Graham Phantom Anti-Skate. Is it effective at all?

I also observe the cantilever when it drops on the record to see if it deflects on impact when it starts it's groove travels. Not a perfect method by any means but works for me with the arms I use

On impact the stylus suspension will be stretched and the elasticity of the joint will be at its maximum. Therefore this is the worst time to measure anti skate using cantilever flex as a measure.
It would appear far more logical to me to check that the cantilever is centred whilst playing. The caveat here is that many folk adjust anti skate without checking that the horizontal bearings in the tonearm are level - a few microns out can affect antiskate requirement significantly. The other trap using listening or channel balance is the possibility that one might use anti skate to compensate for other system issues.

Personally for anti-skate I am looking for accurate set up of turntable and arm first - ensure that both platter is perfectly level, check that the horizontal arm bearings are perfectly level by balancing the arm to zero and check that there is no float in or out, I check that a small tap creates the same travel distance in both directions ( with the arm balanced to zero ). As a high end dealer many years ago my experience is that many arm boards are not level when the platter is level. Furthermore in many gimbal type arms I have found that when the arm base is level, the horizontal arm bearings are not. This is why the float test is very important and will help to minimise anti skate.

Then after careful set up of the cartridge including alignment, VTA, tracking weight ( its an iterative process ) for anti-skate I check that the cantilever is centred whilst playing in the middle of a record. Using your ears before and after each adjustment is helpful.

I would expect that a unipivot arm ( like the VPI ) will have less issues with anti skate than a gimbal arm because the issue of ensuring the horizontal bearings are perfectly level does not arise, and the bearing friction should be considerably lower if the unipivot has been designed correctly.

Dougs point about the anti skate been applied to one end of the cantilever and the skating force being applied to the other is quite correct. However IF one has checked that the cantilever is straight, and not being pulled in one direction or the other whilst playing ( assuming one can find a record that is not eccentric ) then this argument has less weight. In terms of the physics then the effective mass and cartridge compliance come into the calculation. The issue here is that not many cartridge manufacturers specify the lateral compliance, and in my experience some low compliance cartridges are quite compliant in the horizontal plane and vice versa.

Furthermore if a cantilever is leaning one way or the other, then clearly the magnets or coils could be sitting in a position where there is more non linearity within the electrical/magnetic fields.

Many arms have inbuilt anti skate anyway. In the case of the VPI's the twisted loop of tonearm wire is far too stiff in my view and exerts both an anti skate and a rotating force affecting azimuth. In other arms the tonearm wire is often impeding free movement in the bearings as John pointed out.

Antiskate systems themselves can be problematical - spring, hanging weight, elastomer thread, weighted lever - they all have their pros and cons. It is a horses for courses scenario where you end up, but the process for determining the optimum ( or "none" ) anti skate should be consistent.

As an aside, whilst having a hiatus from audio, I used a Shure V15Vxmr ( with stabiliser brush removed ) for about 10 years on an ET2 linear tracker ( high horizontal mass, no skating forces ). I had added a little magnetic damping to the bearing tube motion using the eddy fields generated from the motion of the tube across the magnets. After 10 years the original cantilever was still dead straight and the cartage was sold for more than I originally paid.