Page 12 of 15 FirstFirst ... 21011121314 ... LastLast
Results 111 to 120 of 145

Thread: Clan War statistics taken from the API

  1. #111
    Forum Superstar dorsan's Avatar
    Join Date
    Mar 2015
    Location
    under a rock
    Posts
    3,084
    Quote Originally Posted by urizano View Post
    I might pop in at this point of the discussion. I cite Darian, SC communications manager with "there is huge amount of code behind the CW matchmaking ...". We might discount for the fact that a communications manager is not a developer, and what looks huge to the former is .. manageable to the other.
    I honestly doubt he has ever seen the code himself. I have never worked in a company where employees other than the programmers even had access to the code.

    Quote Originally Posted by urizano View Post
    But assuming matchmaker to be (outright) simple is, in fact, accusing Darian that he lied. I wouldn't go so far, and take it for granted that matchmaking is complicated, at least.
    Nah, not a lie, just a PR talk. I'd discount that any day.

    Quote Originally Posted by urizano View Post
    The point is that we do not have access to the total set of clans looking for a war at a given moment. A neatly designed matchmaker may end up in a large number of clans searching forever. I assume the matchmaker has a simple basis but a number of fixes for this problem made the current matchmaker complicated.
    Even if they would go with a really good matchmaker - like the one I have in mind for over a year now - it would be still a simple and small code. No matter how many fixes you do it's simply not a problem that requires a long code. Then again your definition of 'complex' code might be quite different from mine.
    Leader of clan Purity - #8U82RCR9
    (367/85/20 war score, clan level 13, recruiting th3-th10,
    best win streak: 30, ended on 8/16/2016, best perfect war streak: 13)
    My latest war matchmaker solution: https://forum.supercell.com/showthre...hts?p=10886766

  2. #112
    Skrags's Avatar
    Join Date
    Nov 2015
    Location
    221B Baker Street
    Posts
    4,812
    There's some useful historic info about the MM complexities in the video below (2.20 on). This follows a confirmed direct discussion with the Dev team :-

    https://www.youtube.com/watch?v=BqR3M7Ss_Hs

    Clan Name : SOUTH WEST 45, Clan Tag : #90LP2PL. Highest Win Streak 12!
    Clan Details : Established 2012! Adult, English speaking International Level 10 War Clan (twice a week). Recruitment Thread : https://forum.supercell.com/showthre...H8-s-to-TH11-s My spec: TH10, Heroes 40/40

  3. #113
    Senior Member NeoGen's Avatar
    Join Date
    Jul 2014
    Posts
    206
    I am finding this thread very interesting and helping me to try to figure out where to go next with war,

    Quote Originally Posted by Skrags View Post
    There must be millions of these bases sitting in bronze and below just waiting to be looted!!!
    There may not be millions, but there is a considerable amount, I have a defenceless TH11 and I sit with it between Bronze 1 and 3, but prefer to keep it in Bronze 3, most bases I find when raiding are other defenceless bases. I keep it there as the higher you go, the smarter attackers get and they can take your loot without giving a shield.
    Nex Addoians #YJYP2LPY Relaxed adult war clan with plenty of banter!
    Nex Addo #9Q8RQ999 On a push to Level 10 and in need of good p[layers to help

  4. #114
    Senior Member NeoGen's Avatar
    Join Date
    Jul 2014
    Posts
    206
    I'm a little confused, reading through this, regarding the most popular roster. Have I read it correctly that you are saying the following:

    10:10
    Normal with TH spread of TH10-TH7

    15:15
    Engineered with TH spread of: TH11-TH7

    20:20
    Engineered with TH spread of: TH11-TH7

    25:25
    Engineered with TH spread of: TH11-TH4

    30:30
    Engineered with TH spread of: TH11-TH3

    40:40
    Engineered with TH spread of: TH11-TH3

    50:50
    Engineered with TH spread of: TH11-TH9

    So only 10:10 has a normal clan as the most popular? So if we are a normal clan, then 10:10 with a TH spread of TH10-TH7 is the best? If we go in to any of the others then we stand a higher risk of being matched with an engineered clan?
    Last edited by NeoGen; 1 Week Ago at 01:34 PM.
    Nex Addoians #YJYP2LPY Relaxed adult war clan with plenty of banter!
    Nex Addo #9Q8RQ999 On a push to Level 10 and in need of good p[layers to help

  5. #115
    Junior Member
    Join Date
    Sep 2017
    Posts
    53
    In a very similar process as described by dorsan above I obtained
    559509 clans with clanlevel 10 and higher. I focus on high-level clans
    because there are more likely to haven some sort of strategy which I
    hope will make my analysis clearer.

    From those, 199205 had open logs. Apparently, the high-level clans
    prefer to hide their warlog!

