But if you take a close look at the monsters in B2, you'll see they were really developed presuming 6-sided Hit Dice; that is, it's really made for Original D&D, pre-Supplement I (Greyhawk, which revised hits to d8 dice). Here's some examples where it's easiest to see. We'll assume our hypothesis (B2 hit dice are d6; average roll 3.5) and see if things are consistent.
- H. Bugbear Lair. HD 3+1 -> expect 11.5 hit points for a normal bugbear. Listed guard/male equivalent hit points are: 11, 12, 12, and 10, so that's very consistent. Key piece of evidence: In area #41 there's a 14 hp bugbear described as "the big bugbear"; if HD were d8 then he would be below average; but if HD are d6 then he is indeed bigger than expected.
- J. Gnoll Lair. HD 2 -> expect 7 hp. Listed males are 9, 8, 8. It's more likely that EGG would make these gnolls slightly over-average, than under-average (as they'd be if HD were d8).
- K. #57 Hall of Undead Warriors. Skeletons HD 1 -> expect 3.5, listed as 3 hp. Zombies 2 HD -> expect 7, listed as 8 hp.
- D. Goblin Lair. HD 1-1 -> expect 2.5 hp. Listed males are 3, 3, 3, 4 hp. Again, most of these would under-average if HD were d8.
Now, Gary did a couple of recognizable things when setting hit points (familiar from his later writings), like, if in doubt he'd usually set monster hit points a little bit above the expectation. Most solo "boss" monsters he sets at Max or Max-1 possible hit points (and again you see that a lot in B2, recognizing that hit dice are d6 here).
But what he never did in B2 is set any monster hit points over the max possible with d6 hit dice. For example, there are no kobolds with 4 hp. There are no normal orcs with 7 or 8 hit points. The same can't be said of Mike Carr's module B1, which does in fact list some 4-hp kobolds and 8-hp orcs (for example), demonstrating that the B1 module was in fact developed using post-Greyhawk, Holmes Basic rules.
(Note: A more formal demonstration could be done in a comprehensive spreadsheet, listing all hit points in the module, and running an SSE "sum squared error" analysis on each hypothesis and seeing which comes out closer -- much as I did for the AD&D multiclassing problem: http://deltasdnd.blogspot.com/2008/04/multiclass-hit-points.html. I don't have the time now to do that more formal demonstration, but I'm very confident of what the results would be if it was done. B2 was developed and published assuming d6 Hit Dice, not truly in accordance with the Holmes Basic D&D rules.)