Character Tournament Rankings in Japan

Re: Version B Rankings

Ice-9 - This is going to be a long post. Hope it answers your questions! /forums/images/%%GRAEMLIN_URL%%/laugh.gif

Edit: The more I look at this post, it could be pretty confusing...

Single tournaments only.

The numbers in the ranking charts are based on two calculations: % character distribution and % character advancement. For each calculation, whoever have the best score is given the rank of 1 and the character with the worst score is given the rank of 17. The ranking charts numbers averaged the ranks from the two calculations. For example, Kage rank 1st in % character distribution score and rank 3rd in % character advancement for an average ranking score of 2nd.

The Calculations
The two type of calculations differs on what is being compared. The purpose of using two types of calculations to rank characters is that each calculation may have certain “biases†to characters that are either highly represented or lowly represented in tournaments. Thus, combining two different types of calculations can alleviate some of that bias.

% Character Distribution – If all characters are balanced having no advantage or disadvantage against other characters, you would expect that at every portion of the tournament the % character distribution will remain the same. Hypothetically, if character A is the choice of 5.9% of the total tournament entries, then it should make up 5.9% of the total semifinalists, total finalists, and total tournament winners. However, if certain characters have more or less success in advancing to semifinals, finals, and winning the tournament there will be a change in % character distribution during the later portions of the tournaments.

% Character Advancement – If all characters are balanced having no advantage or disadvantage against other characters, you would expect that each character will have the same chance as other characters in advancing to the next portion of the tournament. Hypothetically, a character has ~12.5% chance of advancing to the semifinals, then 50% chance to advance to the finals, and then 50% chance to win the tournament. However, if certain characters have more or less success in advancing to semifinals, finals, and winning the tournament compare to to other characters, their % chance to advance in the tournament may increase or decrease.

Below is both the raw data I gathered from http://www.virtuafighter.jp/eve_all.html and the calculations I used to make the Version B Ranking Chart in the original post (June 11th, 2007 to July 23rd, 2007). Note that it is not very easy on the eye.

For the most part, yup. If you look at the calculated data chart above, the last row is the standard deviation (SD) value for each column.

Approximately:
High/Low rankings is values > 2 SD
Above Average/Below Average rankings is values > 1 SD
Slightly Above Average/Slightly Below Average rankings is values > 0.5 SD
Average rankings is values < 0.5 SD

I made calculations at various parts of a tournament assuming that at the beginning of a tournament, the players’ skill can range from novice to advanced. The later portions of the tournament should have less novice players (eliminated from the tournament) and more advanced players; thus, the calculated data of the later portions of the tournament may reveal a character’s ability more than data from earlier portions of the tournament.

In hindsight, due to tournament format, rankings for later portions of tournaments are not as statistically sound compare to calculations for earlier portions of tournaments.

Only way to make it more statistically sound, is to increase the data set which is the reason I did the second chart for version B (November 2006 to July 2007). Unfortunately, the raw data is most complete for the last two months of tournaments. The website removes the tournament brackets for the older tournaments and list only the top 4 finishers in each tournament. Thus, I am unable to calculate "usage" and "semifinals" ranks. However, the "finals" and "winners" rankings are as informative as the first version B chart.

Edit: Just to clarify on what each column in the ranking charts represent.

Usage - Ranks which characters were chosen the most or used the most in tournaments.

Semifinals - Of all the characters chosen, what characters used by players have more success in advancing to the semifinals.

Finals - Out of all the semifinalists, what characters used by players have more success in advancing to the finals. Basically this column does not take into consideration the "Entry (Usage)" data and only considers the "Semifinals" data to determine ranks.

Winners - Out of all the finalists, what characters used by players have more success in winning the tournament. This column does not take into consideration the "Entry (Usage)" data and "Semifinals" data. This column only looks at the "Finals" data to determine ranks.

 

Version C - August Rankings

Character Rankings Version C – August

For comparison, Version B ranking chart from original post is next to the updated August chart.



The most dramatic changes in a character’s rank is listed below -
<u>
Semifinals Rankings -</u>
Wolf (-6 ranks)
Pai (-7.5 ranks)

<u>Final Rankings – </u>
Jeffry (+13 ranks)
Brad (+9.25 ranks)
Jacky (+6.25 ranks)
Akira (-6.5 ranks)
Pai (-7.5 ranks)
Lei (-8 ranks)

<u>Winner Rankings -</u>
Wolf (+10.75 ranks)
Jacky (+6 ranks)
Lei (-5.75 ranks)
Lion (-6.25 ranks)
Goh (-10.25 ranks)
Shun (-12.25. ranks)

Whether the difference in the two charts are due to month-to-month variance or version B-version C is debatable. Next update will be October.

 

Re: Version C - August Rankings

Unicorn, sure -



Jigohro - I think its mostly a bad month. Characters that are played less will move up and down the rankings a lot easier than characters that are played more.

 

Re: Version C - August Rankings

Ice-9 - In the raw data charts, the semifinalists are the players who made it to the semifinals but lost at the semifinals.

I agreed with the last sentence in your quote. Beginning of tournaments--the skill level between players will vary considerably. The further into tournament, the skill level between players will be less.

 

Version B - Character-Specifc Matchups

This is based on the Version B finalist (1st and 2nd place) match-ups that I used to make the Tournament Ranking Charts.



How to read the charts:
Looking at the Character's "Rows," you are looking at his/her wins or win ratio, respectively. Looking at a Character's "Column", you are looking at his/her losses or loss ratio, respectively.

For example, let's say we are interested on how Wolf players fair against Kage players for 1st place. Going down the rows in the first chart, I find Wolf's Row. Next, I looked to the right and find Kage's Column. One can see from the Wolf's Row / Kage's Column entry that Wolf have won 6 times against Kage. Looking at the second chart, those 6 victories give Wolf a measly 0.16 win ratio (16%) against Kage. If you want to see how much losses that actually entails, you can go to the first chart and look for Wolf's Column / Kage's Row entry (Wolf loses to Kage 29 times).

Note: The color "rankings" is somewhat deceiving due to low amount of certain character specific matches. So for the most part, I would just ignore the colors. Enjoy!

 

Version C - September Rankings



The most dramatic changes in a character's rank (September from August) is listed below -

Semifinals Rankings -
Blaze (+6 ranks)
Eileen (-5 ranks)
Wolf (+8 ranks)

Final Rankings –
Aoi (-7.5 ranks)
Blaze (+7 ranks)
Goh (-9 ranks)
Jacky (+5.25 ranks)
Sarah (+7 ranks)
Vanessa (-7.25 ranks)

Winner Rankings -
Aoi (+5 ranks)
Brad (-10.75 ranks)
Lau (+6.75 ranks)
Lion (+5 ranks)
Sarah (+8.75 ranks)
Shun (+5.25 ranks)
Vanessa (-5.25 ranks)
Wolf (-6 ranks)