I meant to post this a couple weeks ago, but I guess I didn’t. 
Well, anyway…I’m no authority or expert, but I do want to expand on a thought experiment I’ve hinted at in the past.
He’s right, and probably significantly more nuanced than the quote above. The Mendelian genetic inheritance models, as applied to cannabis breeding, have WAAAAAAAAAAAAAY too many assumptions to be accurate, so the progressive numbers of genetic inheritance you see in Punnett squares and cubing models are best case, not real world. The reality is that every successive generation means selection inaccuracy is compounded, and quite a bit.
The problem is those inheritance models assume perfect observational selection accuracy, which reliably works for selecting color in a simple genome like pea plants, but not for something more complex unless you have much better tools of selection than simple observation. The reason for this is, as chimera said (or implied), in cannabis there are linked traits and a whole spectrum of incompletely dominant traits, rather than the black and white dominance assumed of all plants from Mendel’s pea examples.
In a more complex trait observed and selected from a much more complex genome, we have no way of knowing how good the selection is without the tools of genetic testing and analysis, and that really only works well in two scenarios: 1) either extremely high numbers of plants (thousands) observed with scientifically objective measures applied to differentiate the nuances we can’t sense via observation; or 2) with genetic testing of similarly large numbers of plants after the cannabis genome has been discovered, open sourced, and independently evaluated.
Without one of those two scenarios being true, almost everyone’s observational selection is probably wildly inaccurate, even for just a single trait such as smell. This is because something that seems simple isn’t, and it’s very much biased by the observer. What makes up weed smells? The complex interplay of many, many terps and compounds. How do you know how close you got to the exact same smell in your selection? You don’t. It’s imperfect cuz there aren’t many crosses or plants that have a single overriding monoterpene, and even those have dozens of other terps acting as modifiers. Not to mention there’s a wide variation in the sensitivity and accuracy of people’s noses!
And that’s just smell. Trying to nail down a reliable effect in your bx? Fuhgeddaboudit.
Combine that with the curveball that terroir throws into the mix, and you can basically say that all these seemingly authoritative posts in the canna diaspora about cubing and back/in/out/sidecrossing are all just hand waving. If someone represents themselves as an authority on the subject, run away as fast as you can.
And don’t even get me started on what the addition of epigenetic expression does to the weed bx/cubing theories, especially in the uncontrolled and highly variable conditions within a typical home grow. Bx is a fool’s gambit.
Yet, despite all that, with today’s high end polyhybrids, you’re likely to get decent plants from almost ANY cross…and how many people have the integrity to cheerlead a certain breeder and then report their honest experience of less-than-stellar results at the end? Very, very few.
