That's a profoundly intelligent suggestion. I'm interested to hear if it's possible/relatively easy to implement in practice. If so, I think this principle could be applied to quite a few problems which require massive computational facilities but where the algorithm is reasonably well-known.

For info on distributed high performance computing do a search on "grid computing" on Google

as i've been reading more, i've found a few other interesting projects on Google. you might want to take a look. if there were enough interest in setting a project up, i would definitely get involved. i have a spare server that sits on the web fairly idle right now. it wouldn't scale too far (it's just a wee RaQ4), but we could start organizing the effort and testing the concept on it for sure. if from experimentation we determine a larger roll out to be feasible, i'm sure we could get some support to grow the capabilities.last year i was reading quite a bit about clustering and distributed computing, but never had any chance to implement. to be a success we'd need someone with some parallel and distributed experience to join the effort. i don't remember seeing any comments to that effect on the forum. dcfc, maybe? and then the bigger question is who's got the time to spare?anyone else interested in taking this further? anyone have experience (at an engineering/coding level) with distributed computing over the internet?

MobPsycho it's nice to read somebody on these boards with a "real" opinion.You come off like a pure cash trader? Is that true? The only dispute I have with any of your points is that ECN's add no value. 99% of the guys making markets in equities aren't that smart. You know it and I know it. So why the hell do we need them? Your rant was pretty funny. And market microstructure literature will continue to suck until the guys writing the articles come straight off the trading desks (ie never).

Kapital,thank you for your posts!We're currently evaluating different strategies concerning the future of KapSyn and we are glad the public responds positively and constructively to our past efforts in implementing the Kapitalmarkt Synergetik model!A port of the KapSyn program towards distributed computing is not at least dependent on the purpose the user has in mind. Currently, the program is used to analyse intraday price movements and transaction costs. Therefor I'm currently developing database routines which back up all variables during the simulation run.Test runs of these modified version have shown that even a simple simulation (5 stocks, 20 traders) can result in a data base with one million of data sets and an overall size of tens of megabytes.If we move the program to distributed computing this would prevent us from using this feature as there is no easy way to collect data packets 20 MB from hundred our thousands of users every 5 minutes by internet.Hence, any move to distributed computing may only stem back from the motivation of obtaining RESULTS (intraday prices and transaction costs) rather than HOW this results have come about. This is the reason why I first think of large supercomputer systems rather than distributed computing. A port to a supercomputer system would assure the availability of all data for further analysis.Anyway, the decision about a port isn't primarily up to me. We're a team lead by the Professor Loistl who developed the model and we are regularly discussing such issues. There are still a lot of features which have to be built in before we can think of moving to another platform.Regards,KapSynDeveloper

B2,Your damnation of all microstructure literature is a little harsh (i assume you haveread all of it). What exactly are you looking for? You may find you have the same problemswith modelling in a lot of subjects. Sure, there are a lot of bad papers publishedas there are in all fields, but there is also some good work.As to guys of the desks writing papers, well i assume you mean the 1% of them who aren't dumb.

filthy i meant that the body of literature as a whole isn't that great, not that there aren't some useful individual articles. it's not the academics' fault, really. they just don't have access to the data and never will. you're just simply not going to see some guy using a CRSP data set say, well, XXX bank's last IPO ate it because XXX scummed their biggest buy side account last week and word got around.several flaws in your last statement:1. you assume that it requires intelligence to publish2. 1% is probably too high. but then again the academic %age isn't very inspiring, either.3. since the desk traders ARE the microstructure participants, even if they are dumb don't you think they'd be better at writing about themselves than someone else would?4. why is "intelligence" so desirable, anyways? intelligence has insufficient correlation to success in the market to convince me that it's particularly desirable in that context. while it certainly has aesthetic appeal to me on a personal level, a person's intellectual sophistication can often cloud a decidely simple issue.

the value of microstructure lit is not necessarily that it can tell you something new but thatit can give a theoretical framework for thinking about the stuff you already knew. that is where the value lies in simplifiedmodels in all fields. a model is like a map. for a map to be perfect it would need to be the same size as the landscapeit maps. so we simplify it until it is a useful size. same with models. we simplify until we have the essence of what we need.this is standard practice in science. it is called phenomenological modeling.why is this useful?well, you need a theoretical framework to base new theories on. this is how you knowledge grows.think about “dynamic hedging” (which I see in another post you liked). this didn’t say anything new. rather it restated,in a reasonably coherent framework, what traders knew and did. This was useful.something i often notice with traders is that they often misunderstand why they make moneyor even misrepresent what they actually do. i think having academics look atthe situation from outside is valuable.the data point is valid but is changing. there have been a number of recent papers on price impactthat use real bid/ask data.

QuoteOriginally posted by: filthythe value of microstructure lit is not necessarily that it can tell you something new but thatit can give a theoretical framework for thinking about the stuff you already knew..Just to add my two pence to this thread, I concur with filty , Knowing that there is a bid-ask bounce effect in the market , helps you in better estimating market parameters.There are people in this world who have a very poor opinion on Garch , and if they were to useintraday tick data and estimate model parameters they are going to be way off , simply because of not accounting for the above effect. Most people end up damning the model, whereas the problem lies elsewhere in not accounting for a certain aspect of the market.