"Take care instead of what happened in the previous shoes at the same location." This bothers me. I've done quite a bit of shuffle tracking back in my BJ days. Okay, there's not a player cut in Baccarat anymore which would change the shoe order from one shoe to the next but . . .
Have you actually tracked live shoes by, say, half deck to see how the composition of low/high cards extended from one same color shoe to the next (1,3,5,7)?
How do you back up this claim?
Imo it's quite difficult to get a perfect shuffle of 8 decks either made manually or by CSM.
We know that even one single card burnt or changing position will alter entirely the BP results. Yet the probability to get a high card falling here or there remains, but we might think that per every shoe dealt the number of 8s and 9s (for example) won't be equal on either side.
The same about every other cards class.
It's not news that the game "war" (fundamentally a high card game) is perfectly beatable if any card is removed from the deck, no cards are burnt between hands and the deck is played almost entirely. And actually by now casinos use a CSM and burn a lot of cards (mainly for other reasons).
Baccarat is a more complex version of war but the principle remains the same and we can choose whenever we want which side to bet on.
The side having the two initial cards forming the higher point are largely favorite to win the hand.
Thus we have two opposite forces acting along the way: a natural very slight propensity to get a kind of "chopping" mood and the actual card distribution that tends to deny it as cards are clustered in some way by an imperfect shuffle (thus endorsing the streaks' formation).
Imo the trick is to ascertain when the first force overwhelme the second and vice versa. This could be done by the help of general probabilities and by the actual card distribution.
It's a kind of trend following not solely in terms of actual hands but in terms of cards falling here and there.
If we bet very few hands we could have a better picture of what is happening.