My 2 cents: There's a bit of confusion here about autobias and self bias. Self Bias has been around for decades and has been used by many manufacturers over the years.
Autobias is a modern varient of Fixed bias. Fixed bias is the correct way of saying 'adjustable bias' since what is really going on is when one says 'Fixed' is one is referring to the grid voltage being 'fixed' at a particular voltage- often by means of a potentiometer, and now more recently by a modern servo circuit, which might be computor or analog monitored.
Ideally, the bias in an autobias system should be monitored constantly as tubes can drift substantially upon warmup. Computor (or more aptly, controller) based systems are in some ways more adequate to this task as it is easier to prevent the resulting servo circuit from 'hunting' back and forth across the ideal bias point. In an analog-based system, the timing constants have to be chosen to lag slightly behind the tubes and the voltages measured so that stability in the system can be maintained.
Note that as a result in an autobias system the system will rarely be at the ideal bias point but will always be close by. In practice, many users are not attentive enough to monitor the bias level of their equipment (proper use of test equipment being a major limiation on convinience) so often the fixed bias system may be further off at any given time.
So the result is while an autobias system is probably not going to be exact, it probably will be closer in day to day life then a fixed bias system.
Of course, a self-biased setup will not always be correct either, as tube variations usually mean that the self biased circuit will be off by the variation in that tube. But it will always be off by that variation and none other. So if component values are carefully chosen (for example- matched tubes) a good compromise can be reached.
Ultimately: the autobias system is more likely to be the closest to correct from day to day of the three systems currently in use for push-pull amplifiers. The exception is in SE amplifiers, where the tube is operated self bias and class A. In this special case where no other power tubes are involved, the bias level is very accurate.
Sorry for the long diatribe...