You need to ask the right question to get the right answer. The wrong question that I see most of the foolish out there asking: Is the market random? And, they're thinking primarily about using price to predict the future price. Their thinking about the market in a very limited way. Better questions to ask: What's influencing the market right now? Where is the market going tomorrow? You see something can be predictable but not be dependent on the historical price data. Look at the big dude, what was his name Pauslon? who shorted the housing market. There was nothing in the price that told him to short. All the scientist would say well nothing here: just pure random. But, there was something. There was something that allowed him to predict something that had never occurred which was the ability to anticipate correctly how the market was going to react in the future. He had the ability the model possible future scenarios. I do believe price data can be predictive. But, it can so much more predictive when you can anticipate using hypothesis based reasoning to go with it. In other words, if you can anticipate something happening then you can also monitor to see if it is happening. Most losers are looking backwards while the leaders are looking forward. They are anticipating. Sometimes they anticipate too far ahead which is why every trader has to manage risk. One trader here asked about tape reading as a means too know when the big trader is making a move. Well, if you know what the big trader is looking at and you know how he's going to make his decisions then you can move before he even makes his trade. You can be the big trader. Or you can combine such a hypothesis with the ability the ability to read the market and that's powerful stuff.
I have banned prediction from my strategies (as nearly everybody knows ;-)) ). On the other hand, i do think that price is a factor which exercises some influence on the future price trajectory (although predicting it is another matter). People (and automated systems, consequently) do take decisions based on price, and that, in turn, affects the evolution of the price curve. You are right pointing out that also exogenous factors do contribute. Tom
Here is a repost from my blog, "The Obvious Secret" There is a simple concept that underlies all active trading. It is not a secret and should be obvious. Yet, many traders and some gamblers do not understand this concept. The obvious secret is: The only advantage one can have in an active trading (or gambling) system is a predictive advantage. The implication is that the only way to make money by actively trading in the market is to have some concept of the future. This simple fact is the fact that should drive everything. Most vendors do not want you to understand this, and most people don't want to hear this because they don't believe the future can be predicted. Predicting the future sounds very difficult. You are betting on the future when you make a trade. Expectancy is defined empirically by the win percentage times the average win size minus the loss percentage times the average loss size. But, it is defined essentially by the predictive advantage. Active traders must produce a profit to break even. A trader must overcome the spread, commissions, datafeed charges, and any opportunity costs to turn a true profit. A positive expectancy does not require one to have a greater then 50% probability of a winning trade. However, one must still have a predictive advantage even with a win ratio of less then 50%. In other words, the probability of getting a big winner a small portion of the time will essentially be the same as getting a smaller winner a greater portion of the time if one doesn't possess a predictive advantage. The expectancy will still be zero! This idea that somehow that someone can profit with a 50% or less win ratio does not change the number one rule, the rule that supercedes all rules which is: The only advantage one can have in an active trading (or gambling) system is a predictive advantage. The only statements that one can use to profit from the market will be predictive in nature. A lot of traders don't get it. They don't get the whole point of futures is to predict the future. And very few commentators, if any, will make truly easy to test, verifiable, quantifiable, statements on the market besides for myself! What is edge? Edge is the ability to forecast positive expectancy situations. It is the ability to anticipate that a trade or group of trades will produce a profit. Edge embodies the sentiment that the default expected outcome is not going to be profitable but that select outcomes may be profitable. Many gamblers become mesmerized by martingale systems. Martingale systems are basically systems that vary bet size. Often, a martingale system will use double bets or triple bets to turn a loser into a winner. Martingale systems gain higher win percentage at the cost of suffering catastrophic losses. They do not change expectancy. Virtually all casino games are negative expectancy. A player who plays a negative expectancy game will tend to lose more the more they play. And, this makes sense as the more one plays then the more likely they will be to realize the statistically implied result -- better known as the law of large numbers. There is a lot of focus on optimal position sizing strategies for traders and systems. No position sizing strategy