Page 5 of 9 FirstFirst ... 34567 ... LastLast
Results 41 to 50 of 85

Thread: Deeper in the CWL data mine - part III - average league rosters

  1. #41
    Forum Veteran
    Join Date
    Dec 2013
    Posts
    1,559
    Quote Originally Posted by dmoore1998 View Post
    Even just a single chart of TH12 distribution in something like M3 would probably show enough to be meaningful on some level, since I suspect people who care about "lopsidedness" primarily care about the top and not the bottom.
    Though they should. What has kept us afloat in M3 is having strong defense - no th10s or lower on the war map.

    It would be easier if we could intuitively grasp 5-dimensional shapes in our minds.


    Quote Originally Posted by dmoore1998 View Post
    Totally agree with you about the difference a TH12 makes. I've been in 2 clans in the last 2 CWLs, both in M3. Both talking how hard the previous season in M3 was. Both added a couple of TH12s and crushed opponents. That's a big part of what makes me think there's probably a pretty narrow distribution of TH12s in something like M3..adding even 1-2 seem to push a clan from average to significantly dominant.
    Assuming that they keep their strongest, adding 2 TH12s also means dropping, perhaps, 2 TH10s. That's a big offense and defense boost. (Less so if they added TH12s by having a couple 11s rush to 12.)

  2. #42
    Millennial Club
    Join Date
    Nov 2018
    Posts
    1,296
    Quote Originally Posted by littledoctor View Post
    Though they should. What has kept us afloat in M3 is having strong defense - no th10s or lower on the war map.

    It would be easier if we could intuitively grasp 5-dimensional shapes in our minds.




    Assuming that they keep their strongest, adding 2 TH12s also means dropping, perhaps, 2 TH10s. That's a big offense and defense boost. (Less so if they added TH12s by having a couple 11s rush to 12.)

    Definitely, i cut off part of Onyx's post since I only planned to address the first half. His point on the extra 12s was specifically about recruiting, so yes theoretically dropping your low bases as opposed to the minor progression of an already-included player simply upping the TH.

  3. #43
    Senior Member
    Join Date
    Feb 2015
    Posts
    324
    Quote Originally Posted by OnyxDS View Post
    I did wonder about this, but wasn't sure how to display it. It would be hard to display the distribution of the entire roster at once. A distribution of number-of-TH12s could be done like some of the charts in my earlier posts, and likewise for TH11 etc. But then, there's no way of knowing how they relate. Are the clans with more TH12s also having more TH11, but no TH10 at all, or vice versa?
    I would also like to see this. A distribution of the number of 12s similar to your earlier posts would work. I don’t care about the relationship so much. If you wanted to know the relationship then you could use a conditional probability distribution.

    Something like this...
    Given a roster in M3 with 7 TH12s what is the TH11 distribution? TH10 distribution.
    Last edited by Easyluck; January 24th, 2019 at 11:45 PM.

  4. #44
    New Bloke/Chick
    Join Date
    Feb 2014
    Posts
    43
    Error bars on the charts could graphically display the distribution of each average. Bars could represent standard deviation or a 95% confidence interval

  5. #45
    Senior Member
    Join Date
    Oct 2017
    Posts
    435
    Excellent work for me, @OnyxDS
    You're the be
    st forumer as I see.

  6. #46
    Quote Originally Posted by frainlll View Post
    Excellent work for me, @OnyxDS
    You're the be
    st forumer as I see.
    Agreed...without this data everyone has to assume their individual cwl experience is similar to others. This allows us to see an accurate view of the entire cwl performances and matchups rather then listen to the one guy on the forum who happened to be in the one cwl grouping that had more favorable results.

    I am not sure why sc doesnt provide data like this..I would assume and definitely hope they use data like this to make balancing decisions for the game. Keeping this data hidden doesnt benefit them in any way. If anything I would argue that if sc released official data similar to Onyx to the public and give us their overall idea of where they want the three% rate to be, it would allow us as a community to either accept it or argue a valid reason for changing it.

  7. #47

    Join Date
    Sep 2014
    Location
    Digging another tunnel in the data mine
    Posts
    6,442
    I did have a quick look at distributions, but I want to try and see if I can improve the charts a little before I upload anything. But - they did mostly look like the normal curves you'd expect.

    Quote Originally Posted by Chrys View Post
    Error bars on the charts could graphically display the distribution of each average. Bars could represent standard deviation or a 95% confidence interval
    Sadly the graphing API I'm using doesn't do error bars in a stacked bar chart, so I'm a bit stuck there.

    Quote Originally Posted by CaptainMayhem View Post
    I am not sure why sc doesnt provide data like this..I would assume and definitely hope they use data like this to make balancing decisions for the game. Keeping this data hidden doesnt benefit them in any way.
    Well, they're not exactly keeping it secret - they've made it all available through the API. You could argue this is the kind of thing that it's intended the community do. Likewise with the sort of stats that warmatch produces for the player leagues. I've no idea what SC think of this data on here, but they've not complained yet and they know where to find me if there's something they don't like. So I assume they're reasonably happy.

    The problem with CWL is that the API is rather quirky (and that's putting it mildly). So some things we want are rather hard to extract. But that's a matter of omission rather than secretiveness.

    If anything I would argue that if sc released official data similar to Onyx to the public and give us their overall idea of where they want the three% rate to be, it would allow us as a community to either accept it or argue a valid reason for changing it.
    This I agree with in principle. There have been balance discussions on here for years, and it's been frustrating that SC have never said what the 3* rate is, or what they think it should be.

