That sounds like fun. We pull a lot of trees with 4000 lb tirfors on trails, where the objective is getting the rootball out of the way of new trail construction. They are easier to pull out when you've got the weight of the tree to help, you just get them to a certain point and the tree does the rest.
When I first started doing it, I would climb them and hook them high, but for a couple reasons just started going as high as I could reach from the ground. The first reason was time, some of those jobs we were pulling 100+ trees in the 24-36" range, and that is a lot of time spent getting in and out of spurs, working up through limbs, etc. The second was the risk of breaking them off, if you do that you might as well have just cut it down. Plus the time, energy, and length of pull with the tirfor put into bending the tree before the pull goes to the roots. The third was if I kept the rigging low, I could reach it all to reset under tension.
So I traded mechanical advantage from using the tree as a lever for mechanical advantage from blocks. My standard setup was a 3-1 between the tree I was pulling and one or more trees downhill, with one more block for a redirect back up the hill. I would set up the tirfor uphill, above the tree we were pulling, to get the operator out of the way. I'd put a 2-1 from the tirfor feeding into the 3-1, to make a 6-1. If needed and possible with the layout, I'd feed the tail from the 2-1 back into the system, making either a 7 or 9-1 depending on where I hooked it. That would bring over most of what we needed to pull. If it didn't, I could easily scale it up by feeding the whole thing into one more 2-1, between the 3-1 and the target tree, without changing any of the other rigging. That last 2-1 I'd step up to 1/2" wire rope and a 6" half face Skookum block, it was too much for the 3/8" and cheap 4" toggle pin blocks that I'd use for everything else.
I've never put a dynamometer on it to see what we were actually pulling with that setup. In theory the 12-1 with that tirfor would be 48000 lbs, but in practice it was quite a bit less due to friction and sometimes wide angles through the blocks. Early on we hooked the trees with a half inch choker. Then we snapped it clean off with the 12-1 and a stubborn tree. With the tirfor the tension comes up slow and quiet, so it is different than breaking a cable with a machine. The whole system started to sing as individual wires parted, it was a noise I've only heard the one time. We couldn't tell which piece of wire rope was failing, but we knew something was, so we backed off and waited. After 15 seconds or so, the choker broke and the Skookum block flew about 50' down the hill until it got tangled up and stopped. After that happened, I started hooking with chain. A piece of 3/8 grade 80 with grab hooks on both ends. I would wrap the tree once with each end, leaving a loop in the middle, and put a shackle on the loop, running free, to equalize between the two wraps. I never broke anything after that, with hundreds of trees.
It is a good system, easy to set up and scale up. It does exceed WLL (but not rated breaking strength) on some components, but it isn't overhead, and the way it is set up if something does break it isn't coming back at the operator. The 5/8 and 3/4 wire rope we'd need to stay within WLLs would be too much for the mules that haul our gear in, and much heavier and slower to handle in the brush. I'm sure we pay a cost in cycles to failure, but the average crane or logging operation is doing as many picks in a day or two as we do on a whole project, and we just keep that gear out of any overhead work.
Like Burnham said, it isn't precision work. You can never quite tell where they are going to go. A big root on the downhill side can easily push it 30 degrees off of the line of pull. While there is a lot of variation in how easily they come over, on the whole it has made me a lot more confident choosing spar trees for overhead work. Deliberately pulling them over has given me a much better idea of what a given tree will actually take before it fails.