Anyone knows any software for LPPL? So hard to write everything by myself.
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Sornette LPPL program
- Thread starter GloriaBrown
- Start date
There is some R Code at R-SIG-Finance:
https://stat.ethz.ch/pipermail/r-sig-finance/2009q3/004639.html
https://stat.ethz.ch/pipermail/r-sig-finance/2009q3/004639.html
There is some R Code at R-SIG-Finance:
https://stat.ethz.ch/pipermail/r-sig-finance/2009q3/004639.html
Thanks for your link. I saw this before and look like far from complete. He was asking question and wonder why it didn't work.
From what I understand, the code is actually working
So he gets stuck in the first local minimum in an unnecessary blown up search space.
Do you expect much from LPPL?
I found Sornettes trading simulations rather discouraging.
but crude. He is optimizing over 7 parameters instead of what Sornette proposes in his latest papers (slaving 4 parameters).Though the parameters may be
not within the range that LPPL papers defined, the curve fitting is OK.
So he gets stuck in the first local minimum in an unnecessary blown up search space.
Do you expect much from LPPL?
I found Sornettes trading simulations rather discouraging.
From what I understand, the code is actually working
but crude. He is optimizing over 7 parameters instead of what Sornette proposes in his latest papers (slaving 4 parameters).
So he gets stuck in the first local minimum in an unnecessary blown up search space.
Do you expect much from LPPL?
I found Sornettes trading simulations rather discouraging.
I expect a lot ha ha. Yeah it should be 4 parameters instead of 7. How do you try out LPPL?
Is that irony (i am not a native speaker)?I expect a lot ha ha.
A few years ago we tested it at my former employer. Wasn't really successful.
Sornette published a paper where he combined it with pattern recognition to do actual trading. Could not find it on his page but the results were not impressive.
Anyway you can't use R or matlab to program it properly because you need to intercept gaussian elimination and simplex optimization. Take "numerical recipes" then it should not take more than a week.
Is that irony (i am not a native speaker)?
A few years ago we tested it at my former employer. Wasn't really successful.
Sornette published a paper where he combined it with pattern recognition to do actual trading. Could not find it on his page but the results were not impressive.
Anyway you can't use R or matlab to program it properly because you need to intercept gaussian elimination and simplex optimization. Take "numerical recipes" then it should not take more than a week.
You work in a trading firm so you do backtest for your job?
Not anymore, but this was a part of my last job.
Can you give me some suggestion on how to test if a trend is coming/stop? I still cannot find a good research paper about this yet.
Not really. Maybe regime switching with hidden markov models. Or this stuff:
http://asiaetrading.com/tobias-preis-founder-artemis-capital-video-interview-with-asiaetrading/
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2260189
http://asiaetrading.com/tobias-preis-founder-artemis-capital-video-interview-with-asiaetrading/
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2260189