Wednesday, September 15, 2010

Class notes: 9/15/10

We did a run, here's what happened:

Zach: Group 7 is bidding rand(0,20) for every letter.
Nitin (Group 4): Bids based upon the frequency of getting new 7 letter words
Shishir (Group 2): Calculate an average best-case points for using the letter, then bid it
Neetha (Group 8): Tries to bid a lot, but is underbidding relative to everyone else
Lauren (Group 3): Bid 1 by default. Bids more if it will give us a 7 letter word, plans to increase bidding to just overbid other people
EPK (Group 1): Multiply the face value of the letter by 3
Aditya (Group 8): Focus on letters that are frequent in words. When it's an uncommon letter, put out a random bid

KR: If you aren't bidding high and others are, should you be worried?
Daniel: No. The high bidding players will lose all of their points, while the low bidding players will do OK
Nitin: If there were very few players, the agressive bidder will prevail
Aniket: Conservative players will miss out on the letters they need
Dan: Eventually market values will go down, and the conservative players will prevail for their low investment

KR: What did we learn from this round?
Archie: It's OK to be offensive, as long as nobody else is.
Monica: There are other ways to be offensive, ex, bidding high early on OR just randomly bidding high to block others
John: You should decide when to be agressive based upon how many points are left/where we are in the round. If we are later in the game, then you'll probably pay less (via the Vickrey system)


KR: How does the Vickrey auction fit in with this all?
EPK: There's no advantage to bidding high if others are. Just let them pay as much as they want.
DFed: The value of a letter for each team is not really knowable if there are secret letters. In the relatively short period of time that this game runs on for, it's hard to guess what people might bid. Just bid whatever your own value is, plus some perturbation perhaps.
Neeha: How high you bid should also depend on your present score - you don't want to run out of points
Andrew: For a micro-strategy, still value letters after you have a 7 letter word, to bid low and prevent them from getting anything (or at least making them pay)
Archana: Should keep a history and infer the strategy of other players. Realize that if someone needs an A, avoid it, realizing that it will be cheaper
David: When we track bids, we're too conservative, because we are including stingy player.
Jesse: Nobody wants to waste the points to try to block someone else, because maybe someone else will. Limit blocking to small games only

KR: It might be a good idea to focus on our own strategies before working on counter and counter-counter strategies
Neeha: To really block other players, you might need to get 10-12-15 letters, and it will be costly
Aniket: Use a progressive bidding strategy, valuing letters more as they become less available
DFed: There will be deflation over the course of the game, so you should keep bidding something as you keep going

EPK: Should we bid based upon how many points we have?
Flavio: No, look only at what you spent so far on that word.
Monica: But if you have 60 points and everyone else has 100, it's worth trying to catch up.

KR: What about simulating your strategies... run them in parallel then decide which one to use

Neerja: Focusing on countering someone else always gets you into trouble
Daniel: If you are very close with someone else, you SHOULD try to kill them, if everyone else is far behind, because it won't really sacrifice your ranking
Archie: Separate tactics from strategy: Strategies are long term, but within a game you might have multiple tactics
Pilunchana: You might think that your bidding is high, but it might not actually be relative to others


Deliverable for Monday:
Revised player that is correct (doesn't claim it has letters it doesn't), and does something smart (doesn't just bid random, etc)

No comments:

Post a Comment