Comments on: Why we won’t build a stock-market simulator http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/ A slice of lime in the soda Sun, 26 Oct 2014 19:05:02 +0000 hourly 1 http://wordpress.org/?v=4.2.5 By: kaylabi http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-36081 Thu, 16 Feb 2012 17:16:06 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-36081 Golden Networking has created an instructional DVD for executives and professionals in high frequency trading, entitled, “The Speed Traders Workshop 2012”. This 4-disc DVD set walks professionals and non-professionals through the main issues, challenges and opportunities practitioners face to consistently capture alpha using fast computers. The DVD uses non-mathematical terminology and provides insights to successfully implementing speed trading while avoiding minefields. It is the ideal reference for professionals in the world of alternative investments. Visit http://bit.ly/znIl3E for more information

]]>
By: Christofurio http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34632 Mon, 02 Jan 2012 01:08:15 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34632 Yes, ottorock. This all raises the question — what do we do if we create the market simulator at great expense and it produces the result: 42?

Ask the dolphins for help? They will have left, though they’ll be grateful for the fish.

]]>
By: ottorock http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34623 Sat, 31 Dec 2011 20:53:30 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34623 I can’t seem to get the writings of Douglas Adams out of my head for some reason.

]]>
By: TFF http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34620 Sat, 31 Dec 2011 17:54:15 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34620 Then jomiku, if any attempts to make the market mechanism more stable are fraught with moral hazard, should we go the opposite direction and warn market participants that they are ultimately responsible for their orders?

Suppose the trades at $0.01 had been allowed to stand? Would traders learn to be more careful?

]]>
By: jomiku http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34619 Sat, 31 Dec 2011 17:04:41 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34619 TFF, the idea is to abstract the existing simulator. There is no underlying market mechanism other than what exists in the current simulation that is the stock market. The problem remains: abstracting a simulation introduces differences because simplification through abstraction cannot fully replicate a complex system. The idea is seductive but ultimately nonsense. For example, you may get it right but that context of “right” will only last as long as it does and you won’t know the context has changed until you’re wrong. This again is a version of the long tail problem: you get it right enough for a while and that increases your confidence in the abstraction you’re using, so you trust that more until context shifts more radically and you lose everything.

]]>
By: Greycap http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34611 Sat, 31 Dec 2011 14:08:15 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34611 So the run of the comments here is claiming that 1. markets cannot be proved safe by simulation because 2. the behavior you would like to simulate is radically un-modellable: market agents do not obey any set of rules, or, if they do, the rules are too changeable in ways that cannot be predicted.

The first part of this claim is irrelevant: the original proposition was not that if you could not prove there is a problem, then there is no problem. It was that if you can prove there is a problem, then there is a problem. Good luck disputing that.

The second part of the claim is not as valid as it appears. The reason is that, to first order, all modern market agents are computer programs. Their behavior under any given set of conditions is completely specified. One could imagine a regulatory regime in which it was illegal to trade with an algorithm without submitting the program itself to a central regulatory agency, along with one’s capital and leverage guidelines. These latter need not be deterministic parameters, merely Bayesian priors subject to simulation in a sandbox that includes all programatic agents, plus some emulation, however poor, of the less important merely human agents.

The purpose of such a setup would be to look for deadly interactions between agents; although the conditional behavior of individual agents is completely specified, the systemic interactions are hard to predict. I could easily imagine that such a setup would sometimes uncover genuine problems before they are encountered in the real world, which is all that the original claim is saying.

The problem would be in attribution analysis: what is the regulator supposed to do when a problem is discovered? If there is a problem interaction between programs A and B, is B supposed to be rejected simply because A was there first? If not, how is blame to be divided between the two?

All of which is a long way of saying that I think Felix is right, except that I don’t think that the simulation need be as expensive as he does.

]]>
By: Oenologist http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34609 Sat, 31 Dec 2011 02:50:37 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34609 I agree with tgriffen that it would be almost impossible to build a meaningful simulator.

As with many modelling approaches, you would essentially get out what you put in. You could easily design one seemingly realistic simulator that results in an event like the flash crash, while another seemingly realistic one doesn’t. As always, you would need to make strong behavioural assumptions on the way in which agents react to market conditions. The results would rest crucially on these assumptions, but it may only take a few large market participants behaving in a different manner to the model to result in very different results.

]]>
By: TFF http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34608 Sat, 31 Dec 2011 00:42:15 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34608 jomiku, I think Felix is talking about the market *mechanism* itself. Which could be seen as including HFT trading and other forms of market making, buy/sell queues, futures markets, and so forth. You might then throw random events at the black box and see how it responds?

The Flash Crash resulted in some stocks trading hands at $0.01 per share. That wasn’t a representation of economic activities and/or opinions, that was a breakdown in the market mechanism itself. A “simulation failure” if you like?

]]>
By: jomiku http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34607 Sat, 31 Dec 2011 00:29:09 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34607 Um, the stock market is a simulator. It is a representation of economic activities conducted in the real world: companies make stuff, do their business and their results and prospects are expressed in stock values, option values, derivative values, etc. The simulator is complex; it can be reduced to stock values or stock plus option values or whatever but it extends to encompass huge numbers of variables across the world. These include external inputs ranging from weather to irrationality.

If you built a simulator of the stock market, it would be the stock market. If it is not the stock market, then it will be inaccurate. That is a short summary of the long-tail risk problem: you discover the inaccuracy when it hurts.

]]>
By: lambertstrether http://blogs.reuters.com/felix-salmon/2011/12/30/why-we-wont-build-a-stock-market-simulator/comment-page-1/#comment-34595 Fri, 30 Dec 2011 15:02:00 +0000 http://blogs.reuters.com/felix-salmon/?p=11673#comment-34595 What makes you think we haven’t built a stock market simulator? Assuming it’s technically feasible. If it’s only a question of money, well, there are people for whom hundreds of millions of dollars is not very much, and would have every incentive to invest in the ultimate way to game the system. See Hubertus Bigend in William Gibson’s Zero History, where a stock market simulator is the McGuffin.

And a followup: If a stock market simulator had been built, would we necessarily know about it? Why would you think that?

]]>