I tested a number of trendfollowing models on randomly generated price series data (normal distributed price changes) in Excel. It produced an expected profit of exactly ZERO, just as expected. Then I tested them on 25 years of financial data and they produced superior risk/reward profits compared to buy & hold market by market. If prices were in fact perfectly random that would in turn mean markets were efficient and ANY (!) trading approach would be ineffective in generating any alpha. Value investing, merger arbitrage and global macro just as much as trend following. If you think markets are random why are you trading at all? Ever think of that? The conclusion is that markets are neither random nor efficient.
This where the problem is, its not hard to find systems that 'work', the same way its not hard to find a system that works in that random chart and allows you to capture the trend. Its likely you havent tested enough systems. Unless the random price sample is ultra large, there will be always systems that will 'work'. Furthermore if the random sample is ultra large the 25ys of market data is probably proportionally a lot smaller Markets ARE random a lot of the time, if you have no edge in predicting the minute changes from moment to moment then you dont know what will happen which means its the same as being random even if after the fact it looked obvious that so and so was going to happen. If you have to guess whether the next day ES will go up or down that is pratical terms the same as having to guess the outcome of a coin toss after you adjust for the upward drift of stock prices. But sometimes you can get an edge by better understanding reality(fundamentals) or doing a historical statistical analysis that you believe other market participants are not aware, that shifts the edge of the coin toss and makes it biased
And that's EXACTLY what trend following is all about. A tiny statistical edge. As long as the majority of market participants trade on "market top and bottom calls", fundamental analysis or according to simple bubble/panic herd behavior then trend following will work.
What I see a lot is traders discrediting the trading approaches of others. * Scalpers ridiculing quants * Value investors ridiculing growth investors * Counter trend traders (VN) attacking trend followers * Trend followers ridiculing value investors Few of the above have ever testing nor worked with any of the strategies they criticize. Yet they feel they know all the reasons why their strategy is superior to all the others and why the other strategies by definition can't work. I am trading a number of different mechanical strategies. Amongst them, one statistical counter trend and one future trend following. No strategy is "better" than the other. They all have different pros and cons. All try to exploit different market inefficiencies. Talking oneself into the belief how what one is doing is so much better and how the rest of the world has it all wrong and doesn't know what they're doing is self-delusion IMO.
Check back my posts. Trying to show how TF 'doesnt work' or is inferior to Global Macro wasn't the reason I revised VN points, in fact I was defending that approach a few months back regurgitating the TF arguments, mostly the 'gut feel' one, which is a bad one as a random chart would produce the same thing I want to find out reality, I dont care about how painful it is to acknowledge it. If you show me an automatic IB API that predicts the intraday direction of ES and returns 100% a year using horoscope, I will trade it. Then move to the Bahamas and not read about the global economy again, I will be thrilled
I'm assuming of course the horoscope system predicts the ES with sure high level of accuracy its likely not to be random
Here's reality, from my humble perch: * Markets are neither random nor efficient * Value investing/merger arb/trend following/global macro/relative value etc. all are working more or less well at exploiting these inefficiencies * No strategy is better or worse than all the others all the time * IMO, making consistent excess risk adjusted returns in any of the above strategies is profoundly hard business * If you want to lie at the beach doing nothing then forget about alpha and simply stick with beta. Doesn't mean beta is worthless. And lastly, I wouldn't look at VN for proof on why long-term trendfollowing is impossible just as I wouldn't look at a high-frequency quant for an explanation as to why long-term value investing is dead. Is it so hard to accept that all strategies have their individual merit?
The idea one shouldn't 'look at VN proof on why long-term trendfollowing is impossible' is like saying one shouldn't look at a scientist that is saying water boils at a certain temperature(though there are differences between social and natural sciences). He is making statiscal claims about TF, so far I have yet to see them debunked, but I have read many times that I should 'just close your eyes and keep believing on it, it works, trust me, look at these big boys. stop reading the guy, he blew up'. The quality of that reasoning is so low, my intution tells me its the side that is likely to be wrong Yes, there are different ways to generate alpha, I accept that. I never meant to imply something else. I do however think that GM is more robust than being a one trick SSNR pony, that doesnt mean I wont engage on SSNR tricks if I believe they will continue. It simply means I will always be paranoid about when the edge party will end, which the big TF boys seems to think they are immune to
I've said it before. VN looks at the SP500 or Nasdaq index, tests if buying X consecutive positive daily closes makes any money. He finds it doesn't and hence declares all trend following (which trades 70+ markets, not just one stock index) is a losing strategy over all time frames. No trendfollower is debunking this. Why would one? No trendfollower is buying the SP500 future just because it closed up 8 days in a row. That's non-sense. Yet it is enough for VN to make broad statements. On top of that, many commodity markets include term structure effects (contango/backwardation) and thus are idiosyncratically vulnerable to inefficiencies. VN never even looked at them! Yet he dares to make broad statements about them. Is that the rigorous statistical analysis you're referring to? The best part was when the Blackstar guys presented their "Does trendfollowing work on Stocks?" paper a couple of years ago. It covered survivorship bias, it covered slippage, it covered transaction costs, it covered dividends. In other words everything VN claimed the trendfollowers were "missing". And yes, it still increased risk adjusted returns. VN's reply: "Interesting, this warrants more research but I still think it doesn't work because you can't make money following random prices." LOL The matter of the fact is that VN's analysis is a joke as are his conclusions.
I actually agree with VN regarding the blackstar paper. Look at the chart of the returns against the S&P500. There is virtually no alpha from 1990 to 1999. Then from 2000 to 2008 there is, that system tends to be long smaller cap stocks compared to the S&P500, the outperformance period is also a period where smaller caps beat big caps. I need to see a chart of the TF stock system against a chart of an index of all stocks traded in the US. I doubt the outperformance(if it exists) is big enough to justify all the hassle(not to mention that it could be random)