38 PM

To be honest, I don't think the amount of tournament games is a great deal for the ASR. Moghedien, for example, already played 26 games only within WCO, Winter Tournament 2018 and PRO Ladder 2018. I think such an amount of games has already the ability in the ASR to reflect his abilities pretty well. However, the ratings of his opponents might only be based on 2 - 5 games, which decreases the reliability of his rating for now.

 

This is the exactly the problem with having a small sample size. Unfortunately, you cannot construct an accurate ranking while this persists.

 

>A minimum of 30 for each player, each player's opponent and each player's opponent's opponents

 

Which amount of time would that 30 games needed to be in?

 

It depends on how quickly the player's abilities are changing. For someone like me or you a year is probably fine. For a new player with less than 100 games, you might choose a period as short as 3 months.


I'm in love with Stacy's mom.

#19 roeczak

roeczak

    Captain

  • Tournament Manager
  • 924 posts
  • Coat of arms
  • Platinum Marshal

Posted 09 April 2018 - 07:26 PM

I like this project better than Kleier for online tournaments.
If you enjoy stratego you might want to subscribe to Roeczak. Member of Stratego Captains Club. Actively trying to promote the game and would like to help in any activity towards that goal. Highest Rating : 898 (Platinum Marshal)

#20 Napoleon 1er

Napoleon 1er

    General

  • Honorary members
  • PipPipPipPipPipPipPipPipPipPipPip
  • 2,631 posts
  • Coat of arms
  • Platinum General

Posted 09 April 2018 - 07:53 PM

How can playa1 have mre than me while he did worse than me in all 3 tournaments wco, winter and pro ladder?
If you don't know where you go ... you have a lot of chance to arrive elsewhere ...




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users