couple days ago, a friend of mine suggested i try trading a stock based on moving averages. he said if ma(5 day)>ma(20 day) buy and vice versa.i did monte carlo, and found that not only it didnt give good results, but in any setting it underperformed buy-and-hold. and that's without transaction costs.does anyone here have some suggestions or readings on this topic. i would be thankful.

is it a joke?

Ahh, it seems that you have found a way to make limitless profits. Simply get a prediction from the strategy that you described and then do the exact opposite.

Try www.wealth-lab.com there you can try 100´s of scripts using all kind of moving averages, and you will probably reach the same conclusion as in your MC.So why read more on the topic?

Moving average crossovers work "well" in a trending market. They work "badly" in a sideways market.

I have recently tested some moving average strategies, based on typical technical trading rules, the results of the event studies were mixed, though generally showed little significance. Though again in trending markets there is some bias(i.e/ favorable marginal effects), but it is difficult to discriminate between competing hypothesis in trending markets. Though I am not a chartist at heart, RSI did show overwhelmingly positive results even when isolated in two cases; above 70 and below 30, though trading costs would eat up some profits, inevitably, if this weapon of choice is to be fully exploited. In any event I was not convinced intuitively, intellectually, or empirically by MA crossover strategies. Though I should point out that I am not an expert and my trading experience is limited, I still agree with Soros's thoughts in his Alchemy regarding feedback loops, technical analysis can certainly serve as a catalyst-at least from a cognitive/behavioral standpoint. Just my 2 cents.

Yep maybe it's a joke... But anyway, it's a pity that 99% of the trading litterature on the net is about using technical indicators like MA's... According to my experience this kind of systematic strategies doesn't work! These strategies involve optimizing the MA's values for high profit, high sharpe ratio ('curve fitting') but... these parameters are highly instable on most of the markets! So even if you curve fit and get a nice equity curve, if you use your model on the next period, you will loose everything!

Actually trend following systems do work over the long term, because markets occasionally get very autocorrelated. Thing is you have to have strategies running on a whole bunch of relatively uncorrelated markets at the same time, because in trend following most trades will lose money, and then one trade will make a ton. So diversification helps make it practical. In trend following, your returns will look like you are long an option plus a bunch of noise (a slightly fat right tail). You get lots of little losses because of your stop loss policy (MA strategies have an automatic one) and the occasional big gain when you actually catch a trend. Mean reversionary strategies look the oppoiste - like you are short an option. You have a fat left tail, so you have to make little profits very often to offset the occasional big loss. These tend to work better on spreads than on normal markets.Regarding optimising backtesting, if you use too complex a system and overfit to history your system will be more likely to fail. You need to explore the parameter space around your optimisation. if the P&L is a sharply peaked function of a parameter, then a small change in mkt statistics will cause your system to fail - you need a different system. If your backtested P&L is a broad function of all your parameters then you may be on to something. You should also take an old time series, optimise, and see how it works in the next time period. If it fails, this tells you something too.No systems work all the time, but trends do sometiems appear, so the philosophy is that you are mostly taking little losses but when you hit a real trend you more than make up for it.Nearly any persistent deviation from zero autocorrelation will allow you to make money through some strategy. But it may be too volatile a P&L, or the gains too few and far between to make it a viable hedge fund strategy.

QuoteOriginally posted by: erstwhileActually trend following systems do work over the long term, because markets occasionally get very autocorrelated. Thing is you have to have strategies running on a whole bunch of relatively uncorrelated markets at the same time, because in trend following most trades will lose money, and then one trade will make a ton. So diversification helps make it practical. In trend following, your returns will look like you are long an option plus a bunch of noise (a slightly fat right tail). You get lots of little losses because of your stop loss policy (MA strategies have an automatic one) and the occasional big gain when you actually catch a trend. Mean reversionary strategies look the oppoiste - like you are short an option. You have a fat left tail, so you have to make little profits very often to offset the occasional big loss. These tend to work better on spreads than on normal markets.Regarding optimising backtesting, if you use too complex a system and overfit to history your system will be more likely to fail. You need to explore the parameter space around your optimisation. if the P&L is a sharply peaked function of a parameter, then a small change in mkt statistics will cause your system to fail - you need a different system. If your backtested P&L is a broad function of all your parameters then you may be on to something. You should also take an old time series, optimise, and see how it works in the next time period. If it fails, this tells you something too.No systems work all the time, but trends do sometiems appear, so the philosophy is that you are mostly taking little losses but when you hit a real trend you more than make up for it.Nearly any persistent deviation from zero autocorrelation will allow you to make money through some strategy. But it may be too volatile a P&L, or the gains too few and far between to make it a viable hedge fund strategy.Good post.