How to determine the strength of a trend

Originally posted by ADX_trader


Thank for your help.

I also use ADX and MACD. But waiting the ADX to rise above 30 may make me miss quite a lot of the trend. If the trend is long enough, this is fine. But for a short trend you may enter at or near the peak. A strong trend can also finish quickly.

May be the question is: How to determine the strength and the LENGTH of the trend?
More...

We tested many strategies and indicators (including the ones developed by us) in a sense of trend strength recognition. By that experience we found no very reliable patterns or indicators doing this. You deal with statistics like with any other aspect of trading. There is some percentage of trends continuing and some percentage of trends failing to do so.
Also, through the time, our studies show that such indicators become less reliable, which means it`s harder to make money then many years ago (but I believe that does not need studies to prove - we all feel it :D )
 
A common study is to run adx along with dmi. You run them both in the same frame.

Every trend starts with a dmi crossover, but not every dmi crossover starts a trend.

A crossover accompianied by a rising adx is an indication that the trend may have legs. A falling adx means the current trend may be weakening.

people complain the adx is too slow, that's because they don't understand that it is really too fast. By the time it hits 30, the trend is already well underway, but the adx told you the trend might have legs when it rose from 8 to 12 after the dmi crossover.
 
ProfitSeer hit the nail right on the head when he spoke of using DMI and ADX together.

I've taken it a bit further than that...

I used some of John Ehlers' code to create a modified DMI+, DMI-, and ADX. These automatically adjust to market parameters (dominant market cycle length). I chart them all in the same subchart.

To keep out of the whipsaw brier patch, I created a Signal-To-Noise Ratio (SNR) function. I chart SNR in a separate subgraph.

When the SNR crosses over its threshold, and DMI+ crosses over DMI-, and ADX rises above DMI-, you know a trend is in place on the long side.

When DMI+ crosses below DMI-, or the ADX crosses below both DMI+ and DMI-, or the SNR crosses below its threshold, you know to get out.

Conversely, when the SNR crosses over its threshold, and DMI- crosses over DMI+, and ADX rises above DMI+, you know a trend is in place on the short side.

When DMI- crosses below DMI+, or the ADX crosses below both DMI+ and DMI-, or the SNR crosses below its threshold, you know to get out.

Works great. Gets you in right at the start of the trend, and out right before it ends. Sometimes, the indicators actually get you in before the trend starts.

And the great thing about it is, you have no inputs, so you have no optimization to go through (which can take a long time). The indicators optimize on-the-fly.

DGBrothers
 
Hi,

Here's a screen shot of a recent day. I'm still playing with the time frames, so this one is on a 10 minute chart.

As you can see, on 09/09/02, at around 9:50am, you would have a great setup to enter at around 890.

The SNR almost didn't confirm the trade, but it finally took off.

You would have held the trade until 09/10/02, at around 7:10am, when the SNR dropped below its threshold. You would have gotten out at around 902, for a 12 point runup, a pretty small trade for this system.

From 7:50am to 8:20am and 8:30am to 9:00am on 09/10/02, the SNR is below its threshold, (except for a small pop above it from 8:20 to 8:30, which would have resulted in a scratch trade), keeping you out of any trades. From 9:00am to 9:50am, a weak short signal develops, but you'd be stopped out at 9:30 for a scratch trade due to the stale trade exits (exit strategy that gets you out after a short time if the trade isn't producing enough profit).

From 9:50am to 11:00am, the ADX is below both DMI+ and DMI-, keeping you out of any trades. Just as the ADX begins to rise, the SNR jumps up at 11:10am, giving a weak short trade at 903.

The ADX drops below both DMI+ and DMI- again in short order, getting you out of the trade at 901, basically break-even after slippage. So, on the day, you'd be up around 7 points or so net, not including the long trade you would have gotten in at 12:10pm at around 907 and out the next day at 912 at 11:10am. It was a pretty crappy day.

To the left of the graphic, you can see a small orange dot, that's the ShowMe indicator for a date gap. Although the indicators showed a trade developing right at the open, because of the date gap rules, you would have had to wait 1/2 hour after the open to enter any trades (this gives the adaptive indicators time to adjust to market conditions after a date gap). In this case, this would have resulted in a scratch trade, due to the stale trade exits.

Also keep in mind that I work at a brokerage firm, so commission costs are usually zero for unprofitable trades, we don't charge commissions generally if our trades (that's our trade recommendations, if a client calls an unprofitable trade, we still charge commissions) don't make money. That seriously complicates backtesting, but clients seem to like it.

Keep in mind this system is designed to catch the large moves, greater than 10-15 points, so it falters a bit with the smaller moves. I'm still looking for a way to effectively capture the smaller moves that fly under this system's radar. This system is extremely good for keeping you in the longer trends that last for three or four days, and drop 70-80 points in your lap, without giving you false exit indicators.

I was thinking of doing this to capture the smaller moves:
Calculate the Chaiken oscillator, then plot the CoEffR Pearson Product Coefficient between the Chaiken oscillator and price. This would give the correlation between where price is, and where the Chaiken oscillator says price should be. Basically trade any divergences between the Chaiken oscillator and price. If anybody has tried this, please let me know if it's effective.

DGBrothers
 

Attachments

I'd really appreciate if anybody could tell me if there's a more efficient way of finding the entry and exit points using these indicators. Sometimes I get so wrapped up in one course of action that it's hard to see other ones.

Thanks again,
DGBrothers
 
DGBrothers:

You reminded me an important concept. A system that captures big trends will be very different from the one that captures small trends. I think it is impossible to do both in one system. My experience is a system that captures small trends will make you exit too early in a big one.

I welcome any comment.
 
Also look at RSI and ROC.....a slowing ROC is a danger symbol....if you think you can use Elliot wave and identify the third impulse wave (or identify the impulse extension if not the 3rd - often the 5th for futures) then good for you, but I found this very difficult on a 5 min chart. Then you can follow the rules...you know them...only 1 extension, etc...
 
Actually, I put no credence whatsoever in Fibonnacci, Elliott waves, Gann lines, etc.

They seem to me to be more superstitious mumbo-jumbo than actual trading strategies/indicators...much too subjective.

I mean really, can anyone really tell me without any doubt what Elliot wave the market is in right now?

We have a trader here who uses them, in conjunction with slow stochastics, and you always hear him say, "Well, it could be the start of the third wave up...or, if you look at it on a longer time frame, this will be the fifth wave..." or something similar. It's not surprising that he gets into many bad trades...he's using mumbo jumbo to determine when to get in, and stochastics (which sticks to its extreme position on a trend, thus giving false exit signals) to get out. No wonder he has no confidence in pulling the trigger.
 
Breakout, in the screenshot you posted a while back, I notice that you're using 100-tick bars. If I understand the concept correctly, a new bar will only be drawn after 100 ticks, is that right?

Can you tell me if there is an advantage to using tick bars instead of time bars? If I'm thinking correctly, in a slow market (fewer trades, fewer ticks), there would be fewer bars drawn, and in a faster market there would be more bars drawn, right?

Does this allow one to keep up more easily with price action? How does this affect the EasyLanguage code in a non-discretionary trading system?

I'm playing around with time frames for our new trading system, and trying to determine the best one. I didn't stop to think that the best time frame might be no time frame at all, but tick bars instead.

Let me know of your experience with tick bars, ok?

Thanks,
DGBrothers
 
You must log in or register to reply here.
Back
Top