Comments on: A Discrete Model Gauging Market Efficiency http://www.win-vector.com/blog/2009/09/a-discrete-model-gauging-market-efficiency/?utm_source=rss&utm_medium=rss&utm_campaign=a-discrete-model-gauging-market-efficiency The Applied Theorist's Point of View Thu, 15 Jul 2010 00:13:50 +0000 hourly 1 http://wordpress.org/?v=3.0.1 By: jmount http://www.win-vector.com/blog/2009/09/a-discrete-model-gauging-market-efficiency/comment-page-1/#comment-699 jmount Sat, 12 Sep 2009 14:42:37 +0000 http://www.win-vector.com/blog/?p=809#comment-699 Scott, looking forward to your paper. I am a big fan of the Kelly system (and of one of its famous proponents- Thorp). Bet variation is the usual way you convert a system that has a pronouncement like "so no better can be correct more than 50% of the time" into pure profit (by betting large on the 50% of the time you are going to win). The particle filter stuff is important- but I think it is over sold. In my opinion particle filters are very good at inferring hidden state and very bad at estimating transfer functions (many of them assume you know the function before you start). Also, they often seem like a method to discover autoregression the hard way. Yet there is a huge population that thinks the Kalman filter solves everything. As for libraries- no not really. My method for a while has been to avoid complete frameworks like the plague and say "what if I stuck a slightly more verbose formulation of the problem into a way better commercial optimizer library than the framework supplies?" Scott, looking forward to your paper. I am a big fan of the Kelly system (and of one of its famous proponents- Thorp). Bet variation is the usual way you convert a system that has a pronouncement like “so no better can be correct more than 50% of the time” into pure profit (by betting large on the 50% of the time you are going to win).

The particle filter stuff is important- but I think it is over sold. In my opinion particle filters are very good at inferring hidden state and very bad at estimating transfer functions (many of them assume you know the function before you start). Also, they often seem like a method to discover autoregression the hard way. Yet there is a huge population that thinks the Kalman filter solves everything.

As for libraries- no not really. My method for a while has been to avoid complete frameworks like the plague and say “what if I stuck a slightly more verbose formulation of the problem into a way better commercial optimizer library than the framework supplies?”

]]> By: Scott Locklin http://www.win-vector.com/blog/2009/09/a-discrete-model-gauging-market-efficiency/comment-page-1/#comment-697 Scott Locklin Sat, 12 Sep 2009 07:27:46 +0000 http://www.win-vector.com/blog/?p=809#comment-697 Well, I've got a paper/blog thingee coming up on how I think the early Rentech dudes did what they do: HMM type things figure heavily, as does Kelly bets. Bet sizing systems are rarely thought about, and most I've seen are provably retarded; Kelly is smart, and it's a natural fit to HMM kinds of ideas. Got a favorite Hidden Semi Markov Model library? I guess particle filters are the thing people use these days, but I like fitting: it's sort of mentally parsimonious, and I like to think it works better when your data sucks. Well, I’ve got a paper/blog thingee coming up on how I think the early Rentech dudes did what they do: HMM type things figure heavily, as does Kelly bets. Bet sizing systems are rarely thought about, and most I’ve seen are provably retarded; Kelly is smart, and it’s a natural fit to HMM kinds of ideas.

Got a favorite Hidden Semi Markov Model library? I guess particle filters are the thing people use these days, but I like fitting: it’s sort of mentally parsimonious, and I like to think it works better when your data sucks.

]]> By: jmount http://www.win-vector.com/blog/2009/09/a-discrete-model-gauging-market-efficiency/comment-page-1/#comment-696 jmount Sat, 12 Sep 2009 00:13:38 +0000 http://www.win-vector.com/blog/?p=809#comment-696 Scott, my background is Markov Chains and Hidden Markov Models. I worked with them before I went into finance- I guess if you have a hammer the whole world ... Scott, my background is Markov Chains and Hidden Markov Models. I worked with them before I went into finance- I guess if you have a hammer the whole world …

]]> By: Scott Locklin http://www.win-vector.com/blog/2009/09/a-discrete-model-gauging-market-efficiency/comment-page-1/#comment-695 Scott Locklin Fri, 11 Sep 2009 22:57:05 +0000 http://www.win-vector.com/blog/?p=809#comment-695 You know about the Santa Fe artificial market, and related developments, right? Funny you keep using MM's and HMM's in your papers. I've been obsessed with 'em since I figured out Lenny Baum worked for RenTech. You know about the Santa Fe artificial market, and related developments, right?

Funny you keep using MM’s and HMM’s in your papers. I’ve been obsessed with ‘em since I figured out Lenny Baum worked for RenTech.

]]>