When a resource is locked in transaction A and transaction B waits for the lock to be released, it's a waiting, not a deadlock. A deadlock would accur if transaction B also waited for a resource locked in transaction A, therefore they would be waiting for each other and if the DB server didn't intervene, they would be blocked forever. It's a different scenario.
There is no concept of "duplicated transaction", as far as I know. What Bharani Preetham Peraka describes is a behaviour of a concurrent record update when using optimistic concurrency control. An update and a transaction are two different things - a single transaction can manipulate thousand records in many tables, and you can't say that two transactions are duplicates just because they happen to update a single common record. By the way, RecVersion is indeed a property of record, not transaction.
The primary reason for using transactions is making operations atomic. It means that either the whole operation (e.g. an update of a hundred related records) is performed, or it's not done at all. You won't end up with inconsistent data because the operation failed in the middle.