will change the expectancy of the trades. Yet, they can change the growth rate of the account. It is conceivable if the trades are autocorrelated then the differences in account growth could be significant. But, the more important message is that, again, you can't gain an advantage where one doesn't originally exist. If it is easier to produce a profitable but low winning percentage system then that would suggest certain characteristics of the market. The obvious take away would be that markets tend to trend and that trends tend to continue. Yet, a slightly diffeent way of looking at this would be similar to Taleb's idea that markets tend to undeprice rare, unlikely events thus making it possible to profit from those events. You see if the market does trend that indicates the market was not efficiently moving to those prices which would provide opportunity for the trend follower. As a final word on this issue, we know that to profit that one must have a predictive advantage but it doesn't imply that one must have information from the future to anticipate it. The advantage could be an informational advantage. As an example, it is possible to predict where a car will be that is moving at a given rate of speed merely by knowing the speed, location, and orientation of the car. This is an example of a deterministic system where the next state is completely determined by the previous state
@Fullatuotrading If you have "banned prediction" then you have necessarily "banned" or ruled out the one concept that allows for any profits to be made. Look at the following strategies Directional or momentum trading Goal is to predict when momentum will continue or reverse Event based trading Predict the outcome of the event and how it will influence the market Pair trading/spread trading Predict pairs that are statistically more likely to mean revert Limit orders/Reversion Predict the true distribution of a price series, i.e what limit orders represent unfair prices Trend following/Black swans Predict large profit opportunities at a rate greater then chance (i.e rare events) Stock picking Predict the industries or companies that poised to out perform Global macro trading Predict how worldwide macro economic events will shape future profits ----------- The only trading methods I know that don't require prediction would be pure arbitrage (i.e market making) and possibly rebate trading (negative or flat expectancy + rebate profit = positive). One could also view the long term growth rate of the market as being prediction free (market returns). Some professors have raised strong objections to this notion of stocks for the long run. The only other ways where can get prediction free returns is by "cheating", i.e selling options premium (diametrical risk/reward) or martingale strategies. While I don't know enough about selling premium, a purely martingale strategy is rather dubious and to be implemented effectively requires a prediction of the distribution (imagine averaging down in 1987). One can view the prediction as the outcome of a statistically significant number of events or on an individual basis. The distinction will determine how one approaches the market (placing tons of small trades vs betting bigger on fewer trades) but doesn't really change the requirement. One can choose to predict future direction of price series (i.e directional trading) or the relationship between two prices (pairs trading). One can focus on predicting the profitability trades (profit factor) or winning % or some other criteria but at the end of the day, a prediction must be made. A trade is a prediction. A prediction is a testable statement about the market. For one to make any statement about the market that is meaningful (can be tested) then it is necessarily a prediction.
Personally, I disagree with the above apodictic view [not because of philosophical beliefs, but simply because i know it does not correspond to (my) experience.] There are more facets to the question, and other ways to build an hedge ;-) In my view, it's actually more likely the opposite: it's the attempt itself to "predict" which often leads essentially to a chaotic game and zero (or negative) profits (especially in automated systems)... Tom
This is not a matter one can agree or disagree with. It is a matter that is shown to be self-evident. It doesn't need to be proved, it is an axiom that is self demonstrated which is why I find it quite hilarious that snake-oil vendors manage to skirt the issue. If I'm wrong then show me an example where one isn't required to predict something that can produce a profit (at a rate greater then chance). Other then the cases I asserted, there are not examples where one can produce a consistent profit without prediction. The only example I can think of would be buying in a bull market. You could say, okay, buy any new high and then sell next high. No prediction required. The problem is while it is true that one can make a profit using such a method: it doesn't constitute an edge because when the market goes down then all those profits will be lost. One might say the market tends to rise more often then it falls but then you have a prediction. I'm not stating that one can't produce a profit without a predictive edge. A few lucky people hit it big in Las Vegas I'm sure every year. Of course, they are the exceptions. There is no rational expectation to win without a predictive advantage.
