Page 3 of 5 FirstFirst 12345 LastLast
Results 21 to 30 of 43

Thread: Matchmaking FAQ

  1. #21
    Forum Veteran TankSinatra's Avatar
    Join Date
    Sep 2014
    Location
    California
    Posts
    1,781
    Quote Originally Posted by ryryshouse View Post

    Hope I'm making sense
    I think so, though my math is barely beyond the high school level. Think the issue is I started a "frequently asked questions" without putting the question down. The question in my head is the common fresh spawn post, "OMG my th8 matched a low defense th11 fix this!!!", and the "reordering" I'm thinking of is the sort by defense on the roster after the match is made.
    Level 10 Adult War Clan METAMUCIL.
    And some of us are old enough that its not just a funny word...
    Metamucil keeps the red out of our logs. Small wars, great record, low pressure.
    War Record - https://cocp.it/clan/82GJJV2J

  2. #22
    Forum Veteran TankSinatra's Avatar
    Join Date
    Sep 2014
    Location
    California
    Posts
    1,781
    Quote Originally Posted by dorsan View Post
    It is more likely that the storage values are scaled from the real defensive weight. In which case you might have something like this:
    I'll clear up the language a bit. For the purposes of this FAQ its enough to say its a representation of defense weight (not necessarily exact).

    Quote Originally Posted by dorsan View Post
    Don't agree with this either. As much as defensive weight is measured by storage values, offensive weight is related to the recommended target on the first minute of the war before anyone attacks. But just as with the defensive weights, these results are not 100% representative of the actual values used in the matchmaker.
    I've used the 'rec target' for years and agree it works, but left it out as its more controversial. And I've got "FAQ for noobs" in my head as the post's purpose but wasn't sure what question that answered.
    Level 10 Adult War Clan METAMUCIL.
    And some of us are old enough that its not just a funny word...
    Metamucil keeps the red out of our logs. Small wars, great record, low pressure.
    War Record - https://cocp.it/clan/82GJJV2J

  3. #23
    Millennial Club
    Join Date
    Dec 2014
    Location
    Nashville, TN
    Posts
    1,196
    The other thing to keep in mind is that if you run the same roster over and over, the matches can vary greatly.

    This supports the theory that the margin of error, or "mismatch" is very great.

    Our matches we see poorly matched bases up to 30% of the roster.

    Typically it manifests in the following ways, most common to least common for our roster type.

    1. either TH8 or lower, or lopsided/eng , with us having less EA/IT bases "Roster Engineering" but not always intentional, just a wider TH spread.
    2. Other team having a glut of .5 bases matched to our maxed bases, typically both 9.5 and 10.5
    3. TWC bd ( no TH10s, but all 11 and 9 )


    What is interesting, is that in scenarios 2/3 there is a huge offensive disparity.
    Scenario 1 is harder to gauge, as sometimes it is "heros vs infernos" , but only bc the other opp has 5-6 no hero bases and some bases with teenagers.

    all 3 scenarios you can typically use gold storage weighting only and generate a line that fits. ( which is why you might see me say i havent seen proof of offensive weighting yet )

    I am coming around to dorsan's theory that the inputs to the line fit are not linearly represented by either recommended target and gold storage, this is making the most sense to me. ( there could also be a linear relation but in which offensive plays some weird fudge factor )

    I still think its a fairly remedial line , as that code is freely available in whatever language you want to implement it in. Also makes the most sense.

    look at the break down in our current war. 8/14/18 to 10/21/9 just not sure how their algorithm can suck this bad at this stage in the game.





    8/14/18 breakdown 10/21/9
    15/18/5 remaining 12/26/11
    1 11*** 0
    5 11** 1
    3 10*** 1
    10 10** 6


    **Hitrates**
    1/1 11v11 0/3
    0/0 11v10 1/5
    5/7 10v11 0/0
    0/3 10v10 0/5
    0/0 10v9 4/11
    9/11 9v9 0/7

  4. #24
    Millennial Club
    Join Date
    Mar 2015
    Posts
    1,133
    Given the selection effect that complaints have, some sort of descriptive disclaimer might be useful. I'm not sure how to stop all the "this match sucks, broken matchmaker" threads, but the FAQ should aim an answer at such thread creators.

