Quantitative and Algorithmic Trading | Page 3 | 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

  • #41
  • Jun 24, 2013 5:12pm Jun 24, 2013 5:12pm
Quoting PipMeUp
{quote} I have another point of view. But that's the whole idea of an open discussion I read somewhere that the market are ranging 80% of the time. I still don't understand if that means 8 minutes every 10 minutes or 8 months out of 10 months (makes a difference!). But for the sake of the explanation let's imagine that it implies the market is MR 80% of the time. Well, let's even go as far as pretending that the market is MR 90% of the time. I feel generous today. Let's use Bayes' theorem and let's define N the event "we are under a mean reverting...

Hi Pipmeup,

will come back to that later in this week - simply too busy at the moment


Wish you many green pips in the time between :-)
  • #42
  • Jun 24, 2013 5:17pm Jun 24, 2013 5:17pm
http://www.sovereignman.com/nsa-black-paper/

How To Give The NSA The Finger…Without Them Ever Noticing
  • #43
  • Jun 26, 2013 3:31pm Jun 26, 2013 3:31pm
Quoting infinitus
{quote} Hi Jack, here is some good definition about quant and algorithmic: Trade carried out using algorithms is known as algorithmic trading. Algorithmic trading can be defined as “placing a buy or sell order of a defined quantity into a quantitative model that automatically generates the timing of orders and the size of orders based on goals specified by the parameters and constraints of the algorithm”. The rules built into the model attempt to determine the optimal time for an order to be placed that will cause the least amount of impact on the...
thanks for the link really nice read.
  • #44
  • Jun 27, 2013 2:18am Jun 27, 2013 2:18am
you are welcome :-)

Quoting MikeySym
{quote} thanks for the link really nice read.
  • #45
  • Jul 17, 2013 3:04am Jul 17, 2013 3:04am
test
  • #46
  • Jul 17, 2013 4:19am Jul 17, 2013 4:19am
Quoting PipMeUp
{quote} I have another point of view. But that's the whole idea of an open discussion I read somewhere that the market are ranging 80% of the time. I still don't understand if that means 8 minutes every 10 minutes or 8 months out of 10 months (makes a difference!). But for the sake of the explanation let's imagine that it implies the market is MR 80% of the time. Well, let's even go as far as pretending that the market is MR 90% of the time. I feel generous today. Let's use Bayes' theorem and let's define N the event "we are under a mean reverting...

Hi PipMeUp,


I do feel so sorry to have taken so long to answer. Had too much work besides of trading. I have a few days off, so I will try to answer.

Really interesting what you have wrote


I do calculate my Market Profiles with rolling periods. On all of my TF's I calculate the Market profile with a lookback period of 32.

When my observation is right (only visual inspection), than especially on 15min TF we can often see, that price reaches the +4 or even +5SD (fat tails) and is then unable to mean revert completely.

Then higher TF's are, so to say, in play, and I switch to 1h or 4h Market profile to judge the distribution curve there.


In general I take those TF's, where the curve has the best shape for making the basic judgements.


No I am coming back to mean reverting:

I use the fractal-model of the markets. So we can assume, that we can observe mean reverting on some TF's, when other TF's establish another level (hence: trending).

Furthermore, I do think, to have eliminated the danger of fat tails by using rolling periods. Well, the levels or SD's I do calculate are floating. One have to get used to it.

But with this MTF floating Market profile approach I can work very well.

Yes, it is hard to say, if this approach would stand mathematical/statistical tests. I have read a lot about that stock markets etc. do , most of the time, only show some kind of mean reverting.

My feeling is, at least for currencies, that the various TF's are mean reverting quiet often, especially when one uses rolling periods.

Well, as you said, maybe I only trade the white noise and not really the MR. I do not know right now. Especially since I use my MR system now only for a few month.


You gave the nice example of the +3SD. In Forex trading, mainly EURUSD, I normaly use as threshold the +4SD, since I think, that we do oftentimes see fat tails. But anyway: above +3SD or + 4SD I do not expect the price to mean revert but to at least revert a bit. In a case, where price touches the +4SD, e.g., I am betting not at the mean with a short position, but will take profit maybe at the +2SD level.



