Well that is my point, it is kind of both. Typically your decisions in blackjack are determined by one metric, expected value. So a basic strategy chart, for example, shows which play yields the highest expected value for any given holding. So strictly from the perspective of expected value the play is as Wong describes, namely to split 88 v T if the true count is +7 or lower and to begin to stay at TC +8 or higher. But if you look right on the margin, at say TC 7.5, then you see that the expectation of splitting and staying are very close however splitting incurs significantly more variance. This is a big deal because a winning gambler is at war with variance. After all, if you have the edge, you will win in the long run but short term fluctuations (variance) can beat you. This phenomenon is so pronounced that one would be willing to accept a significant decrease in winrate if you were to simply win at a fixed rate as opposed to suffering the swings entailed with counting cards. We would rather have say $60 / hr guaranteed than a $100 / hr win rate in your typical count game if somehow we were to simply win at a rate rate of 60 cents / hand.
I would definitely stay on 88 v T at +8 and I would definitely split 88 v T @ +5. I think for TC 6 and TC 7 some strong cases can be made for either staying or splitting and I am generally apathetic between choosing to split or to stay.