Quantitative and Algorithmic Trading | Page 8 | Metals Mine

Options

Go to
Permalinks
Print Thread

Similar Threads

Algorithmic/automated trading and emotions 13 replies

Learn Computer and Data Science with Algorithmic Trading 5 replies

Attachments: Quantitative and Algorithmic Trading
Exit Attachments

Quantitative and Algorithmic Trading

  • #141
  • Aug 12, 2014 4:41pm Aug 12, 2014 4:41pm
Experts online: An analysis of trading activity in a public Internet chat room

http://www.sciencedirect.com/science...67268109000341


Abstract
We analyze the trading activity in an Internet chat room over a 4-year period. The data set contains nearly 9000 trades from 676 traders. We find these traders are more skilled than retail investors analyzed in other studies. 55 percent make profits after transaction costs, and they have statistically significant αα s of 0.17 percent per day after controlling for the Fama–French factors and momentum. Traders hold their winners 25 percent longer than their losers. 42 percent trade both long and short, with equal success rates, and almost double the profit per trade when short. The estimates show a strong influence from other traders, with a buy (sell) order 40.7 percent more likely to be of the same sign if there has been a recent post. Traders improve their skill over time, earning an extra $189 per month for each year of trading experience. They also gain expertise in trading particular stocks. Traders who raise their Herfindahl index by 0.1 raise their profitability by $46 per trade.
  • #142
  • Aug 13, 2014 4:51am Aug 13, 2014 4:51am
Quoting infinitus
Statistical analysis of candlestick patterns http://bspace.buid.ac.ae/bitstream/1...1/1/100068.pdf Do Japanese Candlestick Patterns Help Identify Profitable Trading Opportunities? An Analysis on Selected Forex Markets By Aisha Ahmed Ameen 100068 Dissertation submitted in partial fulfilment of the requirement for the degree of MSc in Finance and Banking Faculty of Business Strong evidence of the bullish hammer, the bearish hanging man and the shooting star patterns in some currencies is found. The research finds that the frequency of...
I have a problem with this kind of studies which is the way they define the predictive value of a signal. For the engulfer candlestick pattern for instance how shall I determine whether there is predictive value or not?

Here is an example taken from E/U daily in Aug 2012.
Attached Image (click to enlarge)
Click to Enlarge Name: EURUSD-D1.png Size: 12 KB


If I define that a bullish engulfer "works" by checking if the next day is UP then the #1 fails at being bullish despite it is followed by a 800 pip upward move. The bearish engulfer #2 is considered correct because it is followed by a DOWN move of 7 pips. But the volatility at that time was ~100 pips per day. 7 pips are quite random in comparison.

If I define that a bullish engulfer "works" by checking if the price moves up more than the size of the pattern before it goes down below we see that the #3 works despite it is followed by 4 days of congestion. Not very bullish.
No greed. No fear. Just maths.
  • #143
  • Aug 13, 2014 8:10am Aug 13, 2014 8:10am
Quoting PipMeUp
{quote} I have a problem with this kind of studies which is the way they define the predictive value of a signal. For the engulfer candlestick pattern for instance how shall I determine whether there is predictive value or not? Here is an example taken from E/U daily in Aug 2012. {image} If I define that a bullish engulfer "works" by checking if the next day is UP then the #1 fails at being bullish despite it is followed by a 800 pip upward move. The bearish engulfer #2 is considered correct because it is followed by a DOWN move of 7 pips. But the...

When I remeber correctly, most of those studies look for the return after various periods (after the signal candlestick formation occurred), like 1 day, 3 days, 14 days etc.

But you are right. It is a hard task to perform such studies in a clean way. One has to define many things beforehand, but then the traders gut feeling, which could change in the holding period, is excluded (and that could make a strategy work or not)