    They participated in 104902 wars. In these wars, 18136 of the
    opponents were in my initial list. The remaining 76167 were either not
    in war or in a war with no open log.

    Clan war sizes are
    5:5 38 0.0
    10:10 49228 46.9
    15:15 26251 25.0
    20:20 13019 12.4
    25:25 6040 5.8
    30:30 4484 4.3
    40:40 2978 2.8
    50:50 2863 2.7

    I have no explanation for the 5:5 wars, maybe they are available to
    certain test accounts. I will focus on 10:10 wars, anyway.

    I'm in the middle of downloading player data which will take some more days.
    Meanwhile, I did a first small analysis how good the number and levels and wardens are matched - as this was often complained here in the forum.
    Strategy co-leader of clan 40CW.
    How war matchmaking works - my take: https://forum.supercell.com/showthre...1#post10989427 .

  6. #116
    Junior Member
    Join Date
    Sep 2017
    Posts
    53
    Grand Warden disparity

    Among 4173 10:10 wars where at least on team has a warden, I compared the number of wardens in each team. In 40.3% of the cases, their number was matched. In 42.8% of cases there was a difference of one warden. In 11.8% there was a mismatch of two. There were even cases where one team had 7 more wardens than the other! I'm talking about 10:10-wars, by the way.

    The expectation value of the mismatch of the number of wardens is 0.81 among the wars with a warden involved. In fact, this is way more mismatch than I expected, as having a warden is closely correlated to other things which affect the war strength: town hall level 11, with the option to upgrade war troops and spells, the option to upgrade defensive buildings and build eagle artillery.

    What about the mismatch of combined warden levels? The expectation value is 9.9. In 8.5% of cases it matches. In 5.8% the offset is 1, 6.2% offset is 2, 5.8% offset is 3 and so on.

    To put these numbers into perspective, I simulated completely random matches between parties of 10, where the frequency of warden levels is considered (85% haven't got one, 6% have it at level 20, and so on). The expectation value of the mismatch of the number of wardens of random matchings would be 1.3 while the mismatch of combined levels would be 19.7.

    To sum it up - the disparity of warden number and levels is large. It is somewhat less than in a completely random match. I can basically confirm what people say - the warden doesn't have much impact in matchmaking. Differences in warden number and levels is the rule, not an exception.
    Strategy co-leader of clan 40CW.
    How war matchmaking works - my take: https://forum.supercell.com/showthre...1#post10989427 .

  7. #117
    Forum Contender
    Join Date
    Apr 2016
    Posts
    4,893
    Quote Originally Posted by urizano View Post
    Grand Warden disparity

    Among 4173 10:10 wars where at least on team has a warden, I compared the number of wardens in each team. In 40.3% of the cases, their number was matched. In 42.8% of cases there was a difference of one warden. In 11.8% there was a mismatch of two. There were even cases where one team had 7 more wardens than the other! I'm talking about 10:10-wars, by the way.

    The expectation value of the mismatch of the number of wardens is 0.81 among the wars with a warden involved. In fact, this is way more mismatch than I expected, as having a warden is closely correlated to other things which affect the war strength: town hall level 11, with the option to upgrade war troops and spells, the option to upgrade defensive buildings and build eagle artillery.

    What about the mismatch of combined warden levels? The expectation value is 9.9. In 8.5% of cases it matches. In 5.8% the offset is 1, 6.2% offset is 2, 5.8% offset is 3 and so on.

    To put these numbers into perspective, I simulated completely random matches between parties of 10, where the frequency of warden levels is considered (85% haven't got one, 6% have it at level 20, and so on). The expectation value of the mismatch of the number of wardens of random matchings would be 1.3 while the mismatch of combined levels would be 19.7.

    To sum it up - the disparity of warden number and levels is large. It is somewhat less than in a completely random match. I can basically confirm what people say - the warden doesn't have much impact in matchmaking. Differences in warden number and levels is the rule, not an exception.
    So in 83% of cases the no of wardens matched or was 1 off, and in 94% of the cases the mismatch (in a 10 v 10) war wars at worst 2 difference, an as already stated around 85% of that was matches where the warden was ar most 1 place off.

    i think thats extememly interesring, but unlike you I thought the situation was worse than that.

    is it easy to check If in those cases where there was more than one warden, there was a significant offset in the total Queen levels? Im not sure how complicated that is to check, Im just wondering If Its simple, and If you have the data on that already?
    Last edited by Vikingchief; 1 Week Ago at 01:18 PM.
    Clan - Highland chiefs, Highland rebels, Highlandchiefs2
    Position - clan lunatic
    Accounts - 11 heavy with eagle!!! What have i done, D rushed 11; 10.5, 2 x 10.25; premmi 10.5; 2 x noob rushed 11; super 9.05 (9.11), badly rushed th11, solid super 8.25 (8.10).

