Page 1 of 4 123 ... LastLast
Results 1 to 10 of 32

Thread: Delvings in the CWL data mine II - Attacks and 3* rate

  1. #1

    Join Date
    Sep 2014
    Location
    Digging another tunnel in the data mine
    Posts
    6,442

    Delvings in the CWL data mine II - Attacks and 3* rate

    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 | 4: Start times and other

    Following on from my previous data thread, I've been running some 3* rate calculations. These are mostly interesting at the higher leagues where the bases are more consistent. In all of this I can see the TH level of the attacker and defender, but nothing else - so I cannot determine if a base is maxed, weak, outright engineered or whatever. (In theory the player profile could be obtained from the API, giving hero levels which would give a hint... but my code would need to pull one player at a time, and it's just not worthwhile)

    Champs league
    These are mostly TH12s, so I've run all attacks without considering TH level initially. However I've looked at a couple of other factors. Firstly there are "abnormal" wars. I've arbitrarily defined this as any war where
    • EITHER both sides scored 40+ (there is no doubt that all of these were win-traded with FWA bases; there is a HUGE gap in results between these and the real wars, and so no ambiguity in the middle). The 3* rate for these is meaningless, of course (and usually near 100%)
    • OR the winning margin was over 12*


    The latter is because it's not clear what was going on in these wars, and whether real bases and attacks were used. One clan consistently got hammered by scores like 37-18. I've no idea whether they ended up with a light roster or were tanking or something else. But it's possible they didn't put out real warbases; and possible they were looting. Likewise once the octopus struck, in some of the affected clans the survivors put up results like 40-5 in the last war or two. It's likely the CC were empty on their defence; and they may have been doing loot attacks. So all these wars are suspect. In my first data post I put a chart of the winning margins - and the frequency tailed off to near zero at around 7-10 stars with everything higher looking like outliers.

    The other factor is clans suspected of cheating (listed below as "banned" clans). This list is taken from the Champions War League ban list, PLUS clans that got hit hard by the octopus during this CWL season.

    So, what everyone's been asking for, the strike rates for Champ 1:
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    All 0.6% 7.6% 79.7% 12.0% 2.03 100.00%
    Normal wars 0.6% 7.6% 81.0% 10.8%
    2.02 94.29%
    Banned clans 0.4% 8.1% 76.0% 15.5%
    2.07 8.89%
    Non-banned clans
    0.6% 7.6% 81.5% 10.3% 2.02 85.40%

    The first line is every attack I can see. The second excludes all attacks, from both sides, in abnormal wars - and these are also excluded from the last 2 lines. Lines 3 and 4 split these attacks by whether the attacking clan is suspected of cheating. Effectively the last line is the headline one - attacks by clans that aren't obviously cheating in wars that look to have sensible results, so where both clans probably tried.

    The Avg column is the average number of stars earned. And the final column is how much data is on that line (as a % of all attacks eligible for the table - which is all attacks in the league here)

    I did check the TH split, and 99.9% of all these attacks are 12v12; but as mentioned above, they I can't confirm if they are maxed - but anecdotally the number of non-maxed bases is probably negligible.

    Comments: the headline figure is just over 10%, as mentioned previously. It's quite a bit higher in the cheating clans, although interestingly they also have a very slightly higher 1* rate (script failures?)

    Same again for Champ 2:
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    All 1.0% 11.0% 78.4% 9.7% 1.97 100.00%
    Normal wars 0.9% 10.8% 79.5% 8.7% 1.96 96.91%
    Banned clans 0.6% 10.2% 74.5% 14.6% 2.03 1.50%
    Non-banned clans 0.9% 10.8% 79.6% 8.6% 1.96 95.41%

    In this 99.5% are 12v12s. The rest split between 11v12s and 12v11s (not all successful!). Stripping these out would reduce the 3* rate by about 0.2%.

    Comments: the 3* rate is noticeably lower. It's not clear if all bases are maxed, but probably are close in everything that matters. Presumably the higher rate at Ch1 is skill-based: those clans got promoted to Ch1 because they attacked better!

