Zigzag Bollinger Band

[snailbeard]

One of the reasons that there is no buy trade on the 7 Feb 2014 is that the low-volatility-flag remains true and lags behind actual price movements. Adding more entry patterns won't overcome this restriction. It will be necessary to modify the assessment of volatility. Many traders use ATR, but I don't know how consistent this value will be over a wide range of data.

Then there is the question of whether or not to use multiple ATR values to smooth out the erratic behaviour of a single ATR line. In the following chart using multiple ATRs produces narrow windows of valid volatility, which could turn out to be too restrictive, but in this particular case it actually coincides with the best time to enter a trade:

[snailbeard]

After adding a couple of functions to analyse ATR and rerunning the backtest I still don't get an entry by the normal methods.

Also the breadcrumbs picked up something interesting:

Code: Select all

Found 2581 matches for "st_MultiM15AtrIncreasing: 0".
Found 540 matches for "st_MultiM15AtrIncreasing: 1".

Found 3114 matches for "st_bGoodM15PipsPerHour: 0".
Found 7 matches for "st_bGoodM15PipsPerHour: 1".

Found 2317 matches for "st_bH1BbTooNarrow: 0".
Found 804 matches for "st_bH1BbTooNarrow: 1".

Only 7 matches for st_bGoodM15PipsPerHour?
Thast's one thing to look into...

Out of 10 trades 2 failed:
Both failing trades are due to obvious reversal detection issues which can be fixed.

[snailbeard]

This is the second bad trade due to a reversal at the end of January 2014:
This time we have to work out whether or not it is a pullback before returning to the weekly momentum. This is one case that used to cause losses in NB 10.4
a) The H4 bars are falling not rising
b) The H4 bars close inside and close to the D1 (purple) upper Bollinger band
c) This is a H4 break-in of the D1 Bollinger band
So we should consider price to be in noman's land.
We need to wait for confirmed move out of noman's land.

[snailbeard]

Following on from the previous chart, there is a similar message from the D1 chart which has a supernode D1 zigzag in combination with a bearish engulfing pattern. We know its going down but we have to wait a couple of days for the bulls to give up:

[snailbeard]

On the W1 chart of GBPUSD we can make use of nested Fibonacci levels to find important clues to price barriers. We only need the orange values but they should fit with the longer term levels.

[snailbeard]

The recent changes and bug fixes (I have not added OPRB yet) have improved the results for GPBUSD on the most recent 5 months, initially it was making a loss. I need to investigate where there are two or more sequenctial losing trades but this is a step in the right direction:

[snailbeard]

I'm a bit wary about adding another layer of filters based on a lagging indicator. Experienced manual traders can see falling and rising volatility from the candles without needing the ATR indicator. However, ZzBB needs to derive volatility, which it does from Bollinger band width. Generally a shrinking width means falling volatility and expanding width means rising volatility. So why does ZzBb need more versions of volatility?
Perhaps it doesn't, maybe a better use of BBs would yield the same information.

However, looking at the sequence of 4 losing trades, 3 of them had very low ATR values and two of them should have been picked up by the H4 reversal detection logic. Something else I noticed was that most (but not all) of the good trades were consistent with rising ATR values. However, I noticed one of the good trades that occurred during falling ATR could have been entered later (4 hours later for the same result). Is that signifcant? Open trades are vulnerable to unexpected events therefore keeping trades open longer than necessary adds to the risk. Secondly, open trades reduce the available margin and I have had brokers block entries of other pairs because of this. So if this is a restriction then better to let the most promising trades open first.

The attached chart shows that ATR gives a completely different view of volatility compared to the Bollinger Bands. ATR responds faster to changes in volatility. This is not surprising since ATR uses the faster exponential moving average (EMA), whereas the Bollinger band calculation uses simple moving average. I even tried an EMA version of the BBs and the band contraction was faster, but the expansion was different, slower perhaps but when overlaid on top of the normal BB it might be used to confirm an entry.

[snailbeard]

No shortage of acronyms here but I think I should try to be consistent, therefore I'll try and stick to one abbreviation which will be ORFB :- short for Open-Reverse-Forward-Break.

What about a combination of a Fast-Bollinger-Band crossover combined with ORFB?

In the following chart (at 03:30):
a) Both the black and yellow Bolinger bands are sloping up.
b) The faster Yellow BB has crossed over the Black BB.
c) There is an ORFB between 3:30 and 3:45 on the 29 November.
d) ATR is generally going up but the ATR(50 & 60) are not so what does that mean?

[snailbeard]

To properly use an indicator like ATR it is necessary to find out when it will bite you. It is too simple to say:

if it is going up it is safe to trade and if it is going down it is not safe to trade.

However, it appears more consistent in respect of not entering trades when ATR is falling.
Should we exit if all 5 ATR indicators are falling?
Should we exit if the first three ATRs are falling?
Should we exit if ATR falls below a fixed threshold?

Some traders use a fixed threshold of about 12 pips for entering trades.

In the attached chart we can see how ATR varies over a single trading day.

We can see how it can be safe to stack entries at the start of a swing and really squeeze the pips out of the days move.
The tweezer top candles make excellent ORFB entries.
However, without constraints there can be late entries at the end of the swing.

[snailbeard]

Last year I implemented an ATR module to produce an ATR summary for a range of timeframes but didn't spend any time working out how to make use of it. So now that I have some ideas about M15 I have updated it a bit to produce more info about M15:

Code: Select all

==================================================
GBPUSD : Average True Range for various parameters:
AVG        10        14        20       50       100
M15        4.9       5.6       6.2       8.6      10.1
H1        19.7      17.9      21.0      19.2      18.8
H4        37.7      38.4      40.9      37.3      39.0
D1        88.0     100.5      99.1     103.0      97.0
W1       234.0     234.0     240.2     260.7     238.6
MN1      580.1     533.2     532.0     609.2     713.8
==================================================

==================================================
GBPUSD : Average True Range for various parameters:
AVG        10        14        20       50       100
M15 B      4.9      5.6       6.2       8.6      10.1
M15 A      4.9      5.6       7.4       8.6      10.2
Dir:      0.00      0.00     -1.20      0.00     -0.10
==================================================