Yeah I need to update that first post. Forgot about it tbh. When I get on the computer I will.

As hinted at before, I wanted to check the match-ups of the finished games to see if some clans had more advantage from feasting on weak clans so far.
Meet the ASOS metric for this. ASOS=adjusted strength of schedule, I defined this as the win % of opponents in games against the other teams. If your ASOS is high, it means your opponents won often against other clans, if your ASOS is low, it means your opponents lost often against other clans. In this case: play often against Sninja or TJC for a high ASOS, play often against FCC for a low ASOS. Because of the balanced nature of the competition, all teams will end Clan League with an ASOS of 50%. An ASOS of 55% is considered a difficult schedule, a 60% ASOS is a very difficult one, an ASOS of 45% is considered a weak schedule, while an ASOS of 40% is considered very weak.

Sninja:           2 TJC, 2 WG, 1 Stats, 2 Hydra, 4 FCC, ASOS: 40.55% (games), 39.46% (points)
   TJC: 2 Sninja,        3 WG, 4 Stats, 5 Hydra, 3 FCC, ASOS: 53.48% (games), 52.00% (points)
  [WG]: 2 Sninja, 3 TJC,       3 Stats, 5 Hydra, 5 FCC, ASOS: 42.61% (games), 39.07% (points)
 Stats: 1 Sninja, 4 TJC, 3 WG,          2 Hydra, 2 FCC, ASOS: 51.90% (games), 54.46% (points)
 Hydra: 2 Sninja, 5 TJC, 5 WG, 2 Stats,          3 FCC, ASOS: 52.90% (games), 55.63% (points)
   FCC: 4 Sninja, 3 TJC, 5 WG, 2 Stats, 3 Hydra,        ASOS: 53.74% (games), 54.71% (points)

Before I started, I thought this might not say much actually, but to my surprise, it actually does tell us a lot! You can clearly see that Sninja and WG have been feasting on the weaker teams a lot more! For both clans, more than half of their games have been against Hydra and FCC. For [WG], more than half of their points are from wins against FCC even! (20 out of 39). This gives them a very weak strength of schedule so far. Meanwhile, the other clans have a very similar schedule, all just a little higher than expected to compensate. If you look at this, TJC looks a lot stronger than their record indicates, and [WG] actually looks weaker. Their ([WG]) good team record? Mainly FCC feasting...

I went one step further and used this to (very) naively predict the remaining games/points. I wanted to get a first idea if this actually means that TJC has a reasonable chance to still tie Sninja.
Method (skip this if you don't like the juicy math details) I did this in a Bayesian way by giving every team a 50% prior head start (a theoretical 10/20 points won). For every game won or lost, the number of points were added. This gives Sninja now a hypothetical 45/58=77.58% win rate rather than their real 35/38 record. This is to allow regression to the mean. In each remaining game, I pit the two teams and divide the points proportional to the hypothetical win rate. E.g. a 2v2 between Sninja and Stats (WR=51.61%) would earn Sninja 2.4 point (4 * 77.58% / (77.58%+51.61%)). This is obviously not the best way to do this, but it's the quickest to do without spending another hour writing code (which I might still do). The final results are as follows:

Sninja: 137
   TJC: 121
  [WG]: 113
 Stats: 106
 Hydra:  88
   FCC:  35

I find it noteworthy that according to this method, FCC is really behind the others at one end, while Sninja may not be as dominant as their current record indicates. Hydra is closer to Sninja than to FCC!
Of course, this methodology is very crude and there is obviously a lot of variation in the results, but I found it quite interesting nevertheless.

Very interesting read, thanks for putting it together!