    And last like this for Champ 3: (caution, not all 12v12s)
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    All 1.6% 14.2% 72.2% 12.0% 1.95 100.00%
    Normal wars 1.5% 14.0% 74.0% 10.5% 1.94 94.10%
    Banned clans 0.9% 10.3% 72.3% 16.6%
    2.04 0.18%
    Non-banned clans 1.5% 14.0% 74.0% 10.5% 1.94 93.92%

    The 3* rate starts to go up again, presumably because of weaker, non-maxed bases. In fact although the number of TH11s is small, it's enough to be significant. Here's the "Normal wars" line broken down by TH:
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    12v12 1.4% 13.9% 75.4% 9.3% 1.93 95.25%
    11v11 2.4% 13.8% 48.6% 35.2% 2.17 0.13%
    12v11 0.3% 4.6% 26.1% 69.0% 2.64 2.31%
    11v12 4.7% 26.3% 68.3% 0.7% 1.65 2.24%
    There were some TH10s found, but attacks involving a TH10 totalled only 0.04% of the data, so I've taken those lines out. I haven't split out results for prossibly cheating clans here or in lower leagues as the number of such clans is now so small it's not really significant.
    (Note: in this table abnormal wars are excluded entirely; the final column is now the % of attacks only within 'normal' wars)

    Comments: Once TH11s are involved it will begin to skew the TH12 rates - hard TH12s will be hit by TH11s and only the easier ones attempted by TH12s; and also clans may use anti-2 TH12s in wars with plenty of TH11s. However at this level the number of TH11 is so low (many clans had none) that this effect is probably still weak, and the 3* rate reflects TH12s attacking each other generally. But, of course, they are probably not all maxed now.

    Champ 2 and 3 probably reflect better the strike rates in wars generally than Champ 1 where there is obviously skill selection happening. So these headline rates of around 9% may be the best measure of TH12 difficulty. However, the rate in non-CWL wars may be higher because there is less need to bank the 2*, since another attacker can clean up disasters.

    Masters league
    I've pulled the data for M1-M3 as well, but only the version with TH splits now, as all these have significant numbers of non-TH12. All of these still filter to wars with normal results, and in these leagues this is beginning to filter out a significant number of wars with a 12+ winning margin (the breakdown of winning margins for M1 was in my earlier post). Some of these are likely to be genuine, but lopsided wars. But some may have been clans tanking. And in very lopsided wars it's not clear if both clans will take it seriously. So restricting the data to wars that aren't too lopsided seems better.

    Master 1:
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    12v12 2.2% 17.7% 68.9% 11.2% 1.89 74.03%
    11v11 1.8% 13.1% 49.4% 35.7% 2.19 4.19%
    10v10 3.3% 17.5% 41.6% 37.6% 2.14 0.04%
    12v11 0.5% 6.3% 33.6% 59.6% 2.52 10.47%
    11v10 0.5% 3.1% 14.9% 81.6% 2.78 0.42%
    11v12 5.9% 28.1% 63.4% 2.5%
    1.63 9.90%
    10v12 16.9% 59.6% 22.9% 0.6% 1.07 0.33%
    10v11 8.1% 33.1% 47.2% 11.5% 1.62 0.28%

    Master 2:
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    12v12 2.8% 19.9% 65.1% 12.2% 1.87 49.34%
    11v11 2.1% 13.2% 50.2% 34.6% 2.17 17.78%
    10v10 2.5% 14.8% 47.6% 35.0% 2.15 0.81%
    12v11 0.9% 8.4% 40.7% 50.0% 2.40 13.46%
    11v10 0.5% 3.9% 19.7% 75.9% 2.71 2.79%
    11v12 6.6% 28.1% 59.3% 6.1% 1.65 11.80%
    10v12 17.5% 52.2% 28.8% 1.5% 1.14 0.68%
    10v11 7.8% 25.2% 49.3% 17.8% 1.77 2.30%