    Having said that... having done this data, I'm starting to appreciate that "3* rate" can mean a lot of different things. On here a lot of people are running with the "10%" headline figure from my other thread. But it can be argued that that's both too high or too low. It was for Ch1. But in Ch2, also with max bases, it was dropping nearer to 8%. So I expect going lower still in terms of player skill, it gets even lower on max bases. And Ch1 is tiny - many of the other leagues are 100-1000 times larger. So it could be argued that the max 12 triple rate is quite a bit lower than 10%, excluding the tiny number of players in Ch1.

    But going the other way, the actual triple rate I saw in lower leagues actually went up quite a lot. In fact, in some leagues, the 11v12 rate exceeds 10% IIRC. So the literal 12v12 3* rate across the entire game could well be above 20%. Of course there's a lot of beating up of rushed or weak bases in there (and just god-awful layouts). But this may be where SC's focus is - on the millions of players, not the couple of thousand in Ch1.

    So it is a bit of a minefield.
    For info on CWL see: CWL survival guide and data posts: 1 (stars), 2 (3* rates), 3 (rosters), 4 (start times and other)
    --------------------
    Forum cup organiser. The forum cup discord server is: https://discord.gg/KEUxV2e

  8. #48
    Junior Member DavieC82's Avatar
    Join Date
    Oct 2017
    Location
    Behind you.
    Posts
    78
    Onyx, I tip my hat to you. Informative yet simple to digest. Kudos for the time and effort put into this.
    Player Tag #JU2PYOGY

    TH12, enjoying the grind once more.

    Remember, it's just a game! Don't worry, be happy.

  9. #49
    Junior Member
    Join Date
    Aug 2017
    Posts
    56
    One thing I noticed was that in almost all cases, promoted distributions were far better than relegated distributions from the above league. Ie. Crystal 3 up to Crystal 2.

    A question from a mere mortal like myself, is this likely to change? Will there be a point in time where the relegated distribution will be better then the promoted distribution from the league below?
    Barb Empire
    Level 10 Clan looking for new members of any ability! We’re a friendly bunch, so if you need a clan, then come and have a look!

    #CR8J09UJ

  10. #50
    Millennial Club
    Join Date
    Nov 2018
    Posts
    1,296
    Quote Originally Posted by OnyxDS View Post
    This I agree with in principle. There have been balance discussions on here for years, and it's been frustrating that SC have never said what the 3* rate is, or what they think it should be.

    Having said that... having done this data, I'm starting to appreciate that "3* rate" can mean a lot of different things. On here a lot of people are running with the "10%" headline figure from my other thread. But it can be argued that that's both too high or too low. It was for Ch1. But in Ch2, also with max bases, it was dropping nearer to 8%. So I expect going lower still in terms of player skill, it gets even lower on max bases. And Ch1 is tiny - many of the other leagues are 100-1000 times larger. So it could be argued that the max 12 triple rate is quite a bit lower than 10%, excluding the tiny number of players in Ch1.

    But going the other way, the actual triple rate I saw in lower leagues actually went up quite a lot. In fact, in some leagues, the 11v12 rate exceeds 10% IIRC. So the literal 12v12 3* rate across the entire game could well be above 20%. Of course there's a lot of beating up of rushed or weak bases in there (and just god-awful layouts). But this may be where SC's focus is - on the millions of players, not the couple of thousand in Ch1.

    So it is a bit of a minefield.
    To your point on 3* rate meaning a lot of different things, totally agree. If I were looking at something like C1, I would expect to see a lot of strategy going on. That means I'd expect to see far more base layouts which offered attackers a choice...an easy 2 while trying to limit damage %...or a risky 3 that if failed would likely result in 1.

    You're talking a lot more about strategic decisions of whether to even attempt 3*, knowing that a 3* attempt that fails is somewhat likely to cost your clan the war. I'd expect it to be higher in regular wars where you not only have the advantage of watching replays later...but you also have a HUGE advantage in that if you go for a 3* and fail...you can always have someone else take the easy 2 later (or take the easy 2 first and try for 3 later with no real consequences).

    It's not just about the 3* rate, it's about the rate compared to how often you TRY to 3*. Part of what I think makes CWL more interesting than regular wars is the strategic portion of it. Every attack counts more. I have to look at a base and decide ("Ok, the best way to 3* would be to come from this direction using this army...but if it doesn't work, I'm probably not getting the TH at all...where are we at in war, what is the benefit of getting the 3 vs the cost of failing and only getting 1?").

    The goal of CWL isn't to get as many 3* attacks in as you can, it's to win wars, and then win the league.

    So it COULD be that the 3* rate of people trying for 3* is the same in C1 as it is anywhere else...(yes better defenses and layouts, but also better attackers), but that folks in C1 recognize the cost/benefit of going for 3* (i.e. if a failure on a 3* attempt results in 1 star half the time, and a 2 star half the time...you'd need a 3* rate of 40% to make it worthwile...40% of the time getting 3, 30% getting 2, 30% getting 1...an average of 2.1, compared to just getting a sure 2). The penalty for failing a 3* attempt in lower leagues, with bases that probably aren't designed to really put the attacker to that question probably isn't as much...so they are free to try for 3 with better results should they fail. It's relevant what the results are if a 3* attack FAILS in determining how often people will even legitimately TRY to 3* someone.

Posting Permissions

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