I found it interesting, that most of the pattern only give very small edge, and literally only a handful of patterns were able to give a bigger edge.
  • #144
  • Sep 3, 2014 4:59am Sep 3, 2014 4:59am
These papers are about volatility estimation. They describe a phenomenon they call the incoherent price formation. I didn't know about this and perhaps others don't either. Hence I share. In a nutshell: the estimation of the volatility is strongly biaised at the high frequency resolution because the price doesn't immediately reflect all the available information. The quotes are no longer the price but a noisy observation of the price and the variance of these observations is greater (~2x) than the variance of the price itself. As if the price was a hidden (latent) random walk observed through very noisy observations.
Attached File(s)
File Type: pdf 020221-efficientVolEstimator.pdf   263 KB | 1,038 downloads
File Type: pdf corsi_2001.pdf   204 KB | 1,130 downloads
No greed. No fear. Just maths.
  • #145
  • Sep 3, 2014 9:17pm Sep 3, 2014 9:17pm
Quoting PipMeUp
You talk about MTF and fractals. I propose you an experience to show that the market is not exactly fractal but fractal-LIKE. Open a daily chart on any pair. E/U for example everybody watches this one. Add two MAs. Any kind SMA, EMA, LWMA, funny-whatever-zero-lag-MA. One is applied to the High and the other to the Low. Use the same period, a big value to have the result very smooth. You get a channel. Squeeze the chart to get a rather big sample. Mesure the width of the channel at different places. Trending periods, ranging periods... Quickly estimate...
Thank you =) This is what I had been stressing to others over the years and hence gave up using 'standard timeframe candles'. My Pip Time efficiency studies already show the same results as you described.
  • #146
  • Sep 3, 2014 9:24pm Sep 3, 2014 9:24pm
Quoting infinitus
http://www.forexfactory.com/showthread.php?p=5317784 The Optimum Time Frame for Trading Anyone?
Very good study. Although I wouldn't call the lower timeframes noise thou.
  • #147
  • Sep 3, 2014 9:47pm Sep 3, 2014 9:47pm
Algorithmic Trading
http://numericalmethod.com/introduct...ng-strategies/

Introduction to Algorithmic Trading Strategies

Note: this course is superseded by the CQI introductory course: Introduction to Quantitative Investment

Course Description
This course introduces students to quantitative trading. A trader usually starts with an intuition or a vague trading idea. Using mathematics, s/he turns the intuition into a quantitative trading model for analysis, back testing and refinement. When the quantitative trading model proves to be likely profitable after passing rigorous statistical tests, the trader implements the model on a computer system for automatic execution. In short, quantitative trading is the process where ideas are turned into mathematical models and then coded into computer programs for systematic trading. It is a science where mathematics and computer science meet. In this course, students study trading strategies from the popular academic literature and learn the fundamental mathematics and IT aspects of this emerging field. By working on the class projects, they will gain hands-on experience. After satisfactorily completing this course, the students will have the necessary quantitative, computing, and programming skills in quantitative trading. They are therefore well prepared for a front office role in hedge funds or banks.

Course Outline
Session Topics Readings Homeworks
1 Overview of Algorithmic Trading Lecturer handouthw1
2 Hidden Markov Trading Model Lecturer handouthw2, labprogramming a hidden Markov chain
3 Pairs Trading by Cointegration Lecturer handouthw3, labprogramming a cointegration model; parameter sensitivity analysis
4 Optimal Pairs Trading by Stochastic Control Lecturer handouthw4
5 Pairs Trading by Stochastic Spread Methods Lecturer handout, labprogramming a Kalman filter; parameter calibration
6 Technical Analysis: Linear Trading Rules Lecturer handout
7 Portfolio Optimization Lecturer handouthw5, labstrategy & portfolio optimization
8 Risk Management Lecturer handout

Recommended Readings
Numerical Method’s collection of Quantitative/Algorithmic Trading Literature

Student Projects

  1. The quantitative trading strategies group projects for FE8827, 2011, Nanyang Technological University.
  2. The quantitative trading strategies group projects for QF5205, Fall 2011, National University Of Singapore.

Hope it helps!
I wish you good-luck!
FXcube

  • #148
  • Sep 4, 2014 1:37am Sep 4, 2014 1:37am
Quoting PipMeUp
What you mesure is roughly the average volatility of these TF. H4 is 6 times faster than daily but the volatility is only 3 times smaller. H1 is 4 times faster than H4 but only half the volatility. H1 is 24 times faster than daily but the volatility is only 8 times smaller. A TF is not exactly the replicate of another up to a scaling factor.
Hi PipMeUp, do you think it is possible to have a random process where volatility scales linearly with the timeframe?
  • #149
  • Sep 4, 2014 5:09am Sep 4, 2014 5:09am
Quoting kprsa
{quote} Hi PipMeUp, do you think it is possible to have a random process where volatility scales linearly with the timeframe?
A one-dimensional Wiener process
No greed. No fear. Just maths.
  • #150
  • Edited 6:01am Sep 4, 2014 5:42am | Edited 6:01am
Quoting PipMeUp
{quote} A one-dimensional Wiener process
I believe I have misunderstood you, your measure of volatility is ~variance. I was thinking about the standard deviation.
Sorry.
kprsa
  • #151
  • Sep 4, 2014 7:19am Sep 4, 2014 7:19am
By volatility I indeed mean variance because I don't really care the units are homogeneous. But I get your point that the distance between the 2 MA is in the same unit as the price series and therefore the standard deviation should be used. You're right that there is no reason they scale linearily. But notice they don't scale in square root either!

