Zigzag Bollinger Band

[snailbeard]

continuing on the theme of slippage:

To a significant extent large slippage can be avoided by:
a) having a good broker & Internet connection
b) monitoring spreads
c) avoiding weekends
d) antisipating news events
e) careful selection of times allowed for trading per pair

[nonlinear]

continuing on the theme of slippage:

To a significant extent large slippage can be avoided by:
a) having a good broker & Internet connection
b) monitoring spreads
c) avoiding weekends
d) antisipating news events
e) careful selection of times allowed for trading per pair

As I understand it, the slippage issues arise when the brokers' servers cannot handle a large number of simultaneous orders (to some extent this cannot be avoided, either the spread goes up or you get slipped). So the larger a broker gets, the worse the slippage until they get more servers to handle the bandwidth (usually after customers complain). I think the LP network also affects this somewhat. Ask Jem - he will give you the real deal. For example, I was trading a M15 strategy on a few crims and one had severe slippage (5 second and up, whereas I was getting less than 500 ms on other crims). I complained, and they ended up changing the infrastructure such that now I rarely see large slippage. But who knows how long this will last. You want a crim that stays on top of this, and don't believe what crims report on their websites. One has to constantly monitor the criminal's performance and be prepared to walk away to another. You can try to avoid trading during times of high volume, but that might affect your system's expectation. You get a signal, you want to trade it.

[franknf1]

Hi Snailbead,

Thanks for the response.
I understand that your EA may be much more touchy/hands on for people not versed deeply in its behaviour

Thanks
Nick

[snailbeard]

continuing on the theme of slippage:

To a significant extent large slippage can be avoided by:
a) having a good broker & Internet connection
b) monitoring spreads
c) avoiding weekends
d) antisipating news events
e) careful selection of times allowed for trading per pair

As I understand it, the slippage issues arise when the brokers' servers cannot handle a large number of simultaneous orders (to some extent this cannot be avoided, either the spread goes up or you get slipped). So the larger a broker gets, the worse the slippage until they get more servers to handle the bandwidth (usually after customers complain). I think the LP network also affects this somewhat. Ask Jem - he will give you the real deal. For example, I was trading a M15 strategy on a few crims and one had severe slippage (5 second and up, whereas I was getting less than 500 ms on other crims). I complained, and they ended up changing the infrastructure such that now I rarely see large slippage. But who knows how long this will last. You want a crim that stays on top of this, and don't believe what crims report on their websites. One has to constantly monitor the criminal's performance and be prepared to walk away to another. You can try to avoid trading during times of high volume, but that might affect your system's expectation. You get a signal, you want to trade it.

Yes, it seems that the EA needs to track the time/price/spread at request and the time/price at completion and write this to a CSV file for analysis.

[snailbeard]

Myths and legends

I have read much about the benefit of trading certain currencies during the appropriate session/timezone. However, does anyone have any real data or is this apocryphal?

On the one hand there are definitely quiet times and active times but how does that translate into anything useful?

Yesterday, I added some post simulation analysis - however, since there were only 91 trades in total over twelve months the result of any analysis is not very reliable. Only if there is a very strong correlation between times and outcomes can there be anything useful in it.

AUDUSD: Profits and Losses by day:
Sunday Monday Tuesday Wednesday Thursday Friday Totals
W: 0.00, 1201.82, 1340.70, 785.88, 301.48, 313.62, 3943.50
L: -189.90, -569.36, -284.92, -514.98, -99.00, -486.42, -2144.58
Net profit/loss: 1798.92

Here it seems that Monday and Tuesday were the best days, no reason to avoid Mondays at all.
Sunday didn't ever produce a winner so perhaps we can block Sunday evenings.
Friday was the worst day but not in any overwelming way.


AUDUSD: Profits and Losses by H4:
Sunday Monday Tuesday Wednesday Thursday Friday Totals

W: 0.00, 303.64, 0.00, 0.00, 0.00, 0.00, 303.64
L: 0.00, -183.00, 0.00, 0.00, 0.00, -106.38, -289.38

W: 0.00, 154.48, 190.76, 481.92, 150.00, 30.00, 1007.16
L: 0.00, -284.52, -93.00, -115.14, -99.00, -197.52, -789.18

W: 0.00, 0.00, 150.00, 0.00, 0.00, 149.00, 299.00
L: 0.00, -96.00, 0.00, -102.00, 0.00, -96.00, -294.00

W: 0.00, 465.72, 334.86, 303.96, 151.48, 134.62, 1390.64
L: 0.00, 0.00, -93.00, -79.56, 0.00, 0.00, -172.56

W: 0.00, 66.48, 358.54, 0.00, 0.00, 0.00, 425.02
L: 0.00, -2.92, -98.92, -103.14, 0.00, -86.52, -291.50

W: 0.00, 211.50, 306.54, 0.00, 0.00, 0.00, 518.04
L: -189.90, -2.92, 0.00, -115.14, 0.00, 0.00, -307.96
Net profit/loss: 1798.92

This is a little more interesting we have some weak sessions and one particulary good session.
If you were only able trade for a few hours per day then 12pm to 4pm GMT looks most interesting.

