Page 1 of 3 123 LastLast
Results 1 to 10 of 26

Thread: The bottom of the CWL data mine - Part IV - start times miscellaneous

  1. #1

    Join Date
    Sep 2014
    Location
    Digging another tunnel in the data mine
    Posts
    6,588

    The bottom of the CWL data mine - Part IV - start times miscellaneous

    Disclaimer: this is NOT official data; all the data is from sampling using the API, based on pulling war results pretty much randomly off the system. It was quite a large sample, so the data should be representative, unless of course I messed up my calculations

    CWL data index: 1: Stars and winning margins | 2: Attacks and 3* rate | 3: Average league rosters

    Following my previous posts on CWL data, I've got a few other random bits to post. And first of all, which is probably of interest with the new season about to search, is start times...

    Start times

    These charts show how many clans started CWL (ie matched, not started searching) each hour. The X axis is the number of complete hours after signups opened. And the Y is the number of clans matching in that hour. So "0" is the clans matching between the start and 59:59 minutes. The last, "48+" bar is after signups close, when the system matches all the remaining clans.
    I've done this for a few leagues:







    The higher the bar, the shorter search times are likely to be. There's 200+ clans kept in the search pool, and they seem to be matched at random. So if 40 clans start searching in your league in 1 hour, 40 will be matched, so you would only have a 20% chance of being one of them - with an average search time of 5 hours.

    The scale of the charts varies between leagues. Some of those bars in Gold are many thousands of clans.

    You can see the surge at the end in higher leagues when the system matches all the (200+) clans remaining in the search pool. In the lower leagues it doesn't show as there are more clans than that matching every hour anyway.

    In all leagues there seems to be a huge rush right at the start to sign up. But in the Gold chart it looks like there is some sort of server load limiting in the first few hours as the number of clans starting per hour is almost constant for a few hours. (And the same happens for all lower leagues). So it looks like maybe SC make some clans wait so that they don't have too big a rush of clans starting and finishing wars all within a few minutes, and they spread it out over a few hours.

    So it seems like far more clans search on day 1, especially at the start, and search times then will be much faster (down to a minute or two). BUT, perhaps, if you sign up as soon as the CWL gate opens, you might be made to wait a bit.

    In Champs 1 it looks like they just kept everyone until the end and then matched them. Ch1 will be much larger this season, so we don't know if that will happen again, or if some will get early matches.

    Roster size
    Most of the rest of the charts are not hugely useful, but there are curiousities. Next up the size of the overall weekly roster. Firstly a chart of the average per league:



    Higher clans generally have a bigger roster - no big surprise there. And it looks like most clans don't split up, since they are typically running near 30 on average.

    I also have some frequency breakdowns for roster sizes for a couple of leagues:




    No surprise that there's a bit of a spike at 15 - either clans that didn't understand and entered 15, or were scraping together every player they could find to get 15. (And probably the latter as it becomes more pronounced in lower leagues).
    What baffled me though was the big spikes at every multiple of 5. It looks like maybe there's a big chunk of clans out there who haven't yet (in season 4) figured out that they can enter any number, and might be leaving players out to round down to a multiple of 5!?

    Clan level
    Next up a look at clan levels... These are the distributions of clan levels in various leagues...








    And the average clan level in each league:




    You can see a few sandbaggers - level 1 clans popping up in crystal 1 - but the numbers are very low, and it doesn't really show in other leagues.
    There's a very smooth progression overall with higher level clans being in higher leagues.
    The data for Champ 1 is less of a smooth distribution just because it's a small sample. All of the others are many times larger, so the bigger data set gives a smoother distribution.

    League sizes

    I've avoided stating how big the leagues are (apart from champ 1 which is an oddity), but here's a comparison of the size:


    NOTE: the lowest leagues, especially bronze, are probably much larger. There's a problem with finding these when sampling the data, so for those leagues the sizes here are probably nonsense.

    Although champ 1 gets a lot of attention on the forum, because of the leaderboard, and it's the top league it's worth remembering that it's TINY. Most clans are in gold and silver, and to a lesser extent crystal. Even Master league is several times smaller than the lower ones.

    Seasonal balance change

    And to finish one final pair of charts to look at how balance within leagues is changing. In CWL "matching" threads people often say that they will balance out over time. It's not really been the experience with my clans - they've been very up and down.

    Here's the average winning margin in each league, in the previous season:


    And the same thing for the January season (I put this chart in an earlier thread)

    (in this one the Champ 1 figure was boosted by big wins against banned clans. It may have been around 2 without the bans)

    The concept is that high winning margins are a sign of mismatched, blow-out wars. And... they have come down slightly (by up to 0.5), although the difference is quite small.

    And that is all folks!
    Various data posts: Who plays what? CWL data 1 (stars), 2 (3* rates), 3 (rosters), 4 (start times and other) Data at end of old Legends And (non-data) how max bases are ordered in war
    --------------------
    Forum cup organiser. The forum cup discord server is: https://discord.gg/KEUxV2e

  2. #2
    Pro Member
    Join Date
    Jun 2013
    Location
    Germany
    Posts
    705
    Thank you for all the work you put into that

  3. #3
    Great work Onyx, thanks for all your efforts.
    Never Give Up Without Even Trying!

