Zigzag Bollinger Band

Post Reply
User avatar
snailbeard
Trader
Posts: 615
Joined: Mon Dec 24, 2012 10:54 am
Location: Just above water somewhere between Oxford & Cambridge

Zigzag Bollinger Band

Post by snailbeard »

This autotrader has grown out of a manual trading system that I found to be more consistent than other methods that I have tried in the past. There are lots of examples in the manual trading topic of a similar name
http://www.stevehopwoodforex.com/phpBB3 ... =16&t=2792.

One of the aspects that I like about it, is the consistency of results across different brokers and the robustness over 12 months of data. Although the results are of course not exactly the same, they are at least very similar. This is huge help in determining if changes are beneficial or detrimental.

The current logic does not use pivot levels at all, which can be a source of differences due to the timezone issues when moving between brokers.

However, pivots can sometimes be really useful and some changes I am working on require the daily pivot to help determine the daily bias so I might once again be receiving contradictory results. The solution will be to determine how to get exactly the same pivots regardless of timezone or to use a support/resistance level finder which is independent of timezone candles. One method is to derive levels from the higher time frame Fibonacci levels.

Update: So much has changed and it now uses both weekly and daily pivots as well as Fibonacci levels. However, the user interface still leaves a lot of scope for improvement!

It has reached a long-term performance level which I consider to be adequate but not exciting compared to some of the high risk robots.

I have spent the last few weeks running the robot on live data demo accounts
Global Prime
Oanda
IG - good for testing against high spreads.

Mrs Snailbeard complained that she had a lot of alerts but no trades on IG
- well a trade entry spread limit of 2 pips seems to be too much for some brokers
so I had to increase it to 3++ and added some code to track: high, low and average over 60 minutes.

The results of a few trades are statistically meaningless but a lot of minor issues needed to be fixed which don't show up in simulation. The most important ones were spread issues and the lot size and risk calculations.

Mrs Snailbeard was not impressed with the user interface and I didn't get very far with the argument that a blank screen is very light on CPU usage. So I have created a light weight updater which economically puts out information once per minute.
You do not have the required permissions to view the files attached to this post.
Last edited by snailbeard on Tue Sep 09, 2014 7:48 pm, edited 2 times in total.
User avatar
snailbeard
Trader
Posts: 615
Joined: Mon Dec 24, 2012 10:54 am
Location: Just above water somewhere between Oxford & Cambridge

Re: Zigzag Bollinger Band

Post by snailbeard »

Mrs Snailbeard is not convinced that the status update is much of an improvement over a blank screen.

However, it has the advantage of:

a) not killing the back testing performance
b) working in visual back test mode - I have recently found this to be very useful for checking new functions prior to running a full back test.

The rock bottom performance (12 months) or baseline for improvements is
(AUDUSD)
A profit factor of 1.8
risk per trade 1%
drawdown of 3.76 % per pair
growth (91 trades per annum) : 18% per annum per currency

For the near future there are several ways to improve results
1. Add more entry methods
2. Allow (rentry) overlapping trades on winning trades
3. Improve the near future prediction of price movement
....(this overlaps with additional entry methods which might indicate a reverse bias)
4. optimise stop losses to reduce 'false' losing trades
5. optimise partial take profit for best profit
6. focus on fewer currencies and increase risk per currency to 2%

I am currently mulling over the idea that some wrong trades can be prevented by detecting an entry by other means in the appropriate direction. If this is successful it has a double benefit of gaining a profit and blocking a loss.
User avatar
snailbeard
Trader
Posts: 615
Joined: Mon Dec 24, 2012 10:54 am
Location: Just above water somewhere between Oxford & Cambridge

Re: Zigzag Bollinger Band

Post by snailbeard »

The real cost of multi-pair trading

Although 12 months of back testing and bug fixes have eliminated a lot of issues, it does not prepare us for the real cost of basket trading. How do we know what the long term impact of spreads, commissions and swaps will have on overall performance?

After one week of live demo testing I can see that partial exits are treated as new trades and receive repeat commissons, therefore I am paying several times over for what is meant to be a single trade.
However, this could be a misunderstanding on my part, perhaps half the commission is charged on entry and the other half on exit.

What is the equivalent spread that would have a similar cost to the commissions?

It is also clear that longer trades with positive swap will be the best performers.

Therefore should I add a module to the EA to generate a custom report which includes my adjustments for spreads, commissions and swaps?

Do the swaps change significantly over the period of the back test?
Can I use an average value or do I need to load value from file into a table?

For simplicity I could limit trades to pairs with smaller swaps and I could limit usage to brokers with wider spreads but no commissions.