  8. #118
    Forum Superstar dorsan's Avatar
    Join Date
    Mar 2015
    Location
    under a rock
    Posts
    3,084
    Quote Originally Posted by Vikingchief View Post
    So in 83% of cases the no of wardens matched or was 1 off, and in 94% of the cases the mismatch (in a 10 v 10) war wars at worst 2 difference, an as already stated around 85% of that was matches where the warden was ar most 1 place off.

    i think thats extememly interesring, but unlike you I thought the situation was worse than that.

    is it easy to check If in those cases where there was more than one warden, there was a significant offset in the total Queen levels? Im not sure how complicated that is to check, Im just wondering If Its simple, and If you have the data on that already?
    Honestly, that sounds bad enough for me. Remember we are talking about 10v10 wars. Interesting things to point out about those wars:
    1) a single warden difference can make or break the war in 10v10 due to small war size
    2) this is the war size which has the most clan in it, therefore the matches are the best
    3) the number of players you need to match is the lowest in 10v10

    If you consider those points, it looks pretty bad indeed.
    Leader of clan Purity - #8U82RCR9
    (367/85/20 war score, clan level 13, recruiting th3-th10,
    best win streak: 30, ended on 8/16/2016, best perfect war streak: 13)
    My latest war matchmaker solution: https://forum.supercell.com/showthre...hts?p=10886766

  9. #119
    Forum Contender
    Join Date
    Apr 2016
    Posts
    4,893
    Quote Originally Posted by dorsan View Post
    Honestly, that sounds bad enough for me. Remember we are talking about 10v10 wars. Interesting things to point out about those wars:
    1) a single warden difference can make or break the war in 10v10 due to small war size
    2) this is the war size which has the most clan in it, therefore the matches are the best
    3) the number of players you need to match is the lowest in 10v10

    If you consider those points, it looks pretty bad indeed.
    what numbers would be for you satisfactory? (Just barely acceptable)

    and what would you consider "good".

    finally, do you agree a significant number of those warden advantages have an offset in Iether Queen or infenoes or eagle? Lets say more than 50%? (I have no idea, thats my gut feel, wondering what you believe).
    Clan - Highland chiefs, Highland rebels, Highlandchiefs2
    Position - clan lunatic
    Accounts - 11 heavy with eagle!!! What have i done, D rushed 11; 10.5, 2 x 10.25; premmi 10.5; 2 x noob rushed 11; super 9.05 (9.11), badly rushed th11, solid super 8.25 (8.10).

  10. #120
    Forum Superstar dorsan's Avatar
    Join Date
    Mar 2015
    Location
    under a rock
    Posts
    3,084
    Quote Originally Posted by Vikingchief View Post
    what numbers would be for you satisfactory? (Just barely acceptable)

    and what would you consider "good".

    finally, do you agree a significant number of those warden advantages have an offset in Iether Queen or infenoes or eagle? Lets say more than 50%? (I have no idea, thats my gut feel, wondering what you believe).
    Consider this.

    For 10v10 each clan can have 0-200 total warden levels. If the distribution was random (which it isn't), you would still need only 40.000 clans in the pool to get an exact match on warden levels. Now add a tolerance of 10% on warden levels and you get to a fairly low number of clans needed. Then the number of wardens goes from 0-10. So with random distribution you need only 100 clans in the pool to get exact match for per level. Combine the per number of wardens match with the 10% tolerance on levels, and you'll get to some 1000-2000 clans needed in the pool (since those 2 are in fact correlated). 10v10 has the largest pool, there is no reason for any mismatches at 10v10 at all.
    Leader of clan Purity - #8U82RCR9
    (367/85/20 war score, clan level 13, recruiting th3-th10,
    best win streak: 30, ended on 8/16/2016, best perfect war streak: 13)
    My latest war matchmaker solution: https://forum.supercell.com/showthre...hts?p=10886766

Similar Threads

  1. Clan Statistics
    By Avery99 in forum Ideas & Feature Requests
    Replies: 0
    Last Post: August 7th, 2014, 08:33 AM
  2. clan war statistics
    By lqk98765 in forum Ideas & Feature Requests
    Replies: 0
    Last Post: July 28th, 2014, 04:04 PM
  3. Clan statistics
    By Obeezy in forum Ideas & Feature Requests
    Replies: 2
    Last Post: June 30th, 2014, 09:36 AM
  4. Clan War Statistics
    By Shoegazer in forum Clan Wars
    Replies: 0
    Last Post: April 19th, 2014, 12:58 PM
  5. My Clan Statistics
    By Mudeater in forum Ideas & Feature Requests
    Replies: 0
    Last Post: February 3rd, 2014, 03:10 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •