Which criterion did Sklansky use when dividing starting hands in groups? Computer generated starting hands seems to show a subjective grouping by Sklansky, doesn't it?
The Sklansky hand rating is also inclusive of post flop playability and implied odds. A simply equity calculation would render T8o better then 89s. This clearly isn't the case though.
You have to consider how the hand flops relative to villain range, and it's post flop playability. Do not base hand ranking on pure jam equity because you will be ignoring too many other factors. A hand like 89s can flop flushes, straights etc, more often and better then T8o. Although preflop equity might show different, T8o is often useless postflop- even if you flop TP because you have to fold it too often. With 89s you will often have backdoor equity etc.
From the book
Math of poker of Collin Moshman:
The Sklansky-Chubukov rankings compare starting hands based on how well they perform shoving in a particular situation. Suppose you are playing heads up, and are in the small blind, and you move all-in. As you do, you accidentally turn your hand faceup. Your opponent then has the luxury of calling with precisely those hands that have you beat, or which have the odds to call. How deep can the stacks be so that pushing with your hand face-up is better than folding? (Note that another way of asking this question is which hands are unexploitable at higher stack depths than others.) In No-Limit Hold’em Theory and Practice, David Sklansky and Ed Miller answer that question in absolute terms. Hands like low offsuit aces and deuces are profitable for close to 25 big blinds, whereas 8-7s is only profitable up to eight big blinds. If you put all hands into an S-C table like that on p. 299 of No-Limit Hold’em Theory and Practice, or the ones available online, you then have a ranking of all starting hands from worst to best. This is the hand ranking used by the program SNG Wizard.
Of course, it was his experience which guided Sklansky for the grouping. So yes, he did it "subjectively" and not by some formula.
Also, he didn't have the preflop jam equity in mind, but rather the expected amount of money the hand typically wins. On page 19 of "Hold'em Poker" you find:
These rankings do not reflect how one hand will do head-up against each other. Two nines are ranked below AK, for instance, even though it is a small favorite in "one on one" play. Rather these rankings reflect which hands will win the most pots and more importantly the most money over a period of time in a full game.
Furthermore, you can read there that he assumes a full ring playing limit poker.
The list would have to be substantially altered for no limit games.
I use a "Q-T-8" standard that corresponds essentially to his Level 4 to open. That is no unsuited hands lower than KQ, no suited hands lower than JT, no pairs lower than 8. A level 3 hand would have suited hands down to KJ, QJ, and the 9-9 pair. Level 2 hands have suited hands down to Q's (AQ and KQ), and the pairs J-J and T-T. Level 1 hands have the three top pairs and the top suited hand, A-K.
This standard was based on Sklansky's principles, and is a model of how "Sklansky" is SUPPOSED to work. The problem is that Sklansky himself sometimes goes against his own model. For instance, he and I believe that KQ, KJ, and KT off suit should be levels 4, 5, and 6 hands respectively (levels 2, 3, and 4 for the suited versions). But Sklansky then "deviates" for hands with aces by saying that AT offsuit should be level 5, AJ should be level 4, and AQ should be level 3, and AK should be level 2. I actually agree with AQ and AK, but would hold Sklansky to his "original" standard for AJ (level 5) and AT (level 6).
I created the Q-T-8 model specifically to track Sklansky's deviations from his articulated standards. By design, it is "more royalist than the king, more Catholic than the pope."