But what about you: do you trade MR? When yes - can you describe your system a bit?
  • #47
  • Jul 17, 2013 4:24pm Jul 17, 2013 4:24pm
Very happy to see you back posting!

Quoting infinitus
Furthermore, I do think, to have eliminated the danger of fat tails by using rolling periods. Well, the levels or SD's I do calculate are floating. One have to get used to it.
Something between multiple Bollinger Bands and/or a Keltner channel, I imagine? Your "mean level" or "zero line" drifts. This drift is someone else's trend. And what you trade is noise for him. Look at the picture below. That's a electrocardiogram. Nothing to do with trading you may say. Look at it carefully you'll see that the patterns don't repeat exactly at the same level. They are like carried by a very long period sine wave. This sine wave is due to the patient breathing during the exam. That's not related to the heart: THIS IS NOISE! Noise is not always high frequency.

I agree with you when you write that it's unimportant if the mean reverting component is white noise (MR per definition) or anything else. But this "noise" becomes your signal and you have to filter out the trend (now a nuisance), what you do using a sliding window of 32 bars. The trend becomes YOUR NOISE!
Attached Image (click to enlarge)
Click to Enlarge Name: ecg1.gif Size: 45 KB
No greed. No fear. Just maths.
  • #48
  • Jul 17, 2013 4:26pm Jul 17, 2013 4:26pm
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 the average (roughly no need to calculate it for real)

Using SMA 100 I get something like 150 pips (more or less) on E/U daily
Do the same on H4. I get something close to 50 pips.
Do it on H1. 20 pips.
M15. 15 pips.

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. It's not because you cannot visually distinguish them that they are the same.
It means that a volatility SL is relatively bigger on a smaller TF. The smaller the more "noise". => The R:R is impacted.
But that's not exactly true. Weekly has more in common with hourly than it has with daily.
Think of it next time you read "This system works on any timeframe" in the Trading Systems section
No greed. No fear. Just maths.
  • #49
  • Jul 17, 2013 4:31pm Jul 17, 2013 4:31pm
Quoting infinitus
You gave the nice example of the +3SD. In Forex trading, mainly EURUSD, I normaly use as threshold the +4SD, since I think, that we do oftentimes see fat tails. But anyway: above +3SD or + 4SD I do not expect the price to mean revert but to at least revert a bit. In a case, where price touches the +4SD, e.g., I am betting not at the mean with a short position, but will take profit maybe at the +2SD level.
This is interesting as well. Let's look at it differently. From a trend trader point of view. You're selling a overbought market. But statistically speaking you are doing it in a uptrend. You are counter-trend trading!

Let me explain why I consider CT is bad with the picture below. In blue I marked what you trade: the retracement from +4SD to +2SD. I cherry picked a strong trend to purposely exagerate the picture and make things more obvious. I added a Fibo. When you enter a position you use a 1SD as a SL either direction. So the risk is the same for both the directions you trade the MR. But you can see that the pullback is only 40% or 50% of the in-trend move. For the same risk, the potential reward is much smaller. The R:R is worse.

Let's imagine you go long at the top of this blue swing with no SL at all. Eventually, when the trend resumes you are back in the money. Should you do the same shorting at the low of the retracement, no way you get your money back! A short in a up trend a low probability position.

=> Lower the R:R, lower the win/loss => much lower the expectancy.

Counter-trend trading may be profitable but it is sub-optimal! Instead of swing trading both directions it is better to double the bet in-trend and not trading the retracement at all. You get more profit for the same value at risk.
Attached Image
No greed. No fear. Just maths.
  • #50
  • Jul 17, 2013 7:23pm Jul 17, 2013 7:23pm
Quoting PipMeUp
{quote} This is interesting as well. Let's look at it differently. From a trend trader point of view. You're selling a overbought market. But statistically speaking you are doing it in a uptrend. You are counter-trend trading! Let me explain why I consider CT is bad with the picture below. In blue I marked what you trade: the retracement from +4SD to +2SD. I cherry picked a strong trend to purposely exagerate the picture and make things more obvious. I added a Fibo. When you enter a position you use a 1SD as a SL either direction. So the risk is...

Yes, it is true. Trading mean reverting is not a non-fragile method, as I mentioned in one of my earlier postings.

