After playing Weekend in the Realms, I took a look at the pregenerated characters and doing so inspired some thoughts on the pregens.
First, I think there are two primary considerations with pregenerated characters. First, how will they be able to perform in the module at hand and second, how they will perform in future modules if the players decide to continue in the Living Forgotten Realms community.
As to the first question, I am not one of those people who think that every pre-gen should be hyperoptimized. That said, I do think that they should be constructed so that a player who sits down at a table with custom created characters should be just as good as they are likely to be. New players shouldn't have to think, "how do I get me Character Generator; I could make a better character than this!" In short, I would hope for pregenerated characters to be equivalent to a character who takes a standard charop stat array and the standard feats and powers for a well supported race/class combination. Powers should likewise not be selected on the basis of "if I didn't put this in a pregen, it would never see the light of play." (By contrast, I would consider characters who depend upon specific magic items or special combinations of dragon magazine style feats, diletante/versatile master abuse, and out of character equipment/power selections (like the charop forums' twin striking daggermaster critfisher avengers) to be hyperoptimized).
Continuing in this vein, characters that are simple and straightforward make better pregens than complicated ones. In machines, the more moving parts they have, the more things there are that can go wrong. For characters, the more complex they are, the more system mastery they require in order to play effectively. Since the players most likely to use pregens are those least likely to have mastered the system, simpler=better. For this reason, a rageblood barbarian or an archery focused ranger who can simply attack without worrying about conditional censures or combat advantage are better pregens than an avenger or a rogue. And likewise, an avenger with whirlwind charge (you use it, it works) is a better pregen than an avenger with avenging echoes (it works best with setup or as setup for oath/censure; if it's just used by itself, it is no better than a basic attack).
Likewise, characters should have powers that are going to be effective throughout the module. A wizard with a focus on cold magic, for example, would be a poor pregen for a module where all the monsters were cold-resistant. Doing that would be setting the player up for frustration. Likewise, but slightly less obviously, a charisma paladin would be an unfortunate pregen for a module that featured lots of flying enemies. Said paladin would not be able to keep a divine challenge on them and would not be able to attack them effectively (since his likely stat arrays and proficiencies preclude effective ranged basic attacks).
As to the second question, it is important to consider that the people most likely to play a pregen are new players. And if things go as we hope, those new players will want to continue playing LFR. Since they have experience and a magic item for their pregen, it is to be expected that any who do will want to continue to play that pregen. Consequently, it is not enough for a pregen to be adequate for the adventure at hand. The pregen also needs to be constructed as though it had a future. The player of the wizard shouldn't think "all the other wizards are using this fancy enlarge spell feat, but I don't qualify for it; why did you guys give me a sucky character?" (Fortunately, the wizard pregen is qualified to take that feat--score one in the win column for that pregen). In short, pregen characters should be constructed so that they will be able to take advantage of future feats and powers that are widely thought to be good for their class and should have a stat setup that enables them to take advantage of at least one of the more popular paragon paths--preferably one from the PHB.
As a further consideration along this line, it is important that pregenerated characters fit within the broad mechanical expectations of the designers for the class. This will preclude a number of quirky builds that are fun for experienced players. For example, a friend of mine runs an archer paladin that I am assured is quite effective. However, this would be a poor pregen since a player looking to level up their paladin would find that most of the available powers are not suited to the character's build or playstyle. Likewise, another friend runs a 10-strength warlord who is reasonably effective and will be more so as soon as he hits paragon and takes his bard paragon path that allows him to use charisma for all of his attacks. This would be a poor pregen because the build imposes a hidden consideration for power selection: the character looks to take as many powers as possible that do not require attack rolls or which have an effect that does not depend on an attack roll. A character who simply opened up the PHB and took powers that looked interesting or effective would find that the best power available is rarely the best power for his character. Any quirky character that introduces that level of additional power selection criteria is a poor choice for an introductory game day pregen.
Now, some thoughts on the specific pregenerated characters:
Eladrin Wizard: A solid stat array that is effective in the given scenario and allows players to later expand the character with staple wizard feats like enlarge spell or dual implement spellcaster. The choice of empowered lightning is unfortunate (Acid arrow is even worse, but at least the wizard has another daily option and phantom chasm is a very solid choice) as it is one of the worst wizard powers printed (it is single target with no control aspects, and the wand of accuracy rider (which is uninspiring in general) encourages it to be used in the first round of combat which is usually the ideal time to use an area spell and furthermore complicates the optimal use of wand of accuracy by encouraging its use when it will not necessarily turn a miss into a hit).
IMO, this is the best of the pregenerated characters. The character is effective from the get-go and is set up with a stat array and power array that will serve adequately as a basis for future expansion. Only one retraining session is required to get rid of the abysmal encounter power.
Half-elf Swordmage: This is an unconventional race/class choice but the stat array chosen provides a character that is both effective in the immediate scenario and has room to grow. The selection of a shielding swordmage is also excellent since they are probably the most effective swordmage and are certainly the swordmage that requires the least system mastery. The choice (based on the attack bonuses) of weapon expertise rather than focused expertise, however is a poor one. It doesn't make the character any more interesting--it just makes the character straight up worse than a character that was constructed in a more focused way. I'm not a fan of the encounter or dilletante powers either but the selection of phantom bolt at least goes partway to addressing the swordmage's relative helplessness at range. (I would prefer scorching burst, winged horde, et al). All told, a player who took this character and decided to continue with it would have to retrain at least one item (the expertise feat) but the character is salvageable.
