Clan War Rating, imo. It is wonky but it weights quality of wins (kind of like Elo or QM rating) so it's far better than just counting wins (like territory ranking does) or looking at win%. If you have Muli's userscript and use the Search button on the Multi-Player Dashboard, on the Clans tab it ranks by Clan War Rating. Of course, from a game design perspective, just ranking clans by Clan War Rating (for the purpose of rewards) would change the incentive structure of Clan Wars and probably hamper participation. But if you're just trying to capture a predictive snapshot of which clans are performing well in Clan Wars, then Clan War Rating is the best I can think of.
Here's the top 20 by that:
1. Python - 335.12
2. ONE! - 251.75
3. Lu Fredd - 183.52
4. {101st} - 178.32
5. MASTER Clan - 156.23
6. M'Hunters - 148.19
7. VIW - 143.74
8. FCC - 142.13
9. TLMS - 138.69
10. Avengers - 126.1
11. HAWKS - 112
12. [WG] - 111.93
13. |GG| - 110.16
14. French Community - 103.82
15. [BR] FUDEW - 88.53
16. TSFH - 85.91
17. Vikinger - 81.21
18. Hodopian Dynasty - 75.94
19. ズＫＩＮＧ - 75.64
20. German Warlords 75.62
You'll notice this is wonky, because matchmaking plays a big part in the quality of ratings. TLMS and FCC have ratings that are very close to each other but I suspect they haven't been playing the same quality of opponent. One way to tell whether a rating+matchmaking system yields accurate ratings over time is to see if it converges to 50% win rate for as many clans as possible (basically everyone except for those near the top). Hopefully the reason as to why is intuitive: if your rating system is perfect and your matchmaking system pairs players (or in this case, clans) of similar skill, then everyone but the people at the top and bottom should consistently face- on average- opponents of their own skill level and win about half the time.
I suspect this will not be true for the CW rating system, since the rating system is tuned weird and the matchmaking algorithm does not seem particularly sophisticated. Normally these would be fine but when you have very few matchmaking/rating events (4/day) you start seeing the quirks shine through. Throw in the template selection part of this (e.g., the differentiation in skill between clans that go for mostly SE vs. clans that go for mostly MME) and you really have a Clan War that's multiple sub-competitions where clans like EXCEL can do really well by simply not having to go up against clans like MASTER or FCC.
Anyhow, the fact that there is at least a rating system and a matchmaking algorithm does mean that win-rate over time will still be a noisy, probably useless indicator. So Clan War Rating is the best I think we've got, but we should just take it with a lot of salt.