One thing we can't factor in is that a lot of results out so far are from 1v1 games and due to the schedule (Three 1v1's starting now, three after midseason) there's no telling how the line ups will match up after the midseason :)

After spending 2 hours coding, I came up with a better simulation method (that is still very limited due to, among other things, what dry-clean-only just said above). The results are a little different. Note: because of the way the program is written, I can semi-manually update these simulation every so often if people are interested in that.
Details of the calculation below. (std=standard deviation, PI=prediction interval)

           mean   (std)    95% PI
1. Sninja 138.65 (13.19) [112-164]
2. TJC    116.04 (11.32) [ 93-138]
3. [WG]   108.52 (11.33) [ 86-130]
4. Stats  104.38 (12.72) [ 79-130]
5. Hydra   85.23 (11.44) [ 63-108]
6. FCC     47.17 (11.09) [ 27-70 ]

               1       2       3       4       5       6
  Sninja  83.67%  11.27%   3.74%   1.20%   0.12%   0.00% 
  TJC     10.17%  45.90%  28.47%  13.41%   2.03%   0.02% 
  WG       3.49%  24.28%  35.79%  30.02%   6.40%   0.02% 
  Stats    2.61%  17.52%  27.50%  38.90%  13.36%   0.11% 
  Hydra    0.06%   1.03%   4.49%  16.45%  76.36%   1.61% 
  FCC      0.00%   0.00%   0.01%   0.02%   1.73%  98.24% 

I personally think the simulations do a good job of matching common sense given the results we have so far and looking at the strength of schedule. The only thing I'm a little weary about, is the score of FCC. I think they may not even got to 27 points, the lower bound of the prediction interval. The reason that the simulations are quite bullish about them, is of course the 10 points from the prior. The program still gives FCC somewhat the benefit of the doubt as it does not know that more boots are still coming.




Juicy math details: in this simulation, I calculated for each clan their score on a logit scale and used this to predict the outcome of the remaining games. First, I again counted the points won and lost by each clan, again adding 10 points won and lost for each clan (burn-in prior). More specifically, I assumed 2 points won and lost against each other clan. From this result, I approximated scores on the logit axis that would reflect the results best using the optim function in R. In more detail: I minimized the squared differences between of the observed logodds calculated as log( win%A / win%B ) in A-B matchups (so win%B is 1-win%A) and the expected logodds for the scores of the 6 clans (score(clan A)-score(clanB)) for the 15 pairings.
Then I made an educated guess for the variance of these estimates where I made sure that it was proportional to the product of the number of points won (including the +10 from the start) and the number of points lost (again including this +10). (If I had more time, I could bootstrap this, but alas.) I chose the formula 10/sqrt(points won*points lost) for the standard deviation because it gives a std of 1 in the case of no information, which I found adequately wide for a non-informative prior. (A team with a score of +2 has a 99% probability to win against a team with a score of -2).
From these calculated scores and their standard deviation, for each simulation a score was drawn. The points in the remaining games were then calculated for each clan under the assumption that the true level of the clan was equal to this simulated score. The win% (for team A) was simply the result of the logodds calculation again, resulting in: sqrt(exp(score A-score B))/(1+sqrt(exp(score A-score B))). Because my input from before did not include the difference between 1v1 and teamgames, I was lazy again and simply assigned wins per 3 points remaining (with a final group of 4 or 5 as appropriate). I think this will not meaningfully influence the results though.
Finally, all these scores were summed and voila!

If anyone wants my R code to improve or add more simulations, feel free to send me a PM or ask here. Note that all CL results were manually input though.

Edited 5/21/2018 20:38:46

hmmm, I guess for stage 2 the results will deviate a lot, at-least for clans bringing in newer players to roster.

very interesting stuff Math Wolf.

Hydras got us on Szeurope (which I have never really gotten the hang of, but am forced to play all the time in CL).

I still can't believe that it made it in over Strat 2v2. Two tournaments on EU, FAIL.

We've had multiple templates on the same map many times. At the end of the day, the map itself doesn't matter - it's all about the template itself.

Edited 5/23/2018 01:21:05

Because you are bad on a template doesn't make it a fail.

It's a decent template, there aren't enough 2v2 and 3v3 templates and I think nobody would begrudge anybody from making more team templates.

I do agree that maps being used multiple times is kinda meh, but like beren said, this is nothing new (Gui and 2v2 Gui in CL9).