    Here's my take, feel free to edit, as it's definitely too wordy and invokes math, both often flaws in a FAQ for noobs.
    ------------
    The matchmaker works to find decent matches in a reasonable amount of time. Sometimes it will give results outside normal parameters, either by taking longer than normal or giving a match that is atypically bad (or good). Sometimes it will give results outside normal multiple times in a row through no factor other than unlucky coincidence. However, ongoing trends of matches that are consistently "bad" for the same clan are generally driven by something about the clan itself versus properties of "similar" clans that match.

    Due to the number of matches that are made each day, hour, and minute, there will always be frequent imperfect matches. If you want to convince the forum that you've found an issue with the matchmaker, you will likely need to post detailed logs of the clans involved and document them in a going forward manner. Historical issues are not persuasive due to selection bias in who reports mismatches. 1 in 10,000 clans will have had 4 matches in a row at the 10th worst percentile at any given time. That doesn't mean anything is wrong for that clan (other than being the unluckiest out of 10,000).

    However, if there is an issue, the clan in question can start documenting it and since it's an issue (and not just bad luck) it will continue to occur in a going forward fashion. By looking at future results only, the selection bias of "clan is reporting past bad results" is eliminated. This will be extremely persuasive and demonstrate that "our lineup matches this set/type of lineups, it's bad for reason X, Y, and Z."

  5. #25
    Skrags's Avatar
    Join Date
    Nov 2015
    Location
    221B Baker Street
    Posts
    4,805
    I do think having a FAQ for new forum members is a great idea, especially if it helps with some of the basics that sit outside the 'how War works, prep day' etc etc, and hopefully dismisses some of the assumptions players may have before coming to the forums.

    I also tend to agree with some of the statements posted above about non linearity of weights and how they don't necessarily reflect actual gold storage or RT values, hence I tend to refer to storages as 'approximate' defensive weight (see page 1).

    Other suggestions :-

    'Websites' and spread sheets held on social media should be taken with a pinch of salt.
    Engineering is not cheating.
    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

  6. #26
    Forum Veteran
    Join Date
    Jun 2016
    Posts
    1,855
    Quote Originally Posted by TankSinatra View Post
    In the "Collecting Clan War Matchups" sticky, which is going nowhere, I suggested Darian put up a Matchmaking FAQ sticky instead of sending us on a snipe hunt. I'm not very confident it'll happen, so perhaps we should make our own. Happy to add suggestions, discuss below.

    To clarify, I'm looking to clear up common misconceptions, not present a guide. I'd like to stick with things we have a high degree of confidence about, things that SC should/has come right out to say without giving more detail about MM than they want to/should.
    You should add recent war record to the OP. As we don't agree on how it factors in, perhaps just something like "recent war record is also something that influences war matchups" would be appropriate?

    Quote Originally Posted by urizano View Post
    Current win streak and possibly related historical information are regarded in matchmaking (strongly in the past, less so now).
    That part in parenthesis is an opinion, which can hardly be supported with much data that you've personally gathered b/c the current mm iteration is so new. It's better just to specifically mention "recent war record" as SC does without dumping in our own opinions on the matter.

    Quote Originally Posted by TankSinatra View Post
    This I disagree with. SC mentioned recent win record as going into the matchmaker but never explained how, and certainly never confirmed it was win streak. A better hypothesis I've seen is that there's a simple checkbox for "recent loss rate > 50%" or something like that, to put weaker clans against eachother. But that's just a guess, without some confirmation of specifics I wouldn't put it in a FAQ. Darian could of course...
    How can you put this forward as a true FAQ if you don't even mention recent war record in the OP? Again, SC says that it factors in, we've all seen that it factors in, we just don't have any consensus on HOW it factors in. But it does, somehow, and saying "recent war record matters" or some such will get the point across to people. I certainly know from long experience that it has helped explain many, many otherwise strange matchups over the years.
    Last edited by MossackFonseca; 1 Week Ago at 03:51 PM.

  7. #27
    Forum Veteran
    Join Date
    Jun 2016
    Posts
    1,855
    Quote Originally Posted by Sservis View Post
    Given the selection effect that complaints have, some sort of descriptive disclaimer might be useful. I'm not sure how to stop all the "this match sucks, broken matchmaker" threads, but the FAQ should aim an answer at such thread creators.

    Here's my take, feel free to edit, as it's definitely too wordy and invokes math, both often flaws in a FAQ for noobs.
    ------------
    The matchmaker works to find decent matches in a reasonable amount of time. Sometimes it will give results outside normal parameters, either by taking longer than normal or giving a match that is atypically bad (or good). Sometimes it will give results outside normal multiple times in a row through no factor other than unlucky coincidence. However, ongoing trends of matches that are consistently "bad" for the same clan are generally driven by something about the clan itself versus properties of "similar" clans that match.

