Common transcription methods of Japanese into English (like Hepburn etc.) combine transcription and transliteration approaches - i.e. sticking in many cases to transposing Kana transliteration into English - producing misleading transcriptions of Japanese words.

In my view:

• To an expert, it doesn't really matter much which transcription or transliteration system is used, as he knows "proper" pronunciation and how to decipher (in most cases) the resultant transcribed English words.
• To a laymen, it matters a lot as generally even serious fans of Sumō and Japanese culture are misguided as to proper reading of anglicized Japanese.
• English, for this use, is an especially problematic instrument, as English orthography is very deep - resulting in very problematic guidance as to how the words are pronounced - unlike, for instance, Japanese kana orthography, which is very shallow and thus good.

One of the general criticisms of using phonemic transcription is, for instance with not transcribing "えい" as "ei", that the transcription cannot be used as a reverse transliteration tool then - words cannot be easily transliterated back to Japanese kana by a layman.

Such a point is completely moot in my opinion - as an expert will be able to transliterate back without an issue generally. And a layperson has no need for a reverse transliteration. What's most important that the layman will be guided properly towards a most proper pronunciation, as is possible within the target orthography.

Thus, a general transcription of Japanese to English should:

• Be an orthographic transcription - and not a transliteration - namely for instance not transcribing Japanese "えい" as "ei" as this leads to a complete mangling of the words (thus no "sensei", "keitai", and the like), not transcribing "おう" as "ou" and so on.
• Be phonemic - meaning using English phonemes, matching as closely as possible the proper Japanese phonemes, so as to provide the closest possible pronunciation guide to the layman - thus not transcribing "おう" as "o" or "oh" and so.
• The maximum possible phonemic orthograpic transcription will then be as close to the real pronunciation of Japanese - and not lead to mangling the Japanese words in common English usage.

Therefore, concretely, 相撲:

• Has proper Tōkyō (not Tokyo, being then improperly pronounced /toukyou/ as is common) pronunciation of IPA: ['sɯ̥mo:]
• Is affected by two factors: vowel devoicing and stress both on the "す".
• Due to devoicing the "Su" looses, all but insignificantly, the "u".
• The stress slightly accentuates and extends the "S".
• The "も" is a proper long "mo:", not short "mo".

Thus:

• Proper phonemic orthographic transcription to English becomes "S'mō", leading to a very close pronunciation in English.
• Seeing this form, even the lay person can reasonably guess the proper pronunciation.
• Not "Sumo", leading to wrong, yet common, articulation as /suumou/, /suumo/, /sumo/, and the like.

So, I'm using the following general rules to transcribe, providing in my opinion a very close approximation to the spoken Japanese in transcribed English:

• Devoiced vowels generally dropped, when devoicing significant.
• Stressed consonants generally elongated by "'".
• Vowel hiatus is also marked, where necessary, by "'".

Therefore, the transcription of S'mō phraseology:

Notes and references:

[ https://en.wikipedia.org/wiki/Orthographic_transcription ]
[ https://en.wikipedia.org/wiki/Orthographic_depth ]
[ https://en.wikipedia.org/wiki/Stress_(linguistics) ]
[ https://www.speechactive.com/word-stress-in-the-international-phonetic-alphabet-ipa/ ]
[ https://cjki.org/reference/japphon.htm ]
[ https://en.wikipedia.org/wiki/Japanese_phonology ]