I don't know....you could also look at this as if we've been in a negative phase overall since 2000 give or take. In the grand scheme, 90 years is a pretty small sample size wrt a decadal oscillation, or even smaller sample size as you are suggesting a tri-decadal oscillation. That would only provide roughly 3 measurement periods since 1930.
note that is summer pdo I am using^^ not intentionally cherry picking I promise