  4. #4
    Super Member
    Join Date
    Sep 2014
    Posts
    818
    Very interesting analysis good job!

  5. #5
    Forum Veteran
    Join Date
    Feb 2015
    Posts
    1,854
    As usual, simply awesome.

  6. #6
    Millennial Club
    Join Date
    Oct 2014
    Posts
    1,216
    You have an interest observation on the roster sizes spiking in multiples of 5. I thought most co-leaders/leaders would not try to round out their roster to multiples of 5. They should know that they are limited to 15, so the roster size from 15-50 doesn't have to be in multiples of 5.
    Last edited by move4ward; January 31st, 2019 at 12:16 PM.
    Quote Originally Posted by Ajax View Post
    Well don't play it then. The game is all about having fun. If you don't find that aspect of it to be fun, then don't play it.
    Quote Originally Posted by Ajax View Post
    What evidence I have seen suggests that most people are finding it fun. But obviously, not everybody will.

  7. #7
    Millennial Club
    Join Date
    May 2015
    Posts
    1,034
    Very interesting Onyx. Thanks for putting your time into this and posting for all of us to see!


  8. #8
    Forum Veteran
    Join Date
    Dec 2013
    Posts
    1,825
    Quote Originally Posted by OnyxDS View Post
    I've avoided stating how big the leagues are (apart from champ 1 which is an oddity), but here's a comparison of the size:


    NOTE: the lowest leagues, especially bronze, are probably much larger. There's a problem with finding these when sampling the data, so for those leagues the sizes here are probably nonsense.
    Ah, the chart I've been waiting for.

    Based on some rough guesses about how big those bars are compared to each other, and my spreadsheet-fu, I calculate that all of champs will have around 18000 clans in Dec 2019. Right now, the top 18000 clans reach down somewhere into the middle of masters 2. This is actually a slower rate of promotion/growth than I had expected.

  9. #9
    Forum Veteran Shadrach777's Avatar
    Join Date
    Aug 2016
    Posts
    1,759
    Quote Originally Posted by littledoctor View Post
    Ah, the chart I've been waiting for.

    Based on some rough guesses about how big those bars are compared to each other, and my spreadsheet-fu, I calculate that all of champs will have around 18000 clans in Dec 2019. Right now, the top 18000 clans reach down somewhere into the middle of masters 2. This is actually a slower rate of promotion/growth than I had expected.
    I also find this chart the most interesting, and it was as I expected with the greater distribution of clans being below Master. I was kind of hoping Crystal would have been bigger. It's interesting to see the bulk is within silver and gold. Due to the 50% nature of clan movement, the overall progression should cause Master and Champion to fill up with those from Crystal. It gives me cause for concern, because the competition is pretty even in Master 2 and up, but the unevenness from Crystal will inevitably spill over into Master, thus forcing clans in Master to move up into Champion. We will be back to uneven matches across the leagues. I expect Titan league to get released probably towards the end of the year, but it will require changing the demotion policy to bottom 3 otherwise the problem will become cyclical. Of course they could wait until Legend league gets released before an aggressive demotion policy is in place.

    We see all TH12 in Champion 1 currently, but unless new leagues get released, I doubt by Christmas that will still be the case.
    Location: USA
    TH12: 60/60/30 TH12: 60/60/30 TH12: 50/52/30 TH10: 30/30
    Main #LL2JCQ0C

  10. #10
    Forum Veteran
    Join Date
    Dec 2017
    Location
    Great White North
    Posts
    1,823
    Quote Originally Posted by Shadrach777 View Post
    I also find this chart the most interesting, and it was as I expected with the greater distribution of clans being below Master. I was kind of hoping Crystal would have been bigger. It's interesting to see the bulk is within silver and gold. Due to the 50% nature of clan movement, the overall progression should cause Master and Champion to fill up with those from Crystal. It gives me cause for concern, because the competition is pretty even in Master 2 and up, but the unevenness from Crystal will inevitably spill over into Master, thus forcing clans in Master to move up into Champion. We will be back to uneven matches across the leagues. I expect Titan league to get released probably towards the end of the year, but it will require changing the demotion policy to bottom 3 otherwise the problem will become cyclical. Of course they could wait until Legend league gets released before an aggressive demotion policy is in place.

    We see all TH12 in Champion 1 currently, but unless new leagues get released, I doubt by Christmas that will still be the case.
    I see what you are describing but I dont necessarily see it as a bad thing. As expected, the higher tiers show a norrow win gap. I would posit that this is due to several factors, base progression becomes more uniform, offensive and defensive skill take on a larger role and tolerance for non attacking or throwing in the towel are less likely.

    As you move down in tiers all these factors change in the opposite direction, again as expected.

    I do see a the same shift from masters into champs that you describe but we have to remember that there will be a similar shift from champs 3 to champs 2 and from champs 2 to champs 1. All that will do is compress the top 5 levels (master 2 through to champs1) into the three champs levels. That should change the win gap upwards but given that the gap is fairly tight today, I dont see the star gap going from 3 stars to 12. Likely more like going from 3 to 5.

    It wont actually change the distribution much IMO, it will simply shift the values slightly to the right and level the star gap a teeny bit across the entire bar chart.

Posting Permissions

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