Apr

16

The reasons trend following doesn't work are myriad including ever changing cycles transactions costs, and bid asked spreads, the opportunity to game the system against them, and the ease of triggering mechanical rules and the fact that markets are homeostatic, and supply curves change as prices move up or down.

Ed Stewart writes: 

In my opinion, part of it is that people who mostly trade their own money look at IRR or "cash on cash" returns, and thus see issues of gains and losses more clearly vs. those who only look at marketing documents and time-based returns of recently hot funds.

Larry Williams writes: 

Trend following does not work on just one (or 2, 3, or 4) instrument. Trend followers have to have a large basket of 'bets' on the assumption that someplace in the world a market will trend and that one massive trend—think CL this year and last—pays off the other bets.

It's like betting on all the numbers in Roulette one number pays big odds. Trend followers say they cannot predict which number will show or market will trend, but with enough numbers bet, one will win.

Stefan Martinek writes: 

Larry, you make a great point. TF is more risk/exposure management on a basket than trading. Argument that benefits of diversification end after ~20 markets is such a nonsense (my teacher said that too together with other corporate finance theories; they probably never tested anything outside of equities).

Diversification across groups, styles, markets, and time frames improves risk adjusted returns in a long run. Of course in a short run concentration is great - let's bet all on Apple. TF has a nice barrier of entry which is good: First, some money is needed; second, most operators cannot run 2 years without rewards if necessary. They quit. Philosophically it is somewhere between "systematic macro" and "private equity". In PE you expect that most bets will be a crap unless you are in LBOs and other later stage deals. You expect that some areas will be in slumps maybe for years. Patience is such a great thing if one can afford it.

Orson Terrill adds: 

Well if I hadn't unnecessarily deleted all of my old code I would just spit out some examples… I wrote several functions that tested trend following, and mostly what was observed was that the number of intervals (days, weeks, whatever) in which a trade would be open generally follows an exponential distribution.

For those that do not know what that implies: Lets say trend "A" has been going for 5 days, the probability that the next day will be the end of trend "A" is roughly the same as if trend "A" were 1 day old, or 20 days old. The next day the probability of "A" ending is generally the same, regardless of its age (like a Poisson process for the arrival of the end). The general notion that longer trends are more, or less likely to end, due to their age, is not backed up.

Just because a run is multiple days old/young does not mean it was profitable. In many markets nearly half of the period's range is traded through during the next period, on average (I think this is true on almost all scales in the EUR/USD, but its been a while). This means getting in on momentum greatly increases the likelihood that a trade is entered at such a point where near term downside is slightly more likely than near term upside (assuming its a long equity position).

There were marginal improvements through adding responses to measures of volatility(mostly changes in absolute ranges), rates of change of price medians from multiple length of time intervals, and most significantly in the general case: reversing intra-trend can garner a couple tenths of 1%. Specifically applying those while using several time series which switch regimes in the sharing of strength of running correlations in percentage changes like SPY, TLT, and GLD, might have some interesting results (I eat what I kill, so I had to leave it there).


Comments

WordPress database error: [Table './dailyspeculations_com_@002d_dailywordpress/wp_comments' is marked as crashed and last (automatic?) repair failed]
SELECT * FROM wp_comments WHERE comment_post_ID = '10208' AND comment_approved = '1' ORDER BY comment_date

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search