But the fallacy of thinking in trend trading is: after the fact, the trend on a chart looks obvious. You know then exactly, when to trade :-)

But in real time trading, you do not know that beforehand. Every retracement can in fact develop to a full blown reversal.



Well, in reality, there are two hearts in my chest:

I trade the MR or countertrend opportunities on shorter TF, and I try to get early in in long, multiday/multiweek swings.

The critical phase for a trend trade is the first few days, when you can not be sure, if you got the right timing. After the trade develops, one can work with a convenient SL and manage the trade with trend following tools.


I for one do find it much more easy to get reward from MR trading than from trend following. I simply think, it is due to the fact, that consolidations/sideway movements dominate the market time wise. So it is easier to find opportunities.
  • #51
  • Jul 18, 2013 5:22am Jul 18, 2013 5:22am
Quoting infinitus
I for one do find it much more easy to get reward from MR trading than from trend following. I simply think, it is due to the fact, that consolidations/sideway movements dominate the market time wise. So it is easier to find opportunities.
You MR trade because "sideway movements dominate". This is a marginal probability (http://en.wikipedia.org/wiki/Marginal_distribution). Should you know that the market is going to trend today, would you use your MR strategy?
If you could use the recent PA + MTF analysis to improve the estimation of the likelihood of the event "market will range today" (http://en.wikipedia.org/wiki/Conditional_probability) you could, at least statistically, adapt or turn on/off your MR strategy.

The main part of the job of a quant/algorithmic/discretionary trader is not to pull the trigger when a signal occurs but is to estimate the market context and how it fits his set of strategies.
If you estimate that the market has a high probability to be forming a flat triangle, what is the best way to trade a flat triangle?
If you estimate that the market has a high probability to be trending, what is the best way to trade a trend?
If you estimate that the market has a high probability to be ranging in a rectangle, what is the best way to trade a rectangular range?
... a falling wedge?
... a bullish flag?
... a bullish flag which is against the weekly down trend?
...

That's exactly what I'm trying to do. I try to answer the question "Can I make my computer estimate the market context ("read" the PA)?" in order to select the most efficient strategy accordingly. Can it make it with a good enough acuracy? Can it do it before the profit opportinity is gone?

Quoting infinitus
[trend trading] But in real time trading, you do not know that beforehand.
A trend is a persistent bahaviour. If the market is trending, statiscally it will continue for some more time. On the other hand a trend follower FOLLOWS the trend he doesn't FORECAST the trend. If there is no trend there is nothing to follow. Per definition a trend is a continuation of the move. For something to continue it has to have already begun. Per definition it is a total non-sense to try and catch the "beginning" of the trend. Such a thing doesn't exist!

Quoting infinitus
Every retracement can in fact develop to a full blown reversal.
True. That's why I think trend following is better executed by trading each of the swings separatly instead of holding a melting position hoping for the resumption of the trend. That's what they do in Trend Trading All Pairs (http://www.forexfactory.com/showthread.php?t=208630): when they consider (with 3 MAs) that the market conditions are favorable they trade each swing. The last one fails. No big deal.
No greed. No fear. Just maths.
  • #52
  • Jul 18, 2013 5:24am Jul 18, 2013 5:24am
Quoting infinitus
After the trade develops, one can work with a convenient SL and manage the trade with trend following tools.
My attempt at trailing the stops. Recent Aussie H4, pSAR added for comparison.
That's NOT a stop-and-reverse indicator! That's not a strategy by itself.
Attached Image (click to enlarge)
Click to Enlarge Name: stopTrailer.png Size: 39 KB
No greed. No fear. Just maths.
  • #53
  • Jul 18, 2013 2:04pm Jul 18, 2013 2:04pm
The main part of the job of a quant/algorithmic/discretionary trader is not to pull the trigger when a signal occurs but is to estimate the market context and how it fits his set of strategies.


How do quant forecast what type market is next trading session for every European open ?

Do they use any unknown algorithm or hi-tech methods to project a trend or a range of market for next session?

On Tuesday, wed, Thurs, trend often unfold in intraday.

There is no indicator or code for calculation of mean and mean shift in MT4. So, it is a hard part for target to be set for MR strategy.
  • #54
  • Jul 18, 2013 3:02pm Jul 18, 2013 3:02pm
Quoting Jack168
The main part of the job of a quant/algorithmic/discretionary trader is not to pull the trigger when a signal occurs but is to estimate the market context and how it fits his set of strategies. How do quant forecast what type market is next trading session for every European open ? Do they use any unknown algorithm or hi-tech methods to project a trend or a range of market for next session? On Tuesday, wed, Thurs, trend often unfold in intraday. There is no indicator or code for calculation of mean and mean shift in MT4....

Hi Jack, hi PipMeUp,



well, first let me be clear, that I am in no way an Algo-trader, who does perform automated trading. My trading style is more kind of algo-assisted


I draw all the conclusions and carry out the trades manually. Just want to make that clear, not that some think it would be other :-)


My strategic decisions are based on Elliott wave and S/R.

The algos I use are intended to give me a judgement for probabilities at the levels I determine. In this sense, the above red-marked statement from PipMeUp fits to my approach.


I have problems most of the time with the old adage, that you can make more money by simply following the trend. Most of the time, a critical point remains unaswered:

As a trend follower: at which points are you willing to take a trade?

Oftentime the answer is: after confirmation


But where one sees a confirmation in the aftermath, in real life trading the thought to be confirmation are often only part of more complex corrections. So the trader get whipsawed and often enough stopped out, only to see, that the trend finally was set forth, after he was stopped out.

Confirmation comes with the disadvantage, that you can not reap in the easy pips, when price runs into support or resistance, and the gains are practically "guaranteed".


My short term trading style aims for this "easy" part of the pips. I do not need 50-100 pips to fulfill my goals. Normally 2*20 pips per day are enough for me.

In reaching for this, my algos assist me in my approach.



Algos/Quant Trading has many faces, so any trader can pick the tool he thinks may need.


As I told, sometimes I am going after the really big swings via trend-following. In counter-moves to this trend, when I think, due to Elliott Wave analysis, that the correction is a bigger one, I simply take hedging positions, but let the core position run. In doing so you have to not care about reentering the trend but to only care and concentrate about to close the hedging position. In my experience, this has worked great for me.


The first part of Jacks question should answer PipMeUp, because he seems to mainly use trend-following. The algos I do use are MR tools.


Regarding MT4 indicators:

you only need a data feed, that streams the data either real time or with little delay into Mathlab or Excel, for example, and perform there the analysis.

You can reach this via Indicator or via DDE


If you like to go the DDE way, and are not able to programm a DDE bridge, do a search for "Mexcel Trader". But that one is not free and you have to pay around 250 USD for it, when I remember correctly. I think one or the other broker offers that for free, when you open up an account. But I have no experience with Mexcel.
  • #55
  • Jul 19, 2013 1:45am Jul 19, 2013 1:45am
Quoting PipMeUp
Very happy to see you back posting! {quote} Something between multiple Bollinger Bands and/or a Keltner channel, I imagine? Your "mean level" or "zero line" drifts. This drift is someone else's trend. And what you trade is noise for him. Look at the picture below. That's a electrocardiogram. Nothing to do with trading you may say. Look at it carefully you'll see that the patterns don't repeat exactly at the same level. They are like carried by a very long period sine wave. This sine wave is due to the patient breathing during the exam. That's...

Hi PipMeUP,


nice analogy with ECG - like it


Yep, the noise, e.g. on a 4h TF (let us call it small correction), may be new "trend" on a 5min chart



Sometimes we can see this fairly good on Market Profile charts, when the bell shape is not achieved or rather looks horrible - do not know how to describe it


well, next time I took pictures of it
  • #56
  • Jul 19, 2013 1:49am Jul 19, 2013 1:49am
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...

Yes, very true. Money Management, SL, expectations has to change on each TF

Yes, it is fractal-like. Would it be 1:1 fractals, trading would be more easy :-)
  • #57
  • Jul 20, 2013 4:22am Jul 20, 2013 4:22am
Quoting Jack168
How do quant forecast what type market is next trading session for every European open ?
Quantifying is about mesuring or estimating things. It's not necessarily about forecasting.