Well, i do not think the question reduces to being "wrong" or "right". More simply, based on actual experience, i have always been trading based on "no prediction" assumption (within automation). And i have usually exceeded other automated approaches, based on prediction. see for instance (for some last thread): http://www.elitetrader.com/vb/showthread.php?s=&threadid=222126 This does not signify that personally i don't believe in "prediction". It only means that i don't use it within automated strategies (while i do allow discretionary user intervention to possibly "inject" predictive beliefs and i do think they can, sometimes, improve the outcome on a single realization of the price curve). Weird as it may seem, my current personal belief is just the opposite, that is: sistematically using (or, better, attempting) "prediction", more likely eliminates the possibility of being sistematically profitable (especially with reference to ats, and allowing exceptions on discretionary approaches). I guess that diversity of opinions is what makes the world rich. Tom (PS. btw why the use of the plural in the thread title?)
@fullautotrading Tom, yes I've looked at your system before. While I don't know much about it, it looks impressive. Thanks for sharing! However, I'm stating it doesn't matter what you believe. Prediction is built into your system either implicitly or explicitly. I was coming at this primarily from the angle of the discretionary trader because most auto traders know prediction is built into the system. You are basically fading the market using limit orders, right? If so there is an implicit assumption to your model that price is more likely to reverse after it goes beyond a certain threshold. This implies that price has a certain distribution. The market is known NOT to have a normal distribution which the normal z-scores do not apply. If you are holding open position and then adding new positions then you are using martingale with a normal or non normal distribution implicit, such a system has a small but very real probability of blowing out (tons of c2 forex systems blow out due to just this reason). It is similar to the opposite of the guy buying insurance (any sort of insurance). Most of the time insurance is a cost but it only takes needing it 1x to pay for all the prior expenses. If you are using martingale and fading moves and have a certain account balance, then the position sizes you feel safe to use are based on implicit (and likely wrong ) prediction models. I.e crash of 1987. Some stupid things traders say: I trade what I see. Implicit prediction: There is a momentum to the market and the market will continue in that direction. I don't predict. I trade the trend. Implicit prediction: Again, it implies the market is more likely to continue to move in the direction it has been moving then it is to reverse. Where "has been moving" is very subjective as any 2 arbitrary points will yield a different meaning. It may also imply that the market misprices abnormally large market movements. This is probably more true of why trend following works based on the low win ratios. There is a predictive statement though, namely the market will misprice a large movement enough to pay for all my losses. PS: As for why "The Logics", I added the s to make it distinct. To make it clear, I'm not dealing with Logic but a set of discreet logical considerations. The S makes it unique and clear when people may refer to it, "The Logics" because it is an improper usage except as a reference to my writing. While improper to Englishaticians, it helps from an information-entropy perspective.
Yep. Nobody uses martingales for trading, unless suicidal. It's not so simple. By appropriately overlaying several strategic layers, one can arrive at a point of "balance", where the system will self-hedge. It's not easy and intuitive, nor something achievable in just few years of work. But it's a possibility. For the rest, there is always the possibility of human discretion and judgment ... Again, let me say that in my personal view there is no "prediction" explicitly or implicity. I see it differently. I view it more like the building of a "probabilistic cage". It's not a predictive view. I would describe it more like "absolute of time": past, present and future are all simultaneously immanent. But that is mostly philosophy: after all, all what matters is making $$$. ;-)) which is an objective and inexorable judge and helps us evolving everyday our ideas and making progress ... Tom
Tom, I find your ideas very interesting and would like to learn more. However, I also find that you are "couching" these terms in quite a bit of complexity. When you talk about hedging and multiple instruments then it is reasonable that there are several assumptions (or statements) such a strategy makes about the market, i.e model characteristics. Some of these would be: Correlation or lack of correlation of markets Co-integration or lack of co-integration of markets Auto-correlation or lack of auto-correlation of markets Statements about stationary ----- Can you expand on what you mean by "hedge"? Do you mean if you have a losing trade to take the opposite in a correlated instrument? It sounds like you might be making statements about the probability of runs (consecutive wins and losses), if so this is also making statements about the auto-correlation of an instrument and is likely a variant of martingale. Even if you are just making statements about which instruments to hedge then you are assuming (i.e making statements about) future correlation values. It sounds like you may be taking something from portfolio theory which states if you combine low or non correlated assets you get a better risk adjusted returns. This theory applies to assets with a positive bias-- not to random price distributions.