Besides Hussar's point - that this means there are parts of the game that many purchasers won't use - there is also the complexity issue - PCs can begin at level 1, but you are encouraged to begin at level 4 - "Why am I starting halfway down the number line?"My issue is with systems that leave no room for "zero to hero" style of play.
<snip>
I'd like the game to accommodate both levels of play through the level system. In the game I made (and run), the average settled adult is hit die 4. That means that if I start the PCs at hit die 1, they're well below average, and truly worse than most guards at fighting, most scouts at scouting, etc. However, the advice I would give to avoid this is simple: start them at hit die 4, or even higher. This gives the players and GM the option of starting at "zero" or "average settled adult" or "very experienced" or "heroic" or whatever.
Neither of these may be an issue for a hardcore RPGer like most of those posting here, but I think they are issues for a commercial RPG publishing company like WotC.
This approach requires the abandonment of strong simulation in mechanics - "1st level" or "3 HD" no longer has a concrete meaning within the context of the fiction, but is instead a purely mechanical state which has to be correlated with the fiction on a case-by-case basis (per campaign, per encounter, even per creature perhaps what with minions, elites and solos hanging around).I think what I like best is a game where level 1 PCs can be either zeroes or heroes, dependent only on how things are skinned (demographics of NPC/monster stats), so for an heroic game the PCs start relatively stronger compared to the setting baseline, but the PC stats themselves are identical.
I find that 4e D&D is good for this; 1st level PCs can be relatively weak novice adventurers, or distinctly heroic, depending only on how the opposition and friendly NPCs are statted. Eg in an heroic-from-start game the average human or hobgoblin soldier could be a 3rd level minion (eg the hobgoblin grunt in MM), the 1st level PCs fight squads of guardsmen to rescue the princess from the evil baron. In a novice-at-start game the average human or hobgoblin soldier could be a 3rd level standard monster (eg the human guard or hobgoblin soldier in MM), the 1st level PCs fight goblin brigands to rescue the baker's daughter from their boss, a hobgoblin soldier.
IME 3e and earlier didn't work so well for this approach (NPC power determines PC status) because 1st level PCs were so fragile, no matter how weak the opponents, and because the power gradient was so steep.
Not everyone is going to like that.
This actually came up in my 4e game recently - the PCs ended up in a fight with a wizard much more powerful than them. Initially I had envisaged the encounter happening in 3 or 4 levels time, and had tenatively statted the wizard up as a 19th level elite. When it became clear that the fight might occur much earlier, I restatted him as a 13th level solo. In the course of the fight, I had to explain to the players that it was clear to the PCs that he was a very powerful wizard, who could obliterate any one or two of them with ease, but was being stretched to his resources trying to hold off all five of them at once. And given his 500-odd hit points, I was also having to narrate even large amounts of damage from the sorcerer and ranger as mighty blows that he nevertheless parried with his staff, though showing some strain from the effort . . . until the final blow, when I described him as worn down and unable to defend himself any longer.
Of course I'm not saying this is a deal-breaker - quite the opposite, given I'm a big 4e fan. But it's a serious issue from the point of view both of design and GM advice. (4e is good on the design, but weaker on the GM advice in my view.)