That depends, is smoke coming out of your ears? Alpha would you agree that the optimal complexity of a discretionary trading system depends on the horsepower of the trader? If our brains are like computers, a linux based 1000 MHZ machine might test out as dramatically more capable than a 400 MHZ running windows 95 etc. p.s. that reminds me of an old programmer joke- it said "run windows 95 or better" so i got Linux
Darkhorse, Excellent question about complexity vs. horsepower. This is my favorite type of question -- a question that I thought I had the answer to until you asked it. Whether the optimal complexity of a trading system grows as the horsepower grows depends on the nature of the pattern-finding/trade-system development methodology. <b>More horsepower may be bad for the mechanical trader</b>If the trader is strictly mechanical and uses data mining/artificial intelligence/system optimization methods, then there is an upper limit. The total useful data on the markets creates an upper bound for what can be statistically extracted from that data -- get too fancy and you overfit or find patterns that are not statistically defensible given the massive numbers of patterns that a mechanical optimization/screener system can process. More computer horsepower does not help overcome the finite dataset that is available. I suspect this is why some traders backtest and optimize a system to extreme profitability only to find that it fails rapidly in actual trading (even if they are disciplined). Brute force pattern finders, screeners, and optimizers make it too easy to fit an overly complex model to the data. This style of mechanical system development has an optimal complexity level that is below what current computers can do IMO, so more computer horsepower probably does not help. <b>More horsepower may be good for the discretionary trader</b> For discretionary traders, there is a strong chance that more mental horsepower leads to greater trading performance (with some key exceptions). Being able to weigh a greater combination of macro-economic, company fundamentals, and market psychology factors lets smarter traders foreseen both the long-term fundamental future (the trading bias) and near-term technical trading pattern (setups, entries, and exits). Traders that have the smarts to emulate the thinking of other traders (to know what other traders are thinking before those other traders can act on that knowledge) will have an advantage. Of course, high horsepower discretionary trader have their own limits due to arrogance and over-confidence. Excessive smart traders can also fail if they employ excessively rational models of the markets. For example, Milton Keynes, the Nobel-prize-winning economist, blewup his account (guess that where he got the "animal spirits" stuff from.) Also, even high-horsepower discretionary traders can get trapped in a superstitious feedback loop -- being emotionally stimulated to believe in intricate, but erroneous patterns due to random positive and negative feedback from the markets. <b>More horsepower may be good for the "scientific" trader</b> The most interesting case are those traders with a good theoretical understanding of the markets (including both economics AND trading psychology) that can generate a small number of strongly defined hypotheses about market behavior, test those hypotheses on actual data, and create successful trading systems. These "scientific" traders probably benefit for both mental and computational horsepower. Mental horsepower lets them construct better hypotheses and computation horsepower helps them simulate and test their hypotheses. They do suffer from both the limitations of mechanical system creators (if the scientific trader tries to test too many hypotheses with too little data) and the limitations of discretionary traders (if the scientific trader gets caught in a superstitious feedback loop or tries to impose excessive rationality on the markets.) Me, I would like to think I am in the category of the scientific trader, although some might consider me just another dilettante mechanical trader. RE: "run windows 95 or better" so i got Linux: LOL! The smoke from the ears is dissipating now (although my wife does call me a "hot head" for thermal, not emotional reasons) -Traden4Alpha
Quote from Michael Steinhardt[Steinhardt,Fine + Berkowitz] Personaly I get a lot out of charts the more years i use them including yesterday. Note Mr. Steinhardt's personal opinion,not necessarily my work plan, both of which I respect. Traden4Alpha[interesting nick name], I don't think you are thinking too much;just incorrectly when you imply chart paterns [or price]are random for me. It's not easy and not random. _____________________________ Another non random pattern but different context. ''You call it luck but what I call it is a small sample''John Henry -trend follower-partial Marlins owner'' 8/10/02[Boston Globe]
Actually, I do NOT think that prices are random. I think they are complex mix of stochastic and deterministic components. The challenge for traders is to find and synchronize with the deterministic components of the price action while managing the risk associated with the random components of the price action. Unfortunately, finding a complex deterministic signal embedded in a stochastic signal is fraught with peril. Both discretionary and mechanical systems traders can easily think that they have found a deterministic pattern, when they have not. What is worse, is that the deterministic component is a nonstationary chaotic process (actually, its a self-modifying, complex adaptive system). The mental processes and top-of-mind rules that traders and investors use will shift in time (as do the market clearing mechanisms). This changes the deterministic component to include both modes of price action (e.g., bear, bull, and sideways phases) as well as long-term shifts in spreads, variances, covariances, etc. So, I totally agree with you about the chart and price patterns: It's not easy and not random. Wishing you profitable trading, -Traden4Alpha P.S. My nickname (BTW, I like your's too) comes from my goal to maximize the alpha of my trading activities with respect to the price trajectory of the underlying tradables (Yes, i do also worry about beta and residual risk). I use alpha because it helps me assess whether the timing of my trades is actually superior to that of long-term-buy-hold or random-entry trading. Alpha keeps me from confusing brains for a bull(or bear) market.
I sense here that many of you realize that there are too many variables to find the deterministic motion of the stock market- ergo perhaps there is some random distribution function which will describe the collective motion of the market. This is very common in science when there are so many variables that statistical analysis must be used. Examples abound: a)The motions of molecules in a bottle of air follow Boltzmann statisitics b)The decay of atoms follow Poisson statistics c)the flipping of a coin follows Binomial statistics d)The movement of electrons in a metal follow Fermi statistics So what statistical distribution do stocks follow? Interesting question and perhaps someone will find the answer( or has already found it and is not telling).
