Reference solution in JS is too slow. Please pick a faster one, like mine for instance.

What counts as "struck"? This is not explained.

• when a monster group takes damage
• when a monster group takes sufficient damage in order to have at least 1 monster killed in action

Conflict in spec. It's not clear whether we need to wait for 3 strucks, or the first + 3 next strucks (= 4 strucks).

• If the lightning runs out of targets, or has struck 3 targets already, it stops ..
• .. casts the spell on the enemy creature of their choosing, and it then jumps to the three nearest creatures ..

The same linear solution takes 11886ms in Javascript and 6s in Python. Is it intended ?

Add ruby to issues below (-> random tests are too random).

Link to the picture in the description is dead.

In Python (2 and 3), I timeout even if I just return an empty list without any computing.
Am I missing something?

You're scraping the data from CW user page and not from CW API, which means the data will be a few hours behind the actual data, some values will be inconsistent from values obtained from the API (yes, some values do IIRC), and when CW main site is choking there's no way to run the kata without timing out.

Python and Ruby version has different expected values for creation ratio:

honor ratio
Python: ['myjinxin2015', 'jhoffner', 'A.Partridge', 'g964', 'donaldsebleung', 'GiacomoSorbi', 'SteffenVogel_79', 'Voile', 'smile67', 'Unnamed']
Ruby:   ["myjinxin2015", "jhoffner", "A.Partridge", "g964", "donaldsebleung", "GiacomoSorbi", "SteffenVogel_79", "Voile", "smile67", "Unnamed"]

creation ratio
Python: ['jhoffner', 'g964', 'A.Partridge', 'GiacomoSorbi', 'myjinxin2015', 'donaldsebleung', 'SteffenVogel_79', 'Unnamed', 'Voile', 'smile67']
Ruby:   ["jhoffner", "g964", "A.Partridge", "GiacomoSorbi", "myjinxin2015", "donaldsebleung", "SteffenVogel_79", "smile67", "Unnamed", "Voile"]

Honor ratio and creation ratio is only partially tested: why only test sort_by_... methods and read off the final names when you're already checking the existence of honor_ratio and creation_ratio? As far as the test is currently being at only their relative values matter.

Also Ruby, actual tests:

expected: [..., "\u00E1\u008A\u00A0\u00E1\u008B\u0088\u00E1\u0088\u008D \u00E1\u008A\u00A5\u00E1\u0088\u00B8\u00E1\u0089\u00B1"]
actual: [..., "\u12A0\u12C8\u120D \u12A5\u1238\u1271"]

You need to specify encoding when using nokogiri or the HTML will be parsed as ascii and not utf-8.