    Master 3:
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    12v12 3.6% 21.3% 61.7% 13.5% 1.85 33.42%
    11v11 2.5% 13.7% 50.3% 33.5% 2.15 23.77%
    10v10 2.8% 14.0% 47.0% 36.2% 2.17 4.80%
    12v11 1.5% 10.0% 44.2% 44.2% 2.31 12.00%
    11v10 0.7% 5.4% 24.5% 69.4% 2.63 7.01%
    11v12 7.2% 28.0% 54.6% 10.2% 1.68 9.24%
    10v12 18.8% 51.3% 26.9% 3.0% 1.14 0.80%
    10v11 6.8% 21.3% 46.3% 25.5% 1.91 5.78%
    (12v10 attacks or anything involving TH9 weren't tabulated; the absence doesn't mean that there was necessarily zero of them)

    Comments: Going down the leagues the 3* rate for 12v12 continues to increase. This may be because TH11s are hitting some of the hardest TH12s. But I suspect the biggest difference is that the bases are weaker. You can see this from the fact that the TH11v12 triple rate in M3 exceeds the 12v12 rate in Champs 1 - there were obviously some weak bases to be picked off!

    The 1* rate also continues to rise as we go down. This may be due to worse attackers. But I think will also be because bases become more anti-2. And that's another reason the 3* rate goes up.

    It's interesting that the hit down 3* rates are not exactly enormous. 12v11 is only averaging around 50% in these. The chances are it's weaker TH12s hitting TH11s, but even a couple of troop upgrades is a significant advantage. It is what I see in regular wars however: 12v11 is not exactly a gimme for casual clans.
    I'm surprised the 11v10 rate is this low, however. Although my clan's TH10s do sometimes defend TH11s, it's often the case that even in bad attacks there's simply too much on the attacking side. But this rate is only hovering around 75%

    I'm also surprised the 11v11 and 10v10 rates are almost the same. BUT there may be a selection effect at TH10. The number of attacks here involving TH10s is very small. Clans entering a mostly 12+11 roster with 1 or 2 TH10s on the bottom might be picking their best TH10 attackers from a larger group for inclusion. Whereas all TH11s in the clan are getting in regardless.
    Last edited by OnyxDS; February 1st, 2019 at 11:23 PM.
    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

  2. #2

    Join Date
    Sep 2014
    Location
    Digging another tunnel in the data mine
    Posts
    6,442
    Crystal League
    Same again for C1-C3, but I've included TH9s now...
    Crystal 1
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    12v12 4.6% 22.6% 56.3% 16.5%
    1.85 21.37%
    11v11 3.1% 14.6% 50.2% 32.1% 2.11 21.58%
    10v10 3.2% 14.4% 45.8% 36.5% 2.16 11.79%
    9v9 3.9% 18.2% 35.1% 42.8% 2.17 0.89%
    12v11 2.1% 11.4% 44.1% 42.4% 2.27 10.69%
    11v10 1.1% 6.9% 28.4% 63.7% 2.55 10.30%
    10v9 0.8% 3.7% 12.3% 83.1% 2.78 2.43%
    11v12 8.1% 27.8% 49.3% 14.8% 1.71 6.91%
    10v12 21.6% 45.8% 26.8% 5.8% 1.17 0.73%
    10v11 6.9% 20.0% 44.0% 29.0% 1.95 7.54%
    9v11 12.8% 26.4% 26.5% 34.3% 1.82 0.96%
    9v10 15.4% 29.8% 42.6% 12.2% 1.52 1.41%

    Crystal 2
    Filter 0* 1* 2* 3*
    Avg % of all attacks
    12v12 6.0% 23.8% 49.1% 21.1% 1.85 13.07%
    11v11 4.1% 16.4% 48.5% 30.9% 2.06 16.17%
    10v10 3.9% 15.8% 45.3% 35.0% 2.11 16.88%
    9v9 4.5% 18.8% 34.2% 42.5% 2.15 3.77%
    12v11 2.9% 12.8% 41.9% 42.3% 2.24 9.08%
    11v10 1.6% 8.8% 31.5% 58.1% 2.46 11.46%
    10v9 1.2% 5.7% 16.6% 76.5% 2.68 5.48%
    11v12 9.2% 27.9% 43.5% 19.4% 1.73 5.06%
    10v12 22.5% 40.0% 28.4% 9.0% 1.24 0.69%
    10v11 8.2% 20.9% 43.1% 27.7% 1.90 7.03%
    9v11 11.5% 22.8% 24.3% 41.4% 1.96 1.42%
    9v10 14.7% 29.1% 40.2% 16.0% 1.58 3.06%