Dragonborn barbarian (thaneborn): Since this is widely thought to be the optimal race/class choice for a barbarian, I had to wonder what the catch was. Well, the first catch is that it's a thaneborn barbarian. Don't get me wrong, I like the idea of a thaneborn barbarian. However, my observation is that getting the most out of a thaneborn requires more tactical skill and system mastery than a rageblood barbarian requires in order to achieve similar results. In a pregenerated character that is most likely to be given to players who are new to the system I would much prefer to see simple characters chosen. When it comes to the powers, there is a pretty big mistake as the dragonborn's second at-will power is shown as an encounter power. That has a pretty big impact on the playability of the character especially as the first at-will power is the weaksauce version of howling strike--like howling strike but it comes with an AC penalty and can't be used on a charge. Any player who wanted to stick with the character would want to retrain that power pronto.
Dwarf fighter: Again, this is a conventional class/race choice and a solid stat array was chosen. The character looked like an ideal pregen until I came to the weapon selection. I understand that the picture shows a mace, but it's not as though there is a shortage of WotC art showing a dwarf with a hammer or an axe--and both choices are simply and straightforwardly better. While I don't think that pregens need to be optimized (and thus weapon expertise rather than dwarven weapon training is a perfectly acceptable choice), I don't think that pregens should be deliberately gimped either. The encounter power choice--shield bash--is a different issue. It is fine at level 1 but since it lacks the weapon keyword and the attack bonus is stuck at Str+2, it will very quickly run out of steam by level 4. That's two retrainings that any player who wants to stick with the dwarf will require--and more if he is suckered into taking a +1 frost mace as his treasure instead of doing the smart thing and upgrading to a warhammer, battle axe, warhammer, craghammer, or longsword immediately.
Now we come to the ones that fail.
Gnome cleric: It's hard to know where to begin here. We start out with an unconventional race/class choice, but unlike the swordmage's stat array that reflected an adaptation to the mechanical realities, the stat array for the gnome seems to be constructed to avoid taking advantage of any racial bonuses. 14 strength is not sufficient to make strength powers effective and even if it were, the cleric does not have any of them. 15 Charisma is not bad but does not really make effective use of the gnome's bonus. The gnome's armor and lack of dex interfere with his racial stealth abilities. You could probably make a playable gnomish devoted cleric but this is not it. Even switching the abilities around a little bit would improve the character. 14 Con rather than 14 strength would give it a bit of survivability. A focus on dex rather than Int would make its stealth abilities more useful. If you are inclined to defend this character, I suppose you'll spout some noise about "quirky characters" being interesting for some players and some players enjoying the challenge of making something playable, but there are two points that go against that. First, this is the only leader class pregen available which means that, in a party of pregens, someone is likely going to be stuck with this guy whether they want to play a quirky character or not. Second, what I assume to be the primary target audience for pregen characters--new or inexperienced players are the least likely to recognize that a character is quirky or that it will require extra careful play to be as effective as those constructed to be more effective.
The verdict: no amount of retraining will make this character as effective as a standard custom built character.
Halfling Avenger: This will probably be my most controversial analysis. The stats are well chosen. The powers are well chosen. And the character still fails. Why is that? Because halflings simply do not make good avengers and while the choices made would be fine for a more traditional avenger race (elf, deva, etc), they do not work for a halfling. I will expand a little bit. Avengers have two primary mechanical schitcks. First, at least ignoring the overpowered hide armor expertise rageblood barbarian, they have a reputation as the most survivable striker. Second, they derive their damage primarily by hitting their targets reliably with big weapons with powers that deal multiple large W dice. Even with both of those, however, they tend to struggle to compete with rogues, rangers, barbarians, and sorcerers in the damage department. Like warlocks, avengers are the low end of striker damage. Now, as a halfling, the avenger starts out behind in her strength, is handicapped in her weapon choice, and is destined to be even further behind the damage curve than a tradition avenger (if there is such a thing for a class that is new with 4th edition). The halfling's defenses are limited because her lack of a primary stat bonus left Wis 18/Dex 16 as the best stat array possible, which is fine but will always be behind the Wis 18/Dex (or int) 18 compatriots. Not only that, the aggressively focused stat array also leaves no points for Con for healing surges, hit points, or the boosting of non-AC defenses. A traditional avenger will have higher defenses, more hit points, more surges, and higher non-AC defenses. Moreover, the character's offense suffers. The limited secondary stat makes for a slightly less effective censure but that doesn't come up often in actual play so it is less of an issue. The bigger issue is the weapon limitation that comes with being a halfling. Since avengers start out low on the damage curve and rely upon the size of their W for damage, having two-handed weapons prohibited is a big disadvantage. In a best-case scenario, the character could retrain and gain bastard sword proficiency and take the scrappy feat... and with a one-feat tax would be almost as good as a standard avenger who took a fullblade until 2W and 3W powers and high crit property are considered. At that point, the halfling falls behind even with the feat tax. And since the avenger primarily does damage through W powers and crits, both of those issues are a big deal--doubly so because as a bottom of the totem pole damage dealer, the avenger doesn't have much room to give up damage and still play as effectively as other strikers of the same level.
Again, the verdict: though creative and a solid effort, no amount of retraining will really be able to salvage this halfling barbarian for a player who decided to stick with it.
All of this in no way means that I think those who volunteered their time to create these pregens are stupid, suck or are otherwise bad people. I have made pregens myself and, given my current experience, I wouldn't design them all in the same way if I re-did them. It takes a lot of effort and no-one will get them all right all the time. Still, it is important to think about them and consider what we can learn for the future from each effort.