Aaron,I really liked your bedtime story: the central idea of selecting one’s own payback distribution is terrific. I shall in future swap the possibility of uncertain but possibly very large gains via Kelly, for an ‘almost certain’ modest return, lightly leveraged.I’m a bit out of my depth in the technical forum but I should be interested to know whether anyone has examined other alternatives to Kelly e.g.- the ‘mean variance’ ideas of Ziemba/Zhao et al- the ‘Reaching Goals by a Deadline’ policies described by Sid Browne- any others?You might be interested to know that professional gamblers are as concerned with deadlines and outcome distributions as investors. Although it is possible to diversify risk away over time it is also possible to starve before pay day!Pleasant dreams to father and daughter.Soc.

I have read both:Y. Zhao and Bill Ziemba, "A stochastic programming model using an endogeneously determined worst case risk measure in a risk-return framework for dynamic asset allocation." Mathematical Programming 2001 andSid Browne, "Reaching Goals by a Deadline: Digital Options and Continuous Time Active Portfolio Management." Advances in Applied Probability 31, 551-577, 1999.You notice both were published in journals organized by tools used rather than problem subject area. Both of them are interesting results, but neither one is finance, in my opinion.Anyone with a mathematical neuron in their head and even a little interest in money, looks at stock prices going up and down and tries to model it. Mostly people pick a technique based on tools they are good at using. They pick some sort of model of what's going on, then they have to set an objective function and some constraints to make the problem satisfying: easy enough for them to solve, but hard enough to make it seem like they did something. There's nothing wrong with this, and sometimes people stumble onto useful insights this way. But mostly they come up with mathematical solutions to problems no one has, based on models that do not reflect reality.I call this casino finance, because the prices are treated like roulette spins or dice rolls. Real finance asks where the numbers come from, why stocks are traded, what is a price? Prices are not the raw data, information is.There are a lot of clueless people practicing finance. 10,000 public mutual funds in the United States invest in stocks. None has a track record significantly different from random selection results. That's 10,000 managers, plus analysts and traders, who could set up Excel spreadsheets with lots of RAND()s in them to make all decisions, and no one would notice. These people are desperate for something to calculate, something that reduces the apparent randomness of what they do, something they can use to explain their actions and seem important. These people are candidates for shiny new techniques, and it doesn't matter whether they work or not.Real finance dismissed this sideshow many years ago. There are real, hard, useful problems to solve. But you have to get into the markets to even ask the questions.

Aaron,Agreed - or rather I follow what sounds like excellent advice in an area where I have no expertise.Q. If I have done my analysis of the raw data based on on good understanding of the underlying subject matter, and come up with a set of favourable opportunities, are you suggesting I employ no mathematical techniques whatever?Or put it another way: if you don't like Ziemba or S. Browne's work, what do you like? [Assume the Aaron Browne 'roll your own payoff matrix here' strategy is currently the only one on my newly forming list ]Thanks.----Soc.

Hi Aaron,Thanks for spreading your wisdom around this place. I'm new to this business and am generally getting my head kicked in during interviews, but after reading your article something clicked. I nailed 3 (admittedly simple) questions during an interview (that I probably would have fumbled otherwise) by keeping in mind the general idea of looking out at the frontier of "what you want" and then thinking about how to get there. I know that sounds absolutely trivial to the experts, but for me it was definite progress. Things are coming together slowly but surely. This place is awesome.Thanks,Eric

Thanks, IamEric and Socrates.<< If I have done my analysis of the raw data based on on good understanding of the underlying subject matter, and come up with a set of favourable opportunities, are you suggesting I employ no mathematical techniques whatever? >>No, I'm all for mathematical techniques. But what you've left out of your formulation is the goal. If you're playing around for fun, or trying to develop useful new tools, then what you've described is great and you can even leave out "good understanding of the subject matter." You're the amateur or the consultant to the process.I did a lot of statistical consulting before I got interested in finance, and I loved being given a problem in an area I knew nothing about. I learned a lot of interesting stuff from experts, but I could tell them things they didn't know. When the experts have parsed the problem and gathered the data, it's great fun to be the one to pronounce the solution. But I was only doing statistics, not legal analysis of the meaning of sex discrimination, business analysis of whether to offer the McDLT sandwich, geophysical analysis of groundwater flows, or sales analysis of credit life insurance.Real finance starts with a goal. The simplest goal is "make a lot of money by finding and exploiting profitable trading patterns." But you can't do it. Another goal is "raise money to start and nuture a business." This is possible. Or "manage investments in such a way that people will invest with you and pay your fees." This, too is possible. These are the two general goals, but there are hundreds of subversions for different institutions in different situations, and an equal number of problems for intermediaries in between these problems. And there is the grand unifying problem of solving both these problems at the same time.

Real finance starts with a goal. The simplest goal is "make a lot of money by finding and exploiting profitable trading patterns." But you can't do it. q]i agree with your overall argument but thisdismissal of profitable trading is harsh. after all, scores of arbitrage desks do this all the time. Or do you meana customer traing from home can't do i?

<< i agree with your overall argument but this dismissal of profitable trading is harsh. after all, scores of arbitrage desks do this all the time. Or do you mean a customer traing from home can't do i? >>No, I mean that even spec trading operations have to consider risk. If you give someone a chunk of money and say, "make money for me," you are not likely to be successful. Experience shows that it can work great for a while, but sooner or later the money disappears. If you want to get rich you have to set parameters.Arbitrage desks do indeed exploit profitable opportunities all the time, as do hedge funds. Both entities regularly crash and burn as well. One bad day can wipe out ten years of profit. So you have to manage them with an eye toward longer-term/tail-probability considerations.

[qi agree with your overall argument but this dismissal of profitable trading is harsh. after all, scores of arbitrage desks do this all the time. Or do you mean a customer traing from home can't do i?</i> >>No, I mean that even spec trading operations have to consider risk. If you give someone a chunk of money and say, "make money for me," you are not likely to be successful. Experience shows that it can work great for a while, but sooner or later the money disappears. If you want to get rich you have to set parameters.Arbitrage desks do indeed exploit profitable opportunities all the time, as do hedge funds. Both entities regularly crash and burn as well. One bad day can wipe out ten years of profit. So you have to manage them with an eye toward longer-term/tail-probability considerations. >>

Sorry guys need to learn that quote system. But I strongly agree Aaron. I'm involved in electricity trading and as we all know that the mean reversion, jump difusion analysis will educate us on is that the 1 in 20 year or 1 in 100 year event is the event which sends these companies to the dogs. Your ability to manage and control your risks from a holistic approach with the consideration of each individual risk component and it's correlation to the moment setting the direction. I also agree that Aaron provides interesting reading, thanks for the education