The biggest local-language SEO arbitrage is in section 23

There’s a section of the Search Quality Rater Guidelines written directly for raters outside the anglophone world. Section 23.0 says, almost with a wink: “we know that you can read English (you are reading this document!), but you should only give high Needs Met ratings to English results if users in your locale would expect or want them for a particular query.”

And it completes the thought: an English result most of the locale can’t read should be considered useless — Fails to Meet, the bottom of the satisfaction scale. The examples are categorical: the English Wikipedia page on Barack Obama, with equivalent content to the local-language version, is useless to someone who can’t read English. Latin-script queries change nothing: [ronaldo] searched in Hindi deserves a Hindi result.

The strategic reading for anyone operating in a non-English market:

• Your real competition on a local-language query is only what exists in that language. The world’s best English content isn’t contesting the SERP with you — to the rater (and to the algorithm trained on this scale), it isn’t even a valid result.
• A niche where the best information lives in English = a low-competition SERP with proven demand. The well-made native version doesn’t compete with the foreign source; it fills a vacuum.
• The exceptions are narrow and useful: official sites of global organizations, technical terms the community doesn’t translate (commands, code, function names — the right model is local-language explanation with the terms in the original), and locales where English is a current language.

Thoughtfully translating-and-adapting what only exists in English isn’t a shortcut — it’s literally what Google’s scale demands on behalf of the local user.

Full analysis in the annotated guide: Location, language and the dictionary problem.