That being said, despite my hatred of the template, sze i am ok with, because by saying no duplicate uses of maps, you're basically saying Sze should never be played, because let's be honest, despite buns best effort, 3v3 EU is going nowhere.

If you are being forced to play Szeurope, recruit more players to play it for you. If you don't like the idea of playing different templates on the same map, You can play on templates on different maps. But to say it's a fail is beyond dumb.

Even more dumb that WG voted the template top 3 in the voting templates. Why would you vote it top 3 if you had no players that were going to play it, and complain about being forced to play it?

I didn't vote. Reza did. I was out of town.

Comprehensive Update V:

Hmm... my creative train is not working...

Summons Creativity

Well, the summon creativity enchant didn't work either.

RIP Me

I said I wouldn't do an update so late again. But here I am... after forgetting to do the update. Oops. I guess I can't keep myself to my own word... hence I've got nothing fancy to start off this opening except thinking about how tired I am... again. Well, Ill cut to the chase... I apologize if this one isn't as fancy as you want it to be.

Quick Update: By Points Scored:

TJC 			44
Sninja			43
WG			42
Statisticians		35
Hydra			25
FCC			3

Quick Update: By Max Points:

Sninja			197
WG			175
TJC			174
Statisticians		167
Hydra			160
FCC			135

Completed Games: (Bold detonates winner):

1v1: Guiroma

WG vs FCC			https://www.warzone.com/multiplayer?GameID=15662522 (5/22)
Hydra vs TJC			https://www.warzone.com/multiplayer?GameID=15662521 (5/23)


1v1: French Brawl

Statisticians vs FCC		https://www.warzone.com/multiplayer?GameID=15642524 (5/23)

1v1: Elitist Africa:

Statisticans vs Hydra		https://www.warzone.com/multiplayer?GameID=15546766 (5/24)
Sninja vs Statisticians		https://www.warzone.com/multiplayer?GameID=15478682 (5/24)
Statisticians vs FCC		https://www.warzone.com/multiplayer?GameID=15683761 (5/25)

2v2: Final Earth



2v2 Volcano Island

Statisticians vs TJC 		https://www.warzone.com/multiplayer?GameID=15580812 (5/26)

2v2 Szeurope

Hydra vs WG 			https://www.warzone.com/multiplayer?GameID=15375984 (5/21)

3v3 Europe


3v3 Biomes

TJC vs Statisticians 		https://www.warzone.com/multiplayer?GameID=15560273 (5/23)
Sninja vs Statisticians 	https://www.warzone.com/multiplayer?GameID=15389945 (5/26)

Statisticians completed quite a few games for this period, providing them some clarity in their standing and proving that they are contenders for promotion in this clan league. Hydra wins a team game, making FCC the only clan that has not won any teams games thus far. And they have not scored a single point of anything since week one. 1-14 is NOT the way to go for your first set of 1v1 games. Pair that with four 0-1 and 1 0-2 start for team games, and the ship starts sinking quickly, and perhaps a strong second stretch might not even suffice for them to overcome a floundering Hydra. That petrifying gaze has to affect someone I suppose.

Edited 5/27/2018 16:52:01

Progressions:



Horse Race:



Analysis:

Nothing is 100% in B at this point. But we can be reasonably confident that this is a race between 4 clans for two slots. (Sninja, WG, Stats, TJC). Sninja has an edge on the other three, but this is defo up for grabs. On the back in, it looks like FCC is still fast-tracked for relegation. Otherwise, not much has changed about these trajectories.

WG surpassed TJC for second place, but this could go any way. (Note that the progressions are including the two games that finished after the supposed update time).

My apologies on the late update. Ill try to be a bit more timely about this next time.

Edited 5/27/2018 17:16:23

I believe it is a testament to your tiredness that "bold detonates winners"

Updated simulations!

