I've just revised and (in my opinion, perfected) my HU bot.
It's not really a bot, just a way of playing that requires no thought and forces your opponants to make mistakes repeatedly (mistakes by Sklansky's definition) because making that mistake was thier best move.

I've just played about 30 HU games using the bot stratagy vs. my friend (he's not an amazing HU player, but normally better than me... i explained to him exactly how I will play beforehand too) where we both start with same chip count, I start off as big blind, party poker structure 8k chips in total, 300/600 blinds. I won over 75% of the time.

The entire philosophy behind the bot is to force you take coinflips when you are an underdog. Ironically, it will work better against "good" HU players in comparison to "bad" HU players. It figures that at most, your opponant can get a coinflip on you (exceptions are that he catches PP, but the same exception can be made for the bot so it evens out). Note: your opponant will take at least 6 turns (3 your BB, 3 his BB) to catch on to what you're doing.

You're first to act: Go all in every hand without exception.
You're big blind and he calls: Go all in if your hand is suited, KQ, KA, AQ, any PP, or the higher of your two cards is a diamond. Otherwise, check and fold to any agression unless you hit a monster. In which case, downplay.
You're big blind and he raises but is not pot committed: Same as above.
You're big blind and he raises all in or becomes pot committed: All in with AK, KQ, AQ, or any pocket pair 8 or larger.

If he goes all in 3 hands in a row when he acts first, loosen your calling restrictions.
If he folds when first to act 3 hands in a row, no longer re-raise with diamond high card.

Why this works:
It works because by the time you've figured it out, unless you've caught a really good hand... You're low stacked and it takes two double ups (sometimes three) to take out HU Bot... and HU Bot has a very low chance of being dominated due to its extreme randomness. When facing HU Bot: If you don't catch cards, and aren't willing to make a stand with crap, you lose. If you don't catch cards are are willing to make a stand with crap, you lose 40% of the time and you're back where you started the other 60% of the time. If you do catch cards, you lose 25% of the time and are back where you started 75% of the time. It's rare that you do catch more than one good hand per 5 rounds. If you don't take a stand within 5 rounds, HU Bot has you right back to underdog status. The better the player, the less likely he is to take a stand with crap... so the more likely he is to eventrually lose. The best stratagy against HU Bot is to call an all in with any ace or any king and take whatever odds you get from that. No one "good" uses that stratagy until at least 8 HU hands (4 each position). By this time, you are such a underdog that you need a lot of luck to come back and win.

This stratagy works less well against a big stack as big stack will more likely call you with just about anything once he knows you go all in every hand and you no longer have the stack to take a hit or two. However, starting off understacked 3k to 5k, HU Bot is still hitting a solid 40%.

Try it out, deal some hands, use Party Poker Heads Up structure and try to face down HU Bot. It's earily effective. (Keep in mind, you have to pretend that you don't know it's formula at least for the first couple hands).