Let me expand on what EasyRhino wrote.
It is possible that in actuality the shoe is at -8, and that when you watch it go up by +3, you're still at -5 and you'll lose. But that's no different than if you were playing a new shoe with 25% penetration.
Graphically, let's label high cards "-" and low cards "+", with neutral cards as "o" and unseen cards as X. I'll put vertical bars for reference at 25%, 50%, and 75% penetration.
If you start from a new shoe, the shoe might look like this:
(a) A normal shoe
--o-o-+-o- | o--oo--oo+ | oo++o+-+oo | XXXXXXXXXX
The count would drop to -5 in the first 1/4, to -8 by the halfway point, and back up to -5 by the shuffle (the X's at the end, which you'll never see.
Now, let's say you came in halfway through. You haven't seen the first half of the shoe, so they're all X'd.
(b) Same shoe, but you didn't see the first half
XXXXXXXXXX | XXXXXXXXXX | oo++o+-+oo | XXXXXXXXXX
Your count is at +3 from where you started, but you can't tell where you started. Here is the logical leap: this situation is no different from this scenario, which is basically 25% penetration. All unseen cards are treated the same!
(c) Equivalent to scenario (b) - you only see 25% of the cards.
oo++o+-+oo | XXXXXXXXXX | XXXXXXXXXX | XXXXXXXXXX
Now, just because you have 25% penetration doesn't mean that the game is impossible to beat, just very hard. You can beat it in the exact same manner - wait for a true count of +2 and jump in. The only problem is that true counts of +2 will be extraordinarily rare with 25% penetration!
Hopefully, you're not lost, because it's about to get a lot more complicated. If you're Wonging in at +2, flat betting, and Wonging out at +2, your EV is completely independent of your penetration! This is because EV is dollars won per dollars bet - with 25% penetration, you will win a lot less, but you will bet a lot less as well. Your EV will remain the same.
HOWEVER, your win rate will not be the same. If you're playing with 75% penetration, you will see +2 counts more frequently and bet more frequently. You will be winning money at a faster rate than if you were playing with 25% penetration.
The bottom line is pretty simple: when backcounting, you want to start backcounting as close to the shuffle as possible. The first thing you should notice is how many decks have been played - you're looking for less than 1/2 a deck played from the top, or less than 1 deck until the shuffle.
This is why the best places to Wong are where you have many tables of equally good games, and the worst places to Wong are where you have few tables. If you have a lot of tables, statistically one of them has just shuffled, and you can start backcounting immediately. If you have a few tables, you'll spend most of your time waiting for a shuffle or playing with, essentially, 25%-50% penetration.