    Due to the number of matches that are made each day, hour, and minute, there will always be frequent imperfect matches. If you want to convince the forum that you've found an issue with the matchmaker, you will likely need to post detailed logs of the clans involved and document them in a going forward manner. Historical issues are not persuasive due to selection bias in who reports mismatches. 1 in 10,000 clans will have had 4 matches in a row at the 10th worst percentile at any given time. That doesn't mean anything is wrong for that clan (other than being the unluckiest out of 10,000).

    However, if there is an issue, the clan in question can start documenting it and since it's an issue (and not just bad luck) it will continue to occur in a going forward fashion. By looking at future results only, the selection bias of "clan is reporting past bad results" is eliminated. This will be extremely persuasive and demonstrate that "our lineup matches this set/type of lineups, it's bad for reason X, Y, and Z."
    This is a really good explanation.

  8. #28
    Forum Veteran TankSinatra's Avatar
    Join Date
    Sep 2014
    Location
    California
    Posts
    1,781
    Addressed some issues, stole some Sservis language, & overhauled into a Q&A format.
    Last edited by TankSinatra; 1 Week Ago at 06:02 PM.
    Level 10 Adult War Clan METAMUCIL.
    And some of us are old enough that its not just a funny word...
    Metamucil keeps the red out of our logs. Small wars, great record, low pressure.
    War Record - https://cocp.it/clan/82GJJV2J

  9. #29
    JFSoul's Avatar
    Join Date
    Apr 2014
    Location
    United States
    Posts
    6,534
    Quote Originally Posted by TankSinatra View Post
    Addressed some issues & overhauled into a Q&A format.
    This is GREAT. Bookmarked as a reference. Thanks for doing this, Tank.
    Start Dec '12 | Hi Lg/Lv: Legend / 182
    Accounts:
    TH11Hi(720 Stars)|TH9Max(1335)|TH9Max(1325)|TH11Lo(1020)|TH10Max (770)|TH10Lo(520)|2TH9Hi (300+)
    "The two most powerful warriors are patience and time." Leo Tolstoy





  10. #30
    Junior Member
    Join Date
    Jul 2015
    Location
    USA
    Posts
    77
    Quote Originally Posted by TankSinatra View Post
    1.SuperCell please look at my match, consider my proposal, or fix the matchmaker!


    10. Obviously SC needs to switch to matching by TH only...

    Strict townhall matching cannot work. The war pool is huge, but would need to be an order of magnitude larger to allow exact matching by TH. Running uncommon lineups (odd TH spreads, defenseless or th3 ringers) can lead to strange matching in your favor or against you, long searches, or even no match at all.

    11. Why not add weight to the townhall?

    Because the color of the townhall does nothing in and of itself to affect the course of the war. Weight should be (and is) put on things that actually make a difference - camps, spells, troops, defenses. Perhaps those weights should be changed, but there's no benefit to adding weight to a cosmetic item.
    I don’t think anyone would suggest “strict th matching,”but dismissing a TH as simple cosmetics is absurd. One of the biggest flags in war seems to be the TH itself as I’ve mentioned in several posts. The cosmetic appearance lower in of a TH11 ranked low in war typically tells us, without even looking, that we have a low defense TH with ridiculously high offensive strength. Thus, it makes a lot of sense to assign more weight simply based on the TH level. I haven’t seen a single matchup that throws an 11 in for aesthetic appeal.

    Our clan tag appears below. I’d welcome anyone to look at our matches to see exactly how bad the system is.
    Last edited by Rebecca00; 1 Week Ago at 09:23 PM.
    Proud coleader of Ancient Hunters (#82R2G08U), a US based, adult war clan

Similar Threads

  1. created unranked matchmaking and ranked matchmaking battle type
    By chaoser in forum Ideas & Feature Requests
    Replies: 1
    Last Post: June 13th, 2017, 01:34 PM
  2. Replies: 13
    Last Post: October 16th, 2016, 09:20 AM
  3. Replies: 0
    Last Post: June 7th, 2016, 04:22 PM
  4. Matchmaking Theory: Smart Matchmaking
    By CashConClasher in forum General
    Replies: 0
    Last Post: December 9th, 2015, 02:54 AM
  5. Matchmaking, matchmaking, matchmaking. /quit
    By Sporky in forum General Discussion
    Replies: 35
    Last Post: October 27th, 2015, 04:00 AM

Posting Permissions

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