Simple rules for Complex Systems?
Dear Reader,
I had met my thesis supervising professor today. He is professor for operations research and scientific management and he supervises and helps me with my thesis about inner city routing of fire trucks with real time traffic information. (My childhood dreams come true! Now I am playing with real fire trucks ;-)
We talked about the structure of my thesis and the general problem of fast feedback systems (traffic systems, financial markets) versus slow feedback systems (behaviour of tankers and cargo ships). Slow feedback systems could be modelled stochastically. New information is perceived (a boat will cross the route), the probabilities of scenarios will be adjusted (evasion of the boat, no route change, ...) and than the problem is optimized with the optimal solution as output. In fast feedback systems new information arrives faster than the solution of the optimization is calculated.
I would say, "Hey, I have money, lets buy some more computers!". Bu I am wrong because all market participants will buy more computer if you earn extra money with more computer which diminishes my returns. And this strategy contains huge risks which is explained in more detail in "A Daemon of Our Own Design" by R. Bookstaber. I recommend the book - its a bargain of knowledge! (http://www.amazon.com/Demon-Our-Own-Design-Innovation/dp/0470393750).
Second, a precise modelling of the problem does not lead to better results necessarily, if the model contains a systematic error. An imprecise model with an unsystematic error is better, because it does not advice wrong decisions in a systematic way. (But it advices wrong decisions randomly!).
But what is if the probabilities of these scenarios is unknown and unknowable for me or a third party or the probabilities of scenarios can not be calculated in acceptable time (not enough brain or computer capacity)? Think about the manipulation of traffic lights. You can lengthen the duration of the green phase to increase the traffic flow or shorten it for a decreased flow. The goal is to increase the total traffic flow. If a single traffic light is manipulated, the whole traffic system state differs extremely from the original situation. In other words a huge amount of system states is possible with widely varying possibilities. And this happens day to day while driving a car. My professor advised me to guess and to make assumptions. My first thought was "Pretty scientific Mr. Professor!". My second thought was "Sarcasm is no solution to that problem. And maybe behind his simple answer is some very interesting truth.".
The topic of unknown and unknowable probabilities of system states is in my opinion deeply connected with complex system theory. Mr. R. Zeckenhauser developed in his article "Investing in the Unknown and Unknowable" maxims to handle the unknown and unknowable in the investing world (http://www.hks.harvard.edu/fs/rzeckhau/InvestinginUnknownandUnknowable.pdf). Maybe I can translate his maxims to traffic systems or I find my own maxims which might be transferable to the investing world.
Its quite surprising that simple maxims or rules might be the solution to handle complex systems like financial markets or traffic systems. The first time I read the "Intelligent Investor" I was perplex. Graham introduced very simple rules for investing, like equally weighting your stocks or pay 100 Bucks in your portfolio every month independently from stock market conditions. Maybe he was thinking, that handling the complexities of financial markets is only possible with simple "deterministic" rules. The inverse question is interesting, are few simple rules always better than many simple rules (= complex rule set)? What separates good from bad rules? What are the determining factors?
My brain cooks! I see I have to do more research.
I had met my thesis supervising professor today. He is professor for operations research and scientific management and he supervises and helps me with my thesis about inner city routing of fire trucks with real time traffic information. (My childhood dreams come true! Now I am playing with real fire trucks ;-)
We talked about the structure of my thesis and the general problem of fast feedback systems (traffic systems, financial markets) versus slow feedback systems (behaviour of tankers and cargo ships). Slow feedback systems could be modelled stochastically. New information is perceived (a boat will cross the route), the probabilities of scenarios will be adjusted (evasion of the boat, no route change, ...) and than the problem is optimized with the optimal solution as output. In fast feedback systems new information arrives faster than the solution of the optimization is calculated.
I would say, "Hey, I have money, lets buy some more computers!". Bu I am wrong because all market participants will buy more computer if you earn extra money with more computer which diminishes my returns. And this strategy contains huge risks which is explained in more detail in "A Daemon of Our Own Design" by R. Bookstaber. I recommend the book - its a bargain of knowledge! (http://www.amazon.com/Demon-Our-Own-Design-Innovation/dp/0470393750).
Second, a precise modelling of the problem does not lead to better results necessarily, if the model contains a systematic error. An imprecise model with an unsystematic error is better, because it does not advice wrong decisions in a systematic way. (But it advices wrong decisions randomly!).
But what is if the probabilities of these scenarios is unknown and unknowable for me or a third party or the probabilities of scenarios can not be calculated in acceptable time (not enough brain or computer capacity)? Think about the manipulation of traffic lights. You can lengthen the duration of the green phase to increase the traffic flow or shorten it for a decreased flow. The goal is to increase the total traffic flow. If a single traffic light is manipulated, the whole traffic system state differs extremely from the original situation. In other words a huge amount of system states is possible with widely varying possibilities. And this happens day to day while driving a car. My professor advised me to guess and to make assumptions. My first thought was "Pretty scientific Mr. Professor!". My second thought was "Sarcasm is no solution to that problem. And maybe behind his simple answer is some very interesting truth.".
The topic of unknown and unknowable probabilities of system states is in my opinion deeply connected with complex system theory. Mr. R. Zeckenhauser developed in his article "Investing in the Unknown and Unknowable" maxims to handle the unknown and unknowable in the investing world (http://www.hks.harvard.edu/fs/rzeckhau/InvestinginUnknownandUnknowable.pdf). Maybe I can translate his maxims to traffic systems or I find my own maxims which might be transferable to the investing world.
Its quite surprising that simple maxims or rules might be the solution to handle complex systems like financial markets or traffic systems. The first time I read the "Intelligent Investor" I was perplex. Graham introduced very simple rules for investing, like equally weighting your stocks or pay 100 Bucks in your portfolio every month independently from stock market conditions. Maybe he was thinking, that handling the complexities of financial markets is only possible with simple "deterministic" rules. The inverse question is interesting, are few simple rules always better than many simple rules (= complex rule set)? What separates good from bad rules? What are the determining factors?
My brain cooks! I see I have to do more research.
Kommentare
Kommentar veröffentlichen