Instead of detecting trend, I focus on detecting turning points. Statistical approaches work best for me.

Thanks, Jian. Do you use a technical indicator to detect trend reversals?

I use point and figure charting, that takes the "noise" out of charting, and shows key support and resistance lines, pivot points and thusly, turning points. I use no technical indicators as I believe they are "lagging", especially in today's high speed trading market.

By using Point and Figure, and the Wyckoff Method, I watch volume, supply and demand and support and resistance lines only, and it works.

I use statistical approaches with time deformation. It works fine, but I do not have a continuous detection of trends (that is I am in cash 70% of the time. this discontinuity is the best protection I know of against black swan like the CHF thing of last week).

Thank you all for your inputs. CR.

to us, TREND = All price action - consolidation

So, we defined consolidation, filtered that out of the price action and are left with trend

kind regards

Thanks, Marc. What technique do you use to define consolidation, can you tell?

Help me follow a bit more, Herve, on time deformation?

Time deformation is a simple concept to understand but quite difficult to implement.

Time is relative, that is 2 hours in the arms of your loving wife and 2 hours in the trained hands of a CIA interrogator will not be perceived the same, event though, it still is 2 hours. Likewise, financial markets are sometimes slow, and sometimes fast. You need to take that into account.

I will use a simple example to make myself clear (event though this is not what I do, but it will push you in the right direction). Let's say you use moving average. How many data do you take ? Last 20 ticks ? Last 19 ? Why not 18 or 25 ? If you are using 10 different moving averages, which one will be more applicable to the market as they behave now ?

So, let's say we did a numerical analysis on the parameter called "moving average" and found out that the most significant figure is "20". So, from now on, I will compute a moving average with the 20 last ticks, except that at each new tick, I use volatility to deform these 20, so it becomes something else. Basically, we use Einstein relativity : when volatility goes up, these 20 will very likely become 15 or less, as what happens now is better explained by what happened very recently, and if volatility goes down, it is the other way around, so we do capture what Mandelbrot called the "long memory effect". In fact, if a mass deforms time as Einstein stated it, similarly, volatility does deform market time (and how we perceive it)

Basically, if you do not deform time, you end up negating the dynamics of markets, by remaining too static (what you called "lagging" in your post above and you are absolutely right : no time deformation = risk of lag). For fun, we took most Technical Indicators (Bollinger, MACD, Moving Average, etc.) and added our time deformation algo on them, and we improved their success rate very significantly.

Famous and secretive hedge funds like Citadel and Rentech Medallion do have some processes where they deform time, but from what I know, James Simmons, at Rentech, is not using Einstein relativity, but topology (but this rocket science at this stage, and we are not even trying to figure things out).

I hope this helps and answers your question. BTW, it took us close to 8 years of research to code a robust time deformation algo !

@Cesar Reis

"Thanks, Jian. Do you use a technical indicator to detect trend reversals?"

Yes, MA filters in some of my strategies.

for those focused on support resistance only you might want to take a look at the zip zag indicator

I use the jurik moving average. It does a good job of time deformation. So do tick intervals. As well as range bars.

Which technical approach works best for me in detecting market trends?

My mathematical knowledge! This gave me the tools to check up/backwards/front and leftwards all the TA analysis out there. And once more I realized, no point in following any of this :D

if you have worked for institutions which are big enough to MOVE the market, you realize this technical baloney is a bit wasting money. Per definition you probably will hit 6/7 out of 10 correctly, but I hate losing even 3/4 trades out of 10.

The best approach in order to detect a market trend is your own brain. Common sense ftw.

Thanks for the help on time deformation!

On S/R I agree on the zig zag indicator, I use it along with Dow projections over a two week period.

Tell that to James Simon of Renaissance or David Shaw of D.E.Shaw, both among the top 40 hedge fund quantitative managers in 2012, Harm. Besides, I would take any trading system that consistently has 60-70% winning trades as long as the average gain/trade is higher than the average loss/trade.

A trend clearly has 2 sides - knowing when to get in and when it get out.

I use 4 non correlated pairs on daily charts - along with slow smas (not crosses god forbid) and adx to define when to get in a faster sma to define when to get out. I set very large stops since it's the fast sma that defines when to exit.

Using this method my trades can last for days as eg the eurusd just tumbles down and down and down.

Not rocket science if done this way.

Ichimoku works best for me.

@Hervé,

Many thanks for the illuminating explanation on time deformation.

We have been working on our algo for the past 6 years for short term cyclical movements and long term trend movements in the market price of the security. After having several iterations of several concepts of social and physical sciences, we are concentrating now on the area of high and low activity and price action within those areas.

There are institutional buyers/sellers (technical) and retail buyers/sellers (sentimental) actions so you have to combine two aspects to work out cyclical movements, which then push the price and follow up action to trend movement, in a wave pattern (sine wave).

The fun begins where you analyze the buying and selling to institutions or retails, and how accurately you can assign the right buyer/seller to the buying or selling action at a price range.

The cycle is based on either strong support or resistance or both (pivot) with phases buying exhaustion or selling exhaustion.

We are testing in real time simulation, with experimental design, and test design, in a random pattern to ascertain the efficacy of the variables. Moving average, in any shape or form, is more useful to gauge market sentiments, greed or fear, as we understand.

Our model is more based on (buying and selling state) for pre-action, action, follow up, in a repeatable pattern, for high frequency trading.

Hope that this helps to add another perspective.

Thanks

.

August are you using high time frames? Just Ichi or with another inducator?