For instance I show here how I try to find the best pair to trade: http://www.forexfactory.com/showthread.php?t=435706 (work in progress)
I forecast no price. I just estimate the likelihood of the current trends on the various crosses to estimate the money flow.

Also forecasting is much less interesting than you may think. Do you want to know what price E/U will close on Monday?
Easy! Here is the answer: The same price it will open plus or minus 300 pips with 95% probability.
Why? because the average daily return is zero. The most probable outcome for a bar is a doji and the daily ATR is 150 pips. 300 pips = ~2ATR (2sigma).

Now you can tell me I'm a fucking bastard of scammer; What can you do out of that?!
But hey look! E/U is 1.3140, 300 pips are only 2.3%. I give you a forecast with less than 5% error margin and 95% confidence to be correct and you're not happy. Come on! Weather forecasters do much worse!


Quoting Jack168
Do they use any unknown algorithm or hi-tech methods to project a trend or a range of market for next session?
There is no unknown algorithms that I know of!
There are mathematic models of the financial time series. They range from a simple MA to complex autoregressive models like ARIMA+GARCH. Or things based on Hidden Markov Models. You can find them on the web.


Machine learning (pattern matching) is used: Neural networks, fuzzy logic, gaussian processes... Also on the web. Some algo are even posted here on FF.
For example http://www.forexfactory.com/showthread.php?t=390910 tries to find similar past PA to give possible futures.
No greed. No fear. Just maths.
  • #58
  • Jul 20, 2013 5:24am Jul 20, 2013 5:24am
@infinitus. it never stops to amaze how nobody in the academic world didn't do a backtest with value investing based on the books that cover the warren buffet method. could you please explain to me why people don't look at successful methods and explain/onfirm them and would rather use something else that doesn't work?
  • #59
  • Jul 20, 2013 6:17am Jul 20, 2013 6:17am
Hi. I like your thread. Though I do not use the same terminology, I believe I practice some of the same traits. Rather than calling it a mean, I call it equilibrium.

I have not read the boundless articles and references you posted, so if this point is covered somewhere, then I apologize in advance. In my practical application of equilibrium trading, I split deviation 1 and call it the apex. While it is rarely the actual top or bottom of a move, it is the right entry/exit for consistent profit. I further use control limits of the daily pivot points and track deviation 1 apexes as they near the pivot points. In this fashion I can predict reversals to equilibrium or continuations to the next pivot level.

Give me some time to digest all your data and perhaps I can further explain practical uses of what you are doing in relation to my real world results. I have no interest in sharing my actual strategy because it is very precise and I use it on 1M charts. I cannot allow this type of information to be released for trading on such a low time frame with such a small range between high and low. Try to imagine 10,000 people trading the same 1M chart signal. That wouldn't work. However, I am not opposed to evaluating your idea against mine and possibly providing valid improvements. In addition I have considered applications to higher time frames with a modified strategy that would allow for more hands free trading, or even an EA if possible. I am in continuous search of a strategy that can employ an effective EA utilizing my experience and tools. Maybe I can find something in your thread?

Thank you for the post. I look forward to learning and sharing.
You cannot be extraordinary by being normal
  • #60
  • Jul 20, 2013 8:33am Jul 20, 2013 8:33am
Buy and Hold is the most used method in the world. Almost every grand-child has a saving account. This is investment not trading.

There need no academic research of any kind to buy and hold. Just a quite big starting capital to have early in your life(1); which not everybody has. Also very few people talk about Daren Bufflet --a totally unknown guy who perhaps never existed-- and who purchased the wrong asset 30 years ago, lost all of his money and the money he borrowed(2) and never met a journalist in his life to tell his non-success story.

There are 7 billions people in the world and a single one W. Buffet. Is it really a working method or just a statistical outlayer?
Scratching a ticket is also a method needing not academic researches, very few investment, and there are much more lotery winners then 1 out of 7000000000...

--------
(1) Buffet started at 15 year old, in 1945, with $1200. But $1200 of 1945 are equivalent to $15500 as of today. How many teenagers have this amount to invest?
(2) Buffett Associates, Ltd was funded with money he got from friends and his familly.
No greed. No fear. Just maths.
0 traders viewing now