Ron, I think you are confusing some terms here and to be honest, I think most people do. Edge is not the same thing as being profitable or having a profitable return. I've often joked on ET if that was the case then buying a CD from your local bank could be construed as having edge. Or hell even having a job. After all you "expect" to get a positive stream of cash flows from your employer right? Edge is something entirely different. First let me start off by giving a general definition between using a "technical" approach vs a "quantitative" approach. Technical traders, those who look at patterns or moving averages or doji's are looking to "prove" something works. They do things like "backtest" their pattern and if it yields a positive return you might hear them say they have edge or something like that. But the genesis of their approach is trying to convince themselves (or others) that their indicator or strategy works. The quantitative approach is the opposite. Generally speaking, quantitative math or analysis is actually trying to "disprove" or prove that something does NOT work. Think about to college stats and doing hypothesis testing. You are trying to reject the null. Why would you want to disprove your work? To see if it's robust! I think you used that word. The idea is to qualify your results to determine what role luck possibly played in the outcome. This is where the "expected value" comes in along with one of my favorite words "variance". The idea with quantitative analysis is NOT to solve for expected value. That has no meaning because we have no idea if that expected value is random. So the idea here is to expose your data to rigid "quantitative" analysis and try as hard as you can to prove you are wrong. This in contrast to "technical" analysis where you are working diligently to prove you are "right". Let me make note here that simply proving you are not wrong does not mean you have an edge or that you will even be profitable. The example often used in school is a jury trial. You cannot prove someone is innocent. You can only prove that they are not guilty based on the evidence presented. OJ Simpson for example was determined to be "NOT" guilty. That does NOT mean he was innocent. That's how data analysis works. It's impossible to prove with absolute certainty that something does work but you certainly can prove it doesn't. Or you can prove that it's likely your results are due in large part to the variance of the data. In other words, dumb luck. And as I'm sure you know, there is a mighty fine living to be made convincing people you have something when a high degree of variance is involved. So expected value alone does not reveal anything to you. Neither does your avg winner or avg loser on a backtest. This is often why people who brag about technical analysis get looked down on because their approach is not only unscientific, but it's seriously prone to bias. You are "trying" to make the data work. That is your stated goal. You often "see" what you want to see. It's why you can have two analysts on CNBC looking at the same chart and they are both technicians and one is very bearish and the other is very bullish looking at the same chart. Both are seeing what they want to. I'm not trying to advocate for one approach over the other. I have my own theories on this stuff. But to be fair, it's usually "technical" guys that misuse the concept of expected value. They run a back test on trade station or meta trader and it spits out a positive expected value using their moving avg cross over technique and they jump for joy that they found gold. They basically found exactly what they wanted to find. Not a very robust way to go about things I would say. So to sum up, the prudent way to test things would be to try to prove something does not work. There are many techniques and many variations. Just keep running through the meat grinder as if your goal in life is to expose yourself as a fraud. After some time, while you may never prove you have something that absolutely works, you will find at the very least, that it's highly likely your results did not come from mere chance. There is some other variable driving your results. The fun part comes when you find that variable and begin to optimize it. Sorry for the long explanation.
Fluctuate yes, but if you have a zero or a negative mean, then you don't have a positive expectancy system do you? Of course you would have to have data from enough different market cycles to be reasonably sure, but that is what proper backtesting is for. Scalable is relative, all depends on how big line you are swinging.
This is a bogus dichotomy between TA and QA. The whole idea of backtesting is to disprove a given strategy, be it TA-based or QA-based. The evidence of this are the many, many TA books with countless indicators, PA whatever, etc. and nary a backtest in sight, because backtests would reveal just how bankrupt almost all of these techniques (TA or QA) really are. TA "works" only because <1% of it works, i.e., it is not literally 100% garbage. I don't know if any QA works at all or if they're all still stuck in the paralysis-by-analysis phase of their 'science'. That is the current state of technical trading. Got nothing to do with how you feel about backtesting.
Dude, chill out and go back down to P&R. I never said that TA works or doesn't work. My God you have anger issues.
Seriously you want to compare P&R post counts? Yeah, I didn't think so. And evidently you can't handle the slightest disagreement without hallucinating anger issues. Get help, dude.
Open excel, put all your trades in a workbook in dollar win-loss. Take average, standard deviation. To get Pr(win) just do this for your winners (ie. trades with P&L >0) the opposite for Pr(loss) (ie. trades with P&L < 0) Once you have that use z-score formula to figure out the probability that a trade will be a loss or gain. http://en.wikipedia.org/wiki/Standard_score
I don't think I'm confused at all. You wrote a full page to argue against what I wrote but you still did not provide a definition of an edge. You just used a false dichotomy, as Kut2k2 said, to make some point but your point is subjective and the way you perceive things. Mathematically speaking, making money is strictly equivalent to having a positive expectancy. If you have no edge, at some point the expectancy turns negative. Maintaining profitability is mathematically equivalent to having a positive expectancy. Now, I argued that this is not enough because although it is true after being profitable, i.e. after the fact, relying on computed expectancy based on historical data does not guarantee future profitability as everyone and his brother knows. The rest are irrelevant to this fact. Many authors have concentrated on expectancy without making it clear that this is a random variable and non-stationary due to market returns arising from a non-ergodic process. Therefore it makes sense to look for other measures of an edge based on robustness or antifragility and that was my only point which is based on objective mathematics. I understand the dichotomy between TA and QA you are making but it is irrelevant to me because it is irrelevant to the math. The math accounts only for returns, they do not care what method you used to generate them, you can use the stars or a card reader for all I know, it is irrelevant to the classical definition of an edge, which is equivalent to having a positive expectancy over the period of returns. Although there is some true in that in essence you are mixing up emotions and perspectives with math. No matter how you trade, TA or QA, or what other method you are using, if you average your trades you must have a positive number, i.e. a positive expectancy. Ultimately this is the metric that measures your success or edge but only after the fact and you have no other choice. My point was that you cannot use that before the fact and both TA and QA are based on wrong foundation. BTW, TA tries to get a high expectancy by cutting losses sort and letting profits run while QA tries to do that by a proper combination of win rate and payoff ratio but at the end of the day there is no difference as far as this notion is concerned although QA goes further into analyzing the robustness of expectancy the idea is still ill-conceived from start because the expectancy is not-stationary and changes with changing market conditions. You may not care that much if you are trading 5k in forex but if you have a 100 billion fund you are looking for antifragility rather than high expectancy.
+1 +1 QA also "works" because <1% of it works but it is not the current state I believe as the majority of traders are using s&r, patterns and moving averages. If you are not convinced pay a visit o quantopia or stocktwits. You will have a lot of fun...