Agree (at least mostly)- movement is not truly 'random' but because we don't understand it most of the time, it might as well be. In addition to being a scientific term of debatable application, random is a subjective descriptor depending on the viewpoint of the user (much as 'simple' is relative also). Thus the discretionary trader pays attention at all times but only acts when he comes across a 'clarity pocket'- that small percentage of the time when the correct move is clear. Rules that describe activity and rules that predict activity are two entirely different things though. The 'why' is potentially discernible on a broad scale, the 'where and when' most likely not. You can't solve a puzzle with a trillion pieces that rearrange themselves at will, you can only seek out occasional connections among the pieces you are looking at. p.s. Alpha were u inspired by Ben Warwick's book in yr handle choice? I haven't read it but have heard good things p.p.s. didn't Jackson Brown have a song 'too many angels', traders can mentally edit that as 'too many variables'
You need not complicate the definition of random. Random variables are not subjective, they are very well defined mathematically and scientifically. If the only information you had at your disposal were the closing prices of a stock, then the movement of the stock up or down each day is a discrete random variable. The question that this original thread by yabz directed itself to and the one I have raised is, does this random variable of stock movement follow some known distribution function. Surely there must be an answer to this question.
Not trying to complicate it, just trying to translate how it's used in real time regardless of academic correctness...classrooms and textbooks and experiments are artificially concrete and exact and precise, while the real world is none of those things. A sloppy definition that works in real time is better than a precise definition that can only exist in laboratory conditions. No offense but this is exactly why professors teach, scientists do science, and traders trade....
<b>RE: too many variables to find the deterministic motion:</b> I'm with darkhorse on this one. Although, I certainly do not believe that we can create a 100% accurate predictor of market action, we can create profitable predictors/rules for when to buy & sell. I especially like the notion of finding 'clarity pockets' and would say that both discretionary and mechanical traders should look for these. Certainly, there are "random" effects that perturb the price action. The "random" arrival of buyers and sellers is guaranteed to create minor order imbalance transients (Queuing theory is useful for thinking about this one). News creates so-called exogenous shocks that cause market participants to reevaluate the fundamental value of financial instruments (we'd need a separate thread to discuss whether most really are exogenous or not). I'd even bet that data storms on the internet have an effect by slowing down the trader's cycle of information gathering and trading action. These and other effects then participate in feedback loops that may be positive (e.g., a minor, totally random cluster of buyers leads to a short-term rally and subsequent fall-back on a stock). The examples from science of randomness (Brownian motion, radioactive decay, etc.) are useful for thinking about some aspects of the markets, but they have crucial limits. Whereas an air molecule has no incentive to either hit or avoid a dust mote, traders have strong incentive to hit or avoid a bid. Whereas a coin has no knowledge of its history, traders are acutely aware of the price history of their tradables. That traders (and investors) share common and knowable thought processes and incentives determines the extent that price action follows common and knowable dynamical patterns. <b>RE: Descriptive vs. predictive rules:</b> As darkhorse indicates, this is a tricky area. Its too easy to misapply middle-of-the-chart, descriptive rules (the after-the-fact why) to trading. And, its very hard to maintain a right-edge-of-the-chart mentality when looking at historical data. Just about any kind of backtesting will be prone to survivorship bias. For example, I'd bet that not many traders are backtesting any bullish trading systems against WCOM or ENE historical data -- its too obvious and too easy to avoid data on stocks that we now know did not make it. <b>RE:the distribution:</b>The distribution of price changes (returns) of stocks and stock indexes is nearly normal. But its not quite normal, much to the consternation of believers in the central limit theorem (and an excellent example of darkhorse's comment about the difference between academia and the markets). The distribution of returns differs from normality in being both skewed and heavy-tailed (kurtotic). Some researchers have used what are called stable Pareto-Levy distributions to model the fat-tailed distributions of returns. Others have turned to the t-distribution. Another aspect of the distribution of price changes is that the variance (volatility) changes over time and, historically speaking, returns have a positive bias. If you want to know more and are very mathematically inclined, look at "A Non-Random Walk Down Wall Street" by Lo and MacKinlay or the older, somewhat overlapping "The Econometrics of Financial Markets" by Campbell, Lo, & MacKinlay. And if you want something a bit more introductory, try "Quantitative Methods in Finance" by Watsham and Parramore, although I have found errors in that book. I think this modification of Arthur C. Clark's quote says it best: "Any sufficiently complex pattern is indistinguishable from randomness." This is the essential conundrum for those concerned about whether the markets are random (and all traders should be concerned with this). Nobody can prove that the markets are random, they can only prove that they fail to conform to any of a long and growing list of models. The problem is that along the way, its too easy to find deterministic models that seem to work (for a while) I don't believe the markets are random and I hope that I am right. In the meantime, I'm working to ensure that I am not Fooled by Randomness (or at least not too fooled by it. ). Until I learn that the markets are random, I'm Traden4Alpha P.S. I was first intrigued by alpha when I studied modern portfolio theory. But, more recently, Ben Warwick's very interesting book called "In Search of Alpha" did indeed inspire the nickname.
Now we are appearing to make a little progress in this thread. Traden4Alpha tells us that stock data is not quite normal, but skewed with a large tail- Maybe a Pareto-Levy distribution or a t-distribution. So who will volunteer to tell us(the uninformed) what is a Pareto-Levy distribution and a t-distribution and how do you use it with stock prices?