This attachment gives a true picture of the cost of multi-pair trading:
You do not have the required permissions to view the files attached to this post.
User avatar
snailbeard
Trader
Posts: 615
Joined: Mon Dec 24, 2012 10:54 am
Location: Just above water somewhere between Oxford & Cambridge

Re: Zigzag Bollinger Band

Post by snailbeard »

Comparing commissions with spread charges

The following attachment is a rough comparison between the cost of commissions in comparison to the cost of higher spreads. It seems that Global Prime's commissions work out much cheaper per trade than paying a higher spread from an alternative broker.

By back testing with a spread equivalent to two pips we more than cover the equivalent cost of commissions.

I'll look at the swap costs another time.
You do not have the required permissions to view the files attached to this post.
nonlinear
Trader
Posts: 293
Joined: Fri Jun 01, 2012 2:56 pm
Location: Washington DC

Re: Zigzag Bollinger Band

Post by nonlinear »

I knew you would pop up here eventually, lol. I like your system and will check it out. You are analyzing numbers that in my view need a real account for testing purposes. Commissions, swap, spreads, etc on demo just do not match real feeds, at least on my crims. Thanks!
User avatar
snailbeard
Trader
Posts: 615
Joined: Mon Dec 24, 2012 10:54 am
Location: Just above water somewhere between Oxford & Cambridge

Re: Zigzag Bollinger Band

Post by snailbeard »

Nonlinear,

Thanks for bringing this point up.

It would be useful to know the real money cost when there are requotes, slippage & spread widening.

I came across a couple of interesting sites that compare broker spreads and it is surprising how many brokers still have spreads over 3 pips on major pairs.

http://www.fxintel.com/
http://www.mt4spreads.com/

It is useful to see which ones are real brokers (ECN) and which ones are criminals.

I implemented the Virtual Pending Order for the purpose of not triggering trades when the spread increases or the price spikes briefly. It will be interesting to see how this performs but I don't currently have any analysis of whether or not it makes much difference.

The configurable partial exits took an effort to implement correctly and reliably and I still do not have the ability to change the amounts by pair or by entry method, although I'm thinking about a reliable way to do it.

I have less control over exits, the partial exit logic has a limit on slippage but thats all.

In the end if we have a broker who is a criminal then the spread is not the real cost...
franknf1
Trader
Posts: 12
Joined: Mon Jan 23, 2012 7:10 am

Re: Zigzag Bollinger Band

Post by franknf1 »

Hi,
This sounds interesting.
How do I geta copy of the autotrader?

Thanks

nick
nonlinear
Trader
Posts: 293
Joined: Fri Jun 01, 2012 2:56 pm
Location: Washington DC

Re: Zigzag Bollinger Band

Post by nonlinear »

You are lucky that you have more options available to you than us US sorry folk. I would look at LMAX, Axi, and GP if they would take US clients. Strong executions, no funny business, and decent commission structure. Jem looks to have set up a class act in GP. In the states, I trade with MB, ATC, and Citi and have been pleased so far, but it really depends on your trading strategy.

I should note that those reported spreads you found look all nice and shiny, but slippage is the real silent killer that nobody reports. I mean, who care is you get 1.0 on the euro if one of every five trades slip 10 pips or more? Until you trade the EA live on different brokers on different time zones, you don't really know how your strategy will perform or be affected by delays, bar changes, and such. If you have proven your strategy profitable on back tests and on forward demo, I would try it on a live micro on a few differently timed brokers to get a real feel. That said, I am no pro, so what do I know. ;)
User avatar
snailbeard
Trader
Posts: 615
Joined: Mon Dec 24, 2012 10:54 am
Location: Just above water somewhere between Oxford & Cambridge

Re: Zigzag Bollinger Band

Post by snailbeard »

I agree completely, it'll be the silent slippage which will be the real killer, and no way to find out but to run on live accounts, unless someone out there has reliable statistics.
Can the results from a micro account be applied to a normal account?

The following information in this respect is academic but I'll put here for reference:
(sorry about the small font - I'll have to use the text functions to impove the display)

Oanda(demo)
demo-Oanda-spreads-swaps-0750-07Nov2013.png
Alpari UK(real)
real-Alapri-UK-spreads-swaps-0745-07Nov2013.png
GlobalPrime(demo)
demo-GlobalPrime-spreads-swaps-2013-11-07 0747.png
You do not have the required permissions to view the files attached to this post.
User avatar
snailbeard
Trader
Posts: 615
Joined: Mon Dec 24, 2012 10:54 am
Location: Just above water somewhere between Oxford & Cambridge

Re: Zigzag Bollinger Band

Post by snailbeard »

franknf1 wrote:Hi,
This sounds interesting.
How do I geta copy of the autotrader?

Thanks

nick
Hi Nick,

Your Odyssey EA survived 12 months of backtesting without crashing into losses - that is a great achievement and a starting point for adding some layers to filter entries and turn it into a nice little earner!

My EA is a hideous monster and very touchy - it is better that it stays in it's cage!
Post Reply

Return to “Automated trading systems”