In Knock-Out, we double 10 vs dealer 10 at running count of +4 or more. In this sense, splitting 10 vs dealer 10 at RC >= +20 isn't that bad an idea. Of course push at 20 is highly likely. But we have an extra chance: we may grab an ace. The dealer's hand clearly doesn't have an ace!
Splitting 10 vs dealer A requires a lower RC, the reason being that the dealer's hole card is not a 10. We have a higher chance to win.
The advice is solid, but some thoughts to ponder...
Firstly, splitting 10s will garner unwanted attention, especially if you are at max bet. Surveillance may pick up on this move pretty quick, because it is very rare that anyone would split a 20 against a dealer's 10. If you are already pegged as a good BS player, they may figure you know more than just BS to make that play. Getting barred can hurt your long-term EV.
Splitting 10s vs. 10 will increase your variance. It is fairly uncommon for the RC to get up to +20. If something unusual such as catching stiffs on both hands when the max bet is out were to occur, the player can be left with a few max bets lost. This can take quite a while to recover from. Several hours of grinding may be the result.
So, when you consider camouflage, variance, bankroll, EV of a 20 vs. 10, it is probably better from a risk-averse standpoint to stand instead of split. Especially to avoid barring.
A running count of 20 is not that big. There are still a lot of low cards out.
You analogy with doubling 10v10 is not valid. Doubling the 10 is roughly like splitting the 10s (actually, splitting the 10s ahs higher EV, because you can take another card), but the alternative play is very different. In the case, you alternative is to play a 10, in the other case it is to stand on a 20. Standing on a 20 is much more valuable than hitting a 10.
If you had true count of +52, then I would agree with you that splitting was profitable. But if I ever had a true count of +52, I would suspect I had either made a mistake, or I was being cheated. Or even both.
I do not think a running count system can detect a true count of +52, but I haven�t studied that.
In the above pots, I referred to a true count off +52, I was thinking here of +52 over "normal", which I think is 48 over pivot.
Regarding what KO in running count mode can detect: the highest RC is when all the low cards are removed, and all the big cards remain. Over a single deck, this would be IRC + 30. At such a running count, we have nothing left buts 10s, Aces, 8s, and 9s. We don�t know how many 8s and 9s are left.
Whether or not to split the 10s depends on the relative portions of the 8s, 9s, 10s, and Aces. If there were nothing but 8s and 9s left, you would not split. If there were nothing buts 10s and Aces, then splitting would have positive EV. If there only 10s left, it wouldn�t matter what you do, since you will push until cut card comes out.