There is periodic confusion over which system is "best" or most profitable. To some extent this depends on conditions and spread. But if you define the question properly then there is an answer.
Quite simply, the best system is the one that wins the most money for a given level of risk. This is equivalent to maximizing Sharpe Ratio (or Desireability Index in Don Schlesinger's _Blackjack Attack_). To measure this properly you should use optimal betting; otherwise you penalize some systems in favor of others. For example, the Hi-Opt I system has slightly lower volatility than High-Low. Consequently betting proportional to truecount would give higher return and higher risk using High-Low. You need to take the ratio of return/risk to get a good measure. Again, you want the betting to be optimized for the system.
We can categorize systems by single parameter or multi-parameter (side counts). Within single parameter systems we have ace-neutral, ace-compromised, and ace-normal. We have balanced and unbalanced. And we have level 1, 2, and 3.
While there have been multi-parameter systems based on medium cards, the overwhelming issue at blackjack is counting the ace. Basically the ace should count negative for betting, but positive for playing. You can add a ace-side-count to any system, so this is really a separate feature. I'll proceed to discuss only single-parameter counts.
You can reasonably measure the performance of blackjack systems in terms of betting correlation and playing efficiency. At one time there were deeply dealt single deck games where playing efficiency was important. But betting correlation is most important in modern multi-deck games. In this respect ace-normal is clearly superior to ace-neutral (without side counts). The optimal ace-tag for betting is roughly -1, whereas the optimal value for playing is roughly 0. Here is the Zen insight. At the optimal tag value, the first-order effect of tag adjustments is zero, and the tag error is a second-order effect. In other word by using a compromised ace-tag of -.5 you will get roughly 75% of the betting and playing benefits, not a mere 50%.
The discussion above omitted insurance correlation, which is similar to playing efficiency (ace-tag ideally equals zero). But for optimal insurance correlation you don't want to count the "9".
If you true-count a system then it doesn't really matter whether it is balanced or unbalanced. Any count will allow you to use a running count at the pivot. For example, the High-Low system has a balanced pivot of 0. When the running count is zero then you have a neutral-deck-edge, regardless of penetration. In contrast the KO system has a pivot +4. When you are at the KO pivot then you have a large advantage regardless of penetration. If you have problems with calculating truecount then High-Low will be more accurate in neutral decks and KO will be more accurate in very positive decks.
See www.qfit.com/card-counting.htm for system tag definitions. The most powerful system would be ace-compromised without counting the "9". Some systems are Level 1 (plus-minus). These systems can't be ace-compromised. High-Low is Level 1 ace-normal balanced, KO is Level 1 ace-normal unbalanced, and Red-Seven is Level 1 with an intermediate pivot. Zen is Level 2 ace-compromised balanced, Unbalanced Zen is Level 2 ace-compromised unbalanced. There is little need to go beyond Level 2. However, sometimes "unbalancing" tags allows a little fiddling to make tags more accurate without going to a higher level. Zen {-1,1,1,2,2,2,2,1,0,0,-2} doesn't count the "3" or "5" quite high enough. We could try to get more exact betting with a balanced Level 5 monstrosity like {-3,2,3,4,5,4,3,2,0,-5}. But the Unbalanced Zen just improves the "3" tag with {-1,1,2,2,2,2,1,0,0,-2}. And I coined the Adjusted KO by improving the "5" tag {-1,1,2,2,3,2,1,0,0,-2}. Simulations have shown this is slightly more profitable in shoe games. Technically the Adjusted KO is Level 3, but it isn't really much harder or much better. The important difference is where you want your pivot.
Summary: If you use an ace-compromise then you get roughly 75% of the benefit of an ace-side count. Systems shouldn't count the "9". You should decide whether you want the running count "pivot" to be most accurate in a neutral deck, a slightly profitable deck, or a really profitable deck. Then pick your system:
Level 1 balanced - High Low,
Level 1 small pivot - Red Seven,
Level 1 big pivot - KO,
Level 2 balanced - Zen,
Level 2 small pivot - Unbalanced Zen,
Level 2 big pivot - Adjusted KO.