A process where the standard deviation increase linearily with time needs the variance to increase with the time squared. This requires the variance of the innovations to increase linearily with time. So no time invariant process can do that.

(+1 for a very good observation)
No greed. No fear. Just maths.
  • #152
  • Sep 5, 2014 6:47am Sep 5, 2014 6:47am
Quoting PipMeUp
These papers are about volatility estimation. They describe a phenomenon they call the incoherent price formation. I didn't know about this and perhaps others don't either. Hence I share. In a nutshell: the estimation of the volatility is strongly biaised at the high frequency resolution because the price doesn't immediately reflect all the available information. The quotes are no longer the price but a noisy observation of the price and the variance of these observations is greater (~2x) than the variance of the price itself. As if the price was...

Hi PipMeUp,


thanks for sharing
  • #153
  • Nov 22, 2014 4:50am Nov 22, 2014 4:50am
Quantum Mechanics Made Easy
Chapter 11, "Quantum Time," of
From Eternity to Here:
The Quest for the Ultimate Theory of Time
Sean Carroll

http://www.preposterousuniverse.com/...ohere/quantum/
  • #154
  • Nov 22, 2014 9:46pm Nov 22, 2014 9:46pm
Quoting infinitus
Quantum Mechanics Made Easy Chapter 11, "Quantum Time," of From Eternity to Here: The Quest for the Ultimate Theory of Time Sean Carroll http://www.preposterousuniverse.com/...ohere/quantum/
Ok you now have my interest. Based on your posts I thought there was a bit of a physicist in you. Sean Carroll is a high calibre physicist who keeps away from any 'woo'. His pop science writings are always a good read. If you are interested and haven't already, have a read of Amanda Gefter's 'Trespassing on Einsteins Lawn'.. Not a physicist per se but a gifted scientific journalist who picks the brains of some contemporary talent out there in string theory and loop quantum gravity....anyway enough about speculative quantum gravity theories and back to this game..... :-)
  • #155
  • Nov 22, 2014 10:13pm Nov 22, 2014 10:13pm
Quoting Copernicus
{quote} Ok you now have my interest. Based on your posts I thought there was a bit of a physicist in you. Sean Carroll is a high calibre physicist who keeps away from any 'woo'. His pop science writings are always a good read. If you are interested and haven't already, have a read of Amanda Gefter's 'Trespassing on Einsteins Lawn'.. Not a physicist per se but a gifted scientific journalist who picks the brains of some contemporary talent out there in string theory and loop quantum gravity....anyway enough about speculative quantum gravity theories...

thank you very much


will try to get the book
  • #156
  • Nov 22, 2014 10:43pm Nov 22, 2014 10:43pm
Quoting infinitus
Analysis about the Forex Trading Volume {image}
Nice quote
Trading 24/h
  • #157
  • Nov 22, 2014 10:58pm Nov 22, 2014 10:58pm
that is interesting:

http://www.zerohedge.com/sites/defau.../picture-5.jpg
One Of The Most Striking Equity Market Anomalies Explained


snipset:

It is surprising how little attention academic literature has devoted to understand equity market returns around the turn of the month, despite the observations of Lakonishok and Smidt (1988) and McConnell and Xu (2008) among others that most of the returns accrue during a four-day period, from the last trading day to the third trading day of the month.
We find that the market returns are abnormally high also on the three days before the turn of the month.
In fact, combining the two observations, we find that since 1926, one could have held the S&P 500 index for only seven business days a month and pocketed almost the entire market return with forty percent lower volatility compared to a buy and hold strategy.
  • #158
  • Nov 24, 2014 3:36am Nov 24, 2014 3:36am
Time series prediction with Long Short Term Memory

https://en.wikipedia.org/wiki/Long_short_term_memory
  • #159
  • Nov 24, 2014 3:37am Nov 24, 2014 3:37am
Long Short-Term Memory: Tutorial on LSTM Recurrent Networks

http://people.idsia.ch/~juergen/lstm/
  • #160
  • Nov 24, 2014 3:40am Nov 24, 2014 3:40am
Jürgen Schmidhuber's page on
Recurrent Neural Networks
(updated 2014)
Why use recurrent networks at all? And why use a particular Deep Learning recurrent network called Long Short-Term Memory or LSTM?

http://people.idsia.ch/~juergen/rnn.html
0 traders viewing now