    Crystal 3
    12v12 7.6% 24.7% 41.3% 26.4%
    1.86 7.72%
    11v11 5.6% 18.2% 45.3% 30.8% 2.01 10.62%
    10v10 5.0% 18.0% 44.2% 32.7%
    2.05 17.49%
    9v9 4.9% 19.9% 33.1% 42.1% 2.12 9.02%
    12v11 4.1% 14.7% 38.7% 42.5% 2.20 6.84%
    11v10 2.4% 11.1% 32.6% 54.0% 2.38 10.93%
    10v9 1.8% 8.3% 20.8% 69.1% 2.57 8.51%
    11v12 10.8% 28.2% 37.8% 23.3% 1.74 3.57%
    10v12 22.8% 35.9% 29.3% 12.0% 1.30 0.71%
    10v11 10.1% 23.2% 41.5% 25.1% 1.82 5.58%
    9v11 12.5% 22.6% 23.6% 41.2% 1.94 1.46%
    9v10 14.3% 29.3% 37.6% 18.8% 1.61 4.23%

    (I haven't tried to plot any 12v10 or 11v9 attacks, nor anything involving TH8 or under. There were probably some - so the last column should add up to a little under 100%)

    Note that these still only include wars with a winning margin of under 12*. In C3 this may be under half of wars. It's useful as a reflection of what happens in sensible attacks, but it's worth being aware that the experience of CWL is beginning to drift further from these tables alone now.

    The 26% 12v12 3* rate in C3 may seem quite high, but when you consider that 10v12 has a 12% triple rate, then perhaps not! My clan's TH10 are in C3 and although they see the odd max TH12, there are more of them that are really quite shocking.

    Conclusions
    For TH12 we have a pretty clean data set at Ch1 and Ch2 of big boy TH12s playing proper wars. And the 3* rate overall is around 9% - rising to just over 10% in the better clans (Ch1)

    For TH11 and under, although there is plenty of data there are limits to what it can tell us about balance, because many of the bases may be weak or rushed. Although the 3* rate on both these seems to consistently be around 30%-35%, in those leagues the 12v12 rate is higher than in Champs; and there are significant numbers of 11v12 and 10v11 triples, clearing showing the vulnerability of some bases.

    It does however show us something about player experience - most players aren't playing with all max bases, so this is what most people see in war. It also puts something of an upper limit on the normal 3* rate. Typically clans would use a hit-up to 2* a tough enemy, and use same TH attacks on the weaker ones. So if anything when people are hitting bases equal to their own the 3* rate is likely to be lower that above.

    It's interesting that after the inferno nerf, a few minor defence nerfs, and the huge benefit of having rams, that TH10s still usually don't 3* each other! And people objected to the inferno nerf alone, claiming it would break TH10!

    Missed attacks
    One other thing I counted was the number of attacks used by clans in wars. I considered including something in the tables above, but when a TH11 misses an attack does that go in the 11v11 or 11v10 lines? So, I've made some separate charts for these. These charts are the number of attacks used in each clan's war: with the number of attacks along the bottom and the height of the bar being how frequently clans used that many attacks.

    Starting from the top:

    (In Ch1 some of the missed attacks were caused when people were touched by the octopus - probably most values under 14 were outliers caused by this or some other major event)






    It's scary just how chaotic it gets below Crystal league! And it's worth pointing out that around Gold league is what most people experience. Master leagues seemed to start much smaller than the lower leagues. Although a lot of discussion here focuses on Champ leagues, they are tiny compared to the lower leagues.
    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

  3. #3
    Forum Elder
    Join Date
    Sep 2016
    Posts
    2,114
    Great data compilation! Really interesting read.