Details of the calculation at https://www.warzone.com/Forum/301728-clan-league-10-division-b-official-thread?Offset=63. (that's about 10 posts before this one) (std=standard deviation, PI=prediction interval)

           mean   (std)    95% PI
1. Sninja 144.87 (11.22) [120-167]  (+6)
2. Stats  112.86 (10.16) [ 93-133]  (+8)
3. TJC    109.80 (10.28) [ 90-130]  (-6)
4. [WG]   106.21 (10.70) [ 85-126]  (-2)
5. Hydra   83.68 (10.84) [ 63-105]  (-1)
6. FCC     42.58 (10.40) [ 24-64 ]  (-5)

               1       2       3       4       5       6
  Sninja  94.95%   3.87%   0.93%   0.25%   0.00%   0.00% 
  Stats    2.32%  43.29%  32.23%  19.58%   2.58%   0.00% 
  TJC      1.68%  30.88%  34.93%  28.69%   3.85%   0.00% 
  WG       1.03%  21.30%  29.34%  40.53%   7.80%   0.00% 
  Hydra    0.02%   0.66%   2.57%  10.95%  84.95%   0.85% 
  FCC      0.00%   0.00%   0.00%   0.00%   0.85%  99.15% 

I'm as surprised as anyone by these results, so I had to double-check. Turns out the algorithm rates Stats better than TJC and WG at this point, probably because Stats has a win against Sninja while TJC and WG have not. WG actually gets a razor thin edge over TJC according to the algorithm as well when it comes to underlying skill at this point, but TJC comes out on top because they already have secured a few more points. The results for 2-4 are all over the place though, so I have to agree with Aura: it's definitely up for grabs (second place).
However, Sninja is now running away with the win according to the algorithm (about 95% now against 84% last time) and has only slightly more than a single percent probability to not promote (5% last update, 8 days ago). Meanwhile FCC has even less chance than that to not relegate (probability to end fifth has halved). At this point, the algorithm thinks it's almost equally likely that Hydra promotes or relegates. (98.5% that we see Hydra in B again next year if they don't drop out of CL).

Edited 5/30/2018 20:47:12

hmmm, from what I understand your algorithm rates Stats better than TJC and WG because Stats are 3-1 vs WG and 2-4 vs TJC, TJC is 1-2 with WG. I guess though Stats have negative win rate vs TJC, because of positive win rate vs WG and TJC's negative win rate vs WG, overall it thinks Stats is suited for 2nd place. Of-course you're weighing the win over Sninja bit more , since they're the clan win the least losses, but places 2-4 can surely go anyway.

@Mega Knight: I agree that is likely the case. It's an optimization procedure, so a little black box, but if it wants to optimize the distances between the different teams, it makes sense that Stats is a little closer to Sninja than the others.

And it's already outdated, with FCC just notching their second win, again versus Hydra. Most things are about the same, but for Hydra and FCC that changes a lot:
FCC is now expected to finish with 47.33 points and has a 2.69% probability not to relegate, more than triple what it was before that game!

@Math Wolf does your algorithm only take account the clans as whole or also templates? (since many clans should be a lot stronger in specific templates due to how they have done their line-ups so if your algorithm only takes into account clans themselves as whole then clans who have had more completed games on their stronger templates could be overrated by your algorithm and vice versa and same for clans who have beat "stronger" clans templates where those "stronger" clans are weaker)

^good point. Another consideration is that each clan has some teams stronger than others. If you are comparing within templates though, that should not matter, barring a sub.

@ max & Chris: I don't take it into account explicitly, because this is simply not possible. If you read the juicy math details, the following part may be of interest though:

"From these calculated scores and their standard deviation, for each simulation a score was drawn. The points in the remaining games were then calculated for each clan under the assumption that the true level of the clan was equal to this simulated score."

Basically, based on finished games, each clan has an underlying skill score, and a measure of variation on that score. When I simulate, I draw from this distribution. So in some runs, a clan may be assumed better than what they seem from the finished games, in other runs, they may be assumed worst. Like this, I take into account that the skill level on different templates may be different and that the finished games are not completely representative (but nevertheless the best guess we have so far).