OK, first, we need to define "worst".
"Worst" does not mean "I'm not interested in this" or "It's not my cup of tea". The 'guy-bra' argument neatly makes this point.
Let's consider some "worst" definitions:
*Value: from a pure "page content versus price" perspective, A&E takes the cake, reprinting material appearing in other books that were widely owned. Odds are, most gamers owned at least some of the sources used for A&E. Coming in a distant second would be the MMII, reprinting web enhancements, dragon magazine monsters, and splat book data, as well as a few SS monsters. And an even more distant third is the SBGB, nicely reprinting an entire chapter from Song & Silence on traps, as though the game designers couldn't decide which book it belonged in, so they beefed up the page count of both.
Of course the clear winner here is the write-in candidate of the HBGB, with all of it's content being available as random name generators & archived message board data, all of it free.
*Concept: this isn't so much a "we don't like it" as it is "should this have been done for the game". Should the ELH have been for levels "21+", or perhaps a more sane range of "21-40"? For storytelling purposes, 'magic' serves the same thematic purpose as 'science' or 'psionics', so are psionics really necessary to tell the kinds of stories of D&D? This is also a catch-all category for the whiney remarks of "why should there be a book telling us what evil is" or "why should there be a book about how to play monsters".
*Execution: Between promise & delivery often lies a hard road. The clear winners here are the ELH (whose Epic Spellcasting chapter alone would get votes if allowed, I suspect) and D&D, which had great promise as a roleplaying suppliment & was instead a Super-Sized monster manual. Coming in a strong second is Bruce Cordell's Psionics Handbook; it's really telling that Bruce has gone on to write multiple 3rd party books with variant & alternate handling of the psionics rules; if it wasn't broke, how come you're tryin' so hard to fix it? The BoVD gets an honorable mention, simply for including the Lords of Hell with CRs in the 20s, but having most of the rest of the monsters & material made for levels 1-12, creating a lovely gap in the material's usability. While I can't make a personal judgement, I suspect many would put OA in 3rd place for the things it didn't do. I'll also give an honorable mention to the MMII, simply because the FF showed us what a good monster manual should include: rules for integrating the material with existing suppliments. The sidebars in the FF are what the web-enhancement for the MMII should have been.
*Usability: How much hammering is necessary to make the book fit, and how often can you use it? This isn't a "my style of play" category, but rather a question of interaction with the existing rules. You can't ding the MotP just because you don't like the idea of the Planes; the core rules have multiple spells & items that use the Planes, and having good support material helps a lot. OTOH, the PsiHB needs multiple patches to create a psion with less stringent ability requirements than a monk/paladin. The SBGB works well with its own internal logic, but puts astronomical numbers into the already absurd D&D economy, and does not hold up well outside of its own context. S&F is the clear winner here, though, needing not only its own FAQ, but two errattas to boot. It also gets extra credit for starting PrC creep in 3E, a crabgrass threatening to choke every suppliment in sight. SS fits between S&F and PsiHB for problematic elements, having Outsider PCs with 1/4 the skill points of Outsider NPCs, giving us the "Roll with it Feat" and countless other stumbling blocks on the road to integration into a game.
I think most gamers consider the Value, Concept & Execution factors most heavily, with each individual putting weight based on preference. The 'iffy' concept of the ELH, combined with the terrible execution make it a 'worst' for many. The terrible value and unnecessary concept of A&E put it on the 'worst' list for others. 'Enemies & Allies' is a question of 'iffy' value and questionable concept. Frankly, the Book of Challenges looked like a very straight forward module book, and I can't understand the hate towards it. Now, the "Hero Builder's Guidebook" is a true trifecta: Terrible value, horrible concept, & wretched execution. (that WotC did nothing to support the concepts brought up in the book is a good sign that even their own R&D people didn't like it)
