Robot Results 2012-05-21 to 2012-05-26

Another week of big moves, but the robots we've been reviewing didn't all catch them. Let's see how they did.




Inside Bar

This robot is very interesting to me, or rather the pattern of inside bars it plots is interesting. Some weeks there are plenty. Others there are none. The EURUSD had none, AUDUSD had two. This robot was able to catch one inside bar candle, and ride it for an impressive $916 profit, the most this robot has ever done. With swap taken into account, final weekly profit was $890.48. 

Note that the price initially triggered the BUY STOP order, but then fell. Luckily it moved enough to the upside to allow the trailing stop to trigger, capturing $1.83 of profit before reversing direction and triggering the SELL STOP. 

Last week this robot made only $97. Total balance now is $7008.65. The demo account expired today, so I had to reset it with a $5,000 balance. So this robot has made a 40% return in the few months it was active.  It's been mostly consistent, at least since switching to the four hour time frame. 

Due to it's seeming success, I'm currently trialing several other robots based on this design,



Pending Reversal Candle

This robot made 10% on it's demo run, and has started with a fresh account, keeping the same settings. Last week lost $100. This week had a couple of losses and three wins, enough for a profit of $181.

I hadn't expected this robot to last this long. While the profit it makes is inconsistent, it's an interesting design, and I'll keep it running a little longer.



Daily Close Break Out

Big losses last week, due to a whipsawing GBPJPY, and ended $750 down, closing balance was $3184. When the balances get this low, I generally declare the robot to be a dud and shut it down.

However, this week it made $259, current balance is $3443.

I didn't get around to reviewing the settings of this robot as I thought to do after last weeks loss. There may yet be something to this robot, but I'm not quite sure what it needs to succeed. It's affected badly by whipsaws, so I'll have to think of a way for this robot to deal with them more effectively.



Weekly Break Out

Last week made $943, mostly from a GBPJPY pair which ended it's whip-saw and broke out to the downside. This week suffers more GBPJPY whip-sawing and gives back $445.

Failed to wring much profit out of a bearish EURUSD trade, for shame. That was easy money.

Current balance is $4402, with one trade open for $177.



Weekly Close + ATR

I forgot to save the screen capture of this robot, and then the demo account expired and had to be reset, so I was unable to save the results.

Last week it made $1547, the largest return I've seen in any of my Forex robots, after capturing a GBPJPY bearish breakout trade and using a lot of leverage.

From memory it suffered the same fate as the Weekly Break Out Robot above, getting badly whipsawed by GBPJPY and failing to capture other big moves, and it gave back about $851.

The ATR function of the robot is supposed to compensate for whip-saws somewhat, but it's obviously not enough!

The closing balance was $4995 before the demo account expired and had to be reset. That much I did manage to record!


Open-Close H4 Bot

Lastly our sad old trend following robot, which thanks to it's new over brought/over sold filter, lost $14.01 last week in a strong down trend. Without the filter, this robot should have made a lot more, since it's designed to follow trends, but the filter thought the trend was too strong and wouldn't let it open sell orders.

This week it made $240, which is the best it's done for a while. Balance is $4589.

EURUSD 2012-05-28

Really not much to report since the EURUSD has been plummeting since the beginning of the month. We're all just hanging out waiting for a bounce. There's signs of a stall about now, but with the European news still not being great, no one is expecting it to go up. The general consensus is to sell into rallies, and keep an eye on other currency pairs to see what they're up to. 

(Daily EURUSD chart)

Robot Results 2012-05-14 to 2012-05-19

Another week of strong trends, which unfortunately my robots didn't capitalize on as well as they could have. Still, it's all a learning experience.


Inside Bar

Last week made $513, this week only $97. Balance is $6118. Still, a positive result, and I'm very pleased with this robot over all. However, I suspect it's doing well because the market is co-operating with it. In choppier conditions, I'm not sure it would survive.

I think the demo account may be about to expire, so it'll be starting again from $5,000, but we'll remember the ending balance.

I also started demo testing two variations of this robot (not shown). One uses a stricter definition of an inside bar, and the other uses highs and lows only for inside bar detection. We'll see how they go.



Pending Reversal Candle

Demo account expired, so it was started afresh. Balance of the previous account was $5548, so it basically made 10% in the 2 or 3 months the demo lasted (I always forget which).

The profit wasn't linear, however. This robot had it's ups and downs, ending up where it started at $5,000, before posting one good trade last week which is the 10% gain.

During these last few months, the settings were run through the back test optimizer, but otherwise the robot was mostly unchanged.

Last week made $515, this week is down $100, but one open trade is +$279. Balance on the new demo account is $4899.



Daily Close Break Out

Up $202 last week, but this week suffers a gigantic early EURUSD loss and was whipsawed by GBPJPY, ending up $750 down. Balance is $3184. I guess the settings changes aren't working for this robot. I'll review them.

The big losses came early in the week, and while it regained a little profit, it was nowhere near enough.


Weekly Break Out

Last week up $57. This week it performed well, but several trades on strongly trending pairs hit their trailing stops early and missed out on a lot more profit.

GBPJPY whipsawed back and forward before making up it's mind and plunging, but the robot was able to catch the move while not taking too many hits and so ended the week up $943, most of the profit from a single GBPJPY trade which alone made $788.10.

Balance is $4848. So close to break even.



Weekly Close + ATR

Not a new robot, but it was added to the weekly review last week so we can compare the slightly different strategy it uses to it's near relative, the Weekly Breakout Robot.

Last week up $267. This week, thanks to it's larger lot sizes and catching the same GBPJPY move that the Weekly Breakout Robot did, is up an astonishing $1749, with the GBPJPY trade accounting for $1547 of that.

That is a 42% return in a single week!

This is the single largest Forex trade I've ever experienced. Shame it was just on a demo account!

This is what these robots are capable of. Just fine tuning them to avoid whipsaws and stopping out too early is the challenge.

Without the GBPJPY trade, this robot still would have made $202, which isn't terrible, but kind of small compared to the risky amount of leverage being used.

Balance is $5893.


Open-Close H4 Bot

Our old comparison trend-catching robot made $267 last week, it's first profit in a long time, thanks to a strong down trend. I added an over brought/over sold filter - it's first code change since November - and it seems to have stopped it taking several trades which would have been profitable! Oh well.

This week it's down $14.01, a pretty miserable result considering EURUSD trended down almost the entire week.

Balance is $4349