Fuzzy logic trend filter

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
Chelonia
Roundtable Knight
Roundtable Knight
Posts: 497
Joined: Mon Apr 30, 2007 3:37 pm

Fuzzy logic trend filter

Post by Chelonia » Wed Jun 20, 2012 3:05 am

Would be interesting to find out more about the fuzzy logic trend filter Michael J. Clarke/Clarke Capital Management is using.

One common feature across all programs is their use of filters to weed out trades during potential choppy periods. One such filter Clarke believes is unique to their programs. He calls it the Fuzzy Logic Filter, and says: “The fuzzy logic trend filter examines trends from very short to very long and every point in between for that commodity, and sees what has worked in the past and it makes a decision whether to go, or not to goâ€

trackstar
Roundtable Fellow
Roundtable Fellow
Posts: 76
Joined: Sun Mar 09, 2008 11:03 pm
Location: Houston Texas

Post by trackstar » Wed Jun 20, 2012 12:09 pm

If anyone has details I would love to help model it.

squaredQ
Full Member
Full Member
Posts: 23
Joined: Wed Feb 01, 2012 5:27 pm

Post by squaredQ » Wed Jun 20, 2012 8:03 pm

The model isn't the difficult part. It's deciding what factors determine a choppy/trending or whatever class of market you are currently in and furthermore, how much these measures help determine a greater likelihood of continuation.

So, IMO, it's more important to focus on factor selection criteria as well as issues like how to beneficially discretize the data first. I prefer to start the process via statistical methods foremost.

stopsareforwimps
Roundtable Knight
Roundtable Knight
Posts: 199
Joined: Sun Oct 10, 2010 1:47 am
Location: Melbourne Australia

Post by stopsareforwimps » Wed Jun 20, 2012 10:10 pm

squaredQ wrote:The model isn't the difficult part. It's deciding what factors determine a choppy/trending or whatever class of market you are currently in and furthermore, how much these measures help determine a greater likelihood of continuation.

So, IMO, it's more important to focus on factor selection criteria as well as issues like how to beneficially discretize the data first. I prefer to start the process via statistical methods foremost.
I find myself being repeatedly lured by the siren songs of "picking the turns" and "distinguishing trending from choppy markets".

Even Ed Seykota made some remarks that suggested he tried to distringuish choppy from trending markets.


I've never found any evidence that I could do either of those things.

Macro
Roundtable Fellow
Roundtable Fellow
Posts: 70
Joined: Sat Dec 03, 2011 11:22 am

Post by Macro » Thu Jun 21, 2012 1:10 am

There does exist measures to identify persisting (trending) and mean-reverting (choppy) states of markets across numerous time-frames simultaneously. It involves fairly high level mathematics and to those inclined, I shall provide this hint: geometric time-series.

Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Post by Chris67 » Thu Jun 21, 2012 2:33 am

YES But that misses the pint somewhat .. markets are mean reverting until they arent and they are persistently trending till they aint !
bit like teh old correlation argument .. no level of maths can tell you when correlation is about to change

I think the case in point with Clarke Capital is an interesting one - theyve had a blow away month so everybody rushes to see what their secret is ? because of one month ? Arent Clarke Capital teh guys who had a 50/60 % draw down in 2009 ? and were virtually out of business - presumably using the same logic that they can identify which markets are likely to trend and which arent (absolutely impossible)

The truth is that TRADING BLOX equips you with enough tools to design simple robust trend following systems - because they dont work for 18 months doesnt mean theyre not robust

I think there is a fair amount of performance chasing going on out there at the moment

squaredQ
Full Member
Full Member
Posts: 23
Joined: Wed Feb 01, 2012 5:27 pm

Post by squaredQ » Thu Jun 21, 2012 3:24 am

Macro wrote:There does exist measures to identify persisting (trending) and mean-reverting (choppy) states of markets across numerous time-frames simultaneously. It involves fairly high level mathematics and to those inclined, I shall provide this hint: geometric time-series.
There are several ways to measure this without even reverting to high level math. For example, just find a program to calculate hurst exponent; there should be a correlation between what we perceive as trending (somewhere >>.5) and reverting (somewhere <<.5). There's stuff like Kaufmann efficiency ratio as well.

The key to look at is how to find whether or not the likelihood of that measure is expected to continue in the near future. That's where probability and statistics comes in and I gather he's using some variant of this.
Fuzzy Logic is just one way of approaching the root problem of filtering for reliable signals-- but without understanding how to properly find and pre-process the input factors first, its pretty much useless.

wguan
Roundtable Fellow
Roundtable Fellow
Posts: 64
Joined: Sun Feb 27, 2011 1:52 pm

Post by wguan » Thu Jun 21, 2012 11:47 am

I find Kaufman's efficiency ratio a pretty good trend / mean treating indicator. One which is modified is called trend strength index by engineering returns blog.

Like chris said, i don't think it's possible to accurately predict when it will trend and when it won't. But we sure can identify environment s that will increase the possibility of trending by looking at past prices: momentum.

Post Reply