[snailbeard]

After a deeper analysis of the data day by day and hour by hour I did find some interesting observations and I setup some session filters according to the best days and hours to trade.

The result were impressively better, but unfortuately since I do not have enough data this could just be curve fitting to results retrospectively.

The twelve month profit factor went up from 1.8 to just under 3,
maximum drawdown was reduced to 3%,
net profit went up from £1798.92 to £2356.80

Overall there might well be some merit in using time based filtering, but results would also depend on method of entry, distance to target, breakeven method and the size of stop losses.

If I could repeat these improvements on other pairs without further curve fitting then I would accept the filters as valid, but until then I'm leaving them off.

[snailbeard]

Further to the above I did some number crunching of H1 candles and variations of range.

The results give a bit more information about variations in volatility for each hour of the day.

I leave it to the reader to draw their own conclusions:

AUDUSD H1
Variations in price volatility

Code: Select all

Hour:               0      1      2      3      4      5      6      7      8      9     10     11
Avg Open-Close:    7.00   9.30  11.10   7.50   7.40   6.60   6.60   7.30   9.30   9.60   8.00   8.00
Max Open-Close:   41.00  75.00  72.00  60.00  52.00  70.00  43.00  29.00  44.00  37.00  40.00  69.00
Avg High-Low:      13.70  18.50  21.00  15.30  14.50  12.90  13.40  15.60  19.30  18.70  16.60  16.30

Hour:              12     13     14     15     16     17     18     19     20     21     22     23
Avg Open-Close:    7.40   8.90  10.00  11.20   9.40   7.30   7.20   7.80   7.40   4.90   3.70   4.60
Max Open-Close:   27.00  50.00  61.00  86.00  51.00  42.00  64.00 107.00  45.00  32.00  29.00  45.00
Avg High-Low:      15.70  20.00  20.60  22.70  20.30  16.20  14.80  15.70  14.80  10.00   8.80   9.60

Biggest Average: 11:20 at 15:00 CET
Smallest Average: 3.70 at 22:00 CET

Single Biggest H1 Move: 107.00 at 19:00 CET

Smallest maximum H1 moves: 7:00, 12.00, 21:00, 22:00, 23:00

Number of samples: 6626

===============================================
Standard ATR
AUDUSD : Average True Range for various parameters:
AVG      01      05      20     100
------------------------------------
M15        9      13      10      11
H1        30      23      26      24
H4        39      53      56      47
D1       115      95     102      74
W1       250     239     174     245
MN1      647     358     478       0
===============================================

[nonlinear]

Overall there might well be some merit in using time based filtering, but results would also depend on method of entry, distance to target, breakeven method and the size of stop losses.

The above is right on in my view. I have one system that typically opens and closes trades in EURUSD one to five times a day. I know with that system, if I get a signal before the London open or after the NYC close then it is not worth trading because the pair has little gas and likely is not going to hit my target. Outside of those hours, I am not typically trading with an edge; it is just a coin flip. So with a system that typically or ideally opens and closes trades intraday, time based filtering may improve results.

[snailbeard]

Bugs, bugs, bugs and market manipulation

Last weeks basket trading on most pairs ended with more losses than gains.

Quite a few of the failed trades were good entries which were caught out by stop hunting.
I observed that a better entry point would give an additional 10 pips and enough room to ride over most of the market manipulation.

These alternative entries are essentially Bollinger band breakouts.

Other losses are occuring due to trades being entered when the price is ranging.

I had to restart the EA this morning and now it is throwing a wobbler with the partial-trade-closure!

[snailbeard]

A look at price ranging and filtering:

In this M15 view we only see that price is zigzaging between upper and lower Bollinger bands. Also the centre of price action is along the daily pivot. The H1 Bollinger band is almost horizontal which is one indication of price ranging.
Or we can say that the mid price of the M15 swings remains close to a major support/resitance level.

So we can say price is ranging if the slope of H1 is almost flat (less than x.yy pips per hour - we can also cross reference with the ATR table)
and the zigzag nodes are 'contained' rather than forming a staircase.

We don't need to restrict observations to this one time frame in the next post I'll discuss what the a longer H1 view tells us.