  4. #4
    Forum Elder JtB's Avatar
    Join Date
    Jan 2013
    Posts
    2,888
    Nice data! Wasn’t someone from Supercell saying the 3 star rate on town hall 12’s was about 30%?

  5. #5
    Quote Originally Posted by JtB View Post
    Nice data! Wasnít someone from Supercell saying the 3 star rate on town hall 12ís was about 30%?
    I would like to mention a fact that CWL for th12s is more of a 2star game. 3star rate in CWL doesn't explain the normal war scenario. Spam attacks are same whether normal wars or CWL, high % 2stars. But to 3star at higher ths via planned attacks, scouts are very important. Troll Tesla, any surprise Tesla that screwed pathing of loons, or traps, type of cc troops, etc so many factors to ruin a fresh attack, so fresh triples are very very difficult at higher ths. And since CWL is all about fresh attacks, so it is really more of a 2star game.
    Last edited by BlazeStormz123; January 21st, 2019 at 04:49 PM.
    Never Give Up Without Even Trying!

  6. #6
    Pro Member
    Join Date
    Apr 2016
    Location
    Still searching for the hog rider in BB...
    Posts
    628
    Great work!

    What surprises me most is the 3* rates for 9vs11, over 40% in Crystal 2 and 3 and still over 30% at Crystal 1! The TH9s seem better than TH10.

    Any ideas why this is so? Could it be the remaining engineers that have stuck at TH9 with great troops and skills? I'd be interested...
    Last edited by BWBX; January 21st, 2019 at 05:02 PM.

  7. #7
    Forum Veteran Shadrach777's Avatar
    Join Date
    Aug 2016
    Posts
    1,759
    Quote Originally Posted by BWBX View Post
    Great work!

    What surprises me most is the 3* rates for 9vs11, over 40% in Crystal 2 and 3 and still over 30% at Crystal 1! The TH9s seem better than TH10.

    Any ideas why this is so? Could it be the remaining engineers that have stuck at TH9 with great troops and skills? I'd be interested...
    Nothing worth speculating since the 9s seem to have a harder time hitting the 10s versus hitting the 11s.
    Location: USA
    TH12: 60/60/30 TH12: 60/60/30 TH12: 50/52/30 TH10: 30/30
    Main #LL2JCQ0C

  8. #8
    Millennial Club NightsWatchMB's Avatar
    Join Date
    Aug 2017
    Location
    #boycott-the-blimp
    Posts
    1,055
    Quote Originally Posted by BWBX View Post
    Great work!

    What surprises me most is the 3* rates for 9vs11, over 40% in Crystal 2 and 3 and still over 30% at Crystal 1! The TH9s seem better than TH10.

    Any ideas why this is so? Could it be the remaining engineers that have stuck at TH9 with great troops and skills? I'd be interested...
    More likely in our experience the engineered TH11s that are either super rushed (level 1 everything) or still just have the one cannon

  9. #9
    Centennial Club
    Join Date
    Jun 2018
    Posts
    190
    Awesome job as usual Onyx. Thanks for sharing.

    I mentioned this in your other thread, but given that you published this Data, I am pretty sure that you know have what I need: would you please share the common rooster distribution (i.e., how many TH12, TH11, etc. on average) per different leagues? Please include sub divisions (e.g, Crystal II) as well.

    Thanks,

  10. #10
    Forum Elder RagnarFurias's Avatar
    Join Date
    Sep 2016
    Location
    Brazil
    Posts
    2,324
    Amazing data. So we can say an average TH12 attacker will be able to triple a 12 in 1 CWL war and 2-star the other six. Itís a good parameter to see how one is doing overall, compared to other players. Nice work, thanks for sharing!
    Last edited by RagnarFurias; January 21st, 2019 at 05:41 PM.
    Nothing here stays the same forever. You must learn to change as the game changes, or clash will force you one way or another.

Posting Permissions

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