Comments on: What is (and what good is) a combinatorial prediction market? http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/ Musings of a computer scientist on predictions, odds, and markets Thu, 09 Feb 2012 08:47:34 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Eric http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-25667 Eric Sun, 24 Apr 2011 09:37:13 +0000 http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-25667 The beauty of LMSR based market maker is the nice properties that you can prove mathmatically. As pointed out by David, sooner or later, a balance needs to be make between maintaining one set of properties over another. However, the profound impact on behavior is often out of reach for formal analysis or even computational analysis. I think a more plausible approach for dealing with arbitrage loophole (a tiny trivial implementation details may trigger a infinite loop of free money) would be implementating another layer to detect such a loophole. (this may even be able to fix the loophole on the fly, if it provides a negative feedback to the main algorithm, with the cost on computation as well as violating the beautity of LMSR and it's nice properties.) I think a better name of "arbitragy attack" in David's sense would be Market Manipulation. This term is better because there are two components for market manipulation: 1. the traders can predict the behavior of the market maker (and if he is certain he would be the only trader over a small period). 2. He can profit if his prediction of the market maker is indeed correct. From my practice of implementing real money maker maker, 1 often holds and the market maker becomes a free money pump... I was dying to deal with market manipulation issue with most popular market making algos and I think it is another important property to be approved to make the algo useful for real money. The beauty of LMSR based market maker is the nice properties that you can prove mathmatically. As pointed out by David, sooner or later, a balance needs to be make between maintaining one set of properties over another. However, the profound impact on behavior is often out of reach for formal analysis or even computational analysis. I think a more plausible approach for dealing with arbitrage loophole (a tiny trivial implementation details may trigger a infinite loop of free money) would be implementating another layer to detect such a loophole. (this may even be able to fix the loophole on the fly, if it provides a negative feedback to the main algorithm, with the cost on computation as well as violating the beautity of LMSR and it’s nice properties.)

I think a better name of “arbitragy attack” in David’s sense would be Market Manipulation. This term is better because there are two components for market manipulation: 1. the traders can predict the behavior of the market maker (and if he is certain he would be the only trader over a small period). 2. He can profit if his prediction of the market maker is indeed correct. From my practice of implementing real money maker maker, 1 often holds and the market maker becomes a free money pump…

I was dying to deal with market manipulation issue with most popular market making algos and I think it is another important property to be approved to make the algo useful for real money.

]]> By: Jump http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-16851 Jump Tue, 15 Feb 2011 08:58:18 +0000 http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-16851 One in 9.22 quintillion. I'll spend a buck on the lottery instead.... One in 9.22 quintillion. I’ll spend a buck on the lottery instead….

]]> By: financial spreadbetting http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-3643 financial spreadbetting Wed, 21 Jul 2010 05:57:48 +0000 http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-3643 I think the market maker has more of an influence and may not act as rationally as you describe. I think the market maker has more of an influence and may not act as rationally as you describe.

]]> By: David Pennock http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-374 David Pennock Wed, 01 Apr 2009 11:15:20 +0000 http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-374 Thanks Luke. Oh, to be more clear by "arbitrage attacks" at the end, I meant infinite loops where bidders can systematically take advantage of the market maker over and over, extracting arbitrary amounts of money. I agree that arbitrage in the usual sense is good for accuracy, but I believe at least for a prediction market it's much better for the auctioneer or market maker to handle all the mechanical "plug and chug" arbitrage and logical inference, forcing traders to focus on providing information, not searching for "free money". Thanks Luke. Oh, to be more clear by “arbitrage attacks” at the end, I meant infinite loops where bidders can systematically take advantage of the market maker over and over, extracting arbitrary amounts of money.

I agree that arbitrage in the usual sense is good for accuracy, but I believe at least for a prediction market it’s much better for the auctioneer or market maker to handle all the mechanical “plug and chug” arbitrage and logical inference, forcing traders to focus on providing information, not searching for “free money”.

]]> By: Luke H http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-373 Luke H Thu, 26 Mar 2009 00:28:36 +0000 http://blog.oddhead.com/2008/12/22/what-is-and-what-good-is-a-combinatorial-prediction-market/#comment-373 Interesting you describe it as "arbitrage attacks" as something bad. Arbitrage on the bundled stocks will adjust the price to more accurate levels, ie, it will reduce the aggregate price of the stocks when they are more than 100%, and increase the aggregate stock prices when it is less than 99%. So arbitrage increases the accuracy of the prediction - what's bad about that? Interesting you describe it as “arbitrage attacks” as something bad.

Arbitrage on the bundled stocks will adjust the price to more accurate levels, ie, it will reduce the aggregate price of the stocks when they are more than 100%, and increase the aggregate stock prices when it is less than 99%. So arbitrage increases the accuracy of the prediction – what’s bad about that?

]]>