I've read that trend-following strategies perform well in high-volatility regime... and I've also read opposite statement saying that trend-following strategies perform well in low-volatility regime...But I seem to find it hard visualize intuitively either case.It seems to me "volatility" is not a good dividing line for good/bad performance of trend-following strategies.I am thinking of "auto-correlation"... but not so sure. If "auto-correlation" is a good dividing line for the good/bad regimes, then is there a way to "forecast" auto-correlation?How about mean-reverting strategies?

Well I don't think anyone has a silver bullet for when/where trend-following vs mean-reverting works. At least if they did they should be sipping margaritas on a yacht in the Riviera. Here though is how I think about it. Auto-correlation whether positive or negative is a function of two things. The first is order flow persistence, i.e. it's a well known fact that if there are more buys than sells in one period, than one would expect all things being equal more buys than sells in the next period. The second is order flow elasticity, i.e. as the price rises there will be fewer buyers and more sellers and vice versa. The former is the logic behind trend-following. The latter is why that logic doesn't always hold, if there was heavy buying action that pushed the price up it might dry up buyers and flood sellers in the next period pushing the price down.I think this paradigm goes a long way towards explaining why trend-following works in certain markets and mean-reversion works in others. For example take one of the most well-known mean-reverting markets, the idiosyncratic return component of equities (i.e. the returns of single stocks neutralized against the major market factors). This formed the core of many stat-arb strategies for a very long time. And with very good reason, one would expect order flow to be highly elastic for idiosyncratic equity factors. Most equities outside of their core factor exposures are very close substitutes for each other. So investors have very little tolerance if the idiosyncratic returns drive prices away from fair value. E.g. if the idiosyncratic component of one firm in the S&P 500 was higher than the others most investors would substitute to the others very easily.In contrast one of the best markets for trend-following tends to be commodities. And with good reason, the buyers and sellers are very price inelastic. For example if the price of oil goes up, consumers can't quickly substitute away from using crude oil. There are massive infrastructure changes that need to be made. Similar if say wheat is falling, it's not like farmers can easily switch over to growing corn without making large capital expenditures. So low elasticity leads to momentum, because when there's an imbalance between buyers and sellers it takes longer for prices to readjust to a point to bring markets back to equilibrium.

Lets say we have 4 regimes:LowVol, Down Market: Just punt LowVol, Up Market: MOM performs wellHighVol, Down Market: MR performs wellHighVol, Up Market: MOM & MR performs okayThis is just an observation I made from my trading strategies over the last few years. I have chatted with several people and most of them agree with this. But there will be exceptions based on the individual models people use.No matter what regime you are in if you are trading MR or MOM then from a portfolio POV you better have strategies that take advantage of both.$.02

I believe the idea is that trend-following strategies can be effectively modeled using lookback straddles, which comes from a famous paper by Fung & Hsieh, "The Risk in Hedge Fund Strategies: Theory and Evidence from Trend Followers". http://faculty.fuqua.duke.edu/~dah7/RFS2001.pdf. It's been a while since I read it, but you may find some intuition as to why this is so in that paper.

if it's too late today to read the whole fung & Hsieh's paper, here is an outline of the strategies described.sure one will know more reading the paper ...

QuoteOriginally posted by: undergrad86Well I don't think anyone has a silver bullet for when/where trend-following vs mean-reverting works. At least if they did they should be sipping margaritas on a yacht in the Riviera. Here though is how I think about it. Auto-correlation whether positive or negative is a function of two things. The first is order flow persistence, i.e. it's a well known fact that if there are more buys than sells in one period, than one would expect all things being equal more buys than sells in the next period. The second is order flow elasticity, i.e. as the price rises there will be fewer buyers and more sellers and vice versa. The former is the logic behind trend-following. The latter is why that logic doesn't always hold, if there was heavy buying action that pushed the price up it might dry up buyers and flood sellers in the next period pushing the price down.I think this paradigm goes a long way towards explaining why trend-following works in certain markets and mean-reversion works in others. For example take one of the most well-known mean-reverting markets, the idiosyncratic return component of equities (i.e. the returns of single stocks neutralized against the major market factors). This formed the core of many stat-arb strategies for a very long time. And with very good reason, one would expect order flow to be highly elastic for idiosyncratic equity factors. Most equities outside of their core factor exposures are very close substitutes for each other. So investors have very little tolerance if the idiosyncratic returns drive prices away from fair value. E.g. if the idiosyncratic component of one firm in the S&P 500 was higher than the others most investors would substitute to the others very easily.In contrast one of the best markets for trend-following tends to be commodities. And with good reason, the buyers and sellers are very price inelastic. For example if the price of oil goes up, consumers can't quickly substitute away from using crude oil. There are massive infrastructure changes that need to be made. Similar if say wheat is falling, it's not like farmers can easily switch over to growing corn without making large capital expenditures. So low elasticity leads to momentum, because when there's an imbalance between buyers and sellers it takes longer for prices to readjust to a point to bring markets back to equilibrium.I guess we are talking about the coexistence of trend-following and mean-reverting in the same markets?

QuoteOriginally posted by: SkipI believe the idea is that trend-following strategies can be effectively modeled using lookback straddles, which comes from a famous paper by Fung & Hsieh, "The Risk in Hedge Fund Strategies: Theory and Evidence from Trend Followers". http://faculty.fuqua.duke.edu/~dah7/RFS2001.pdf. It's been a while since I read it, but you may find some intuition as to why this is so in that paper.Did he say anything about mean-reverting?Also, I found that we have vagueness when we talk about strategies. What's a typical trend-following? Maybe a simple Moving Average strategy is a trend-following strategy. Then it's easier to analyze. What's a typical "mean-reverting" strategy?

I don't recall them talking about mean-reversion, but it's been quite a while since I read it. It's not a tough paper. I think a model using the difference of two moving averages would be a classic and simple trend-following strategy. My understanding is that currencies tend to trend best, or used to. There are tons of technical analysis books with these sorts of ideas in them. Not sure how effective any of them actually are, however. A classic mean-reversion strategy might be some sort of short-term return like the intraday bid-ask bounce, to take an extreme exalmple, or monthly stock returns to take a simpler one. The residuals to a linear equity factor model might qualify.

There is a nice master thesis about modeling the regime switching using hidden markov model.

When you look at a single cycle of a sine wave do you see 2 mean reversions or 3 trends ?The only way I know of differentiating between the two is where you would put a tp or sl.If tp < sl , its MR, if its tp > sl its trending, if tp = sl you don't care and you made your money on the entry.There is no reason why you can't see different things in different time frames and hedge one against the other.I'd be happy to hear others views on this.

QuoteOriginally posted by: ktangThere is a nice master thesis about modeling the regime switching using hidden markov model.where is it?

QuoteOriginally posted by: Skip My understanding is that currencies tend to trend best, or used to. There are tons of technical analysis books with these sorts of ideas in them. q]what are some factors that affect whether a security/product tends to revert or trend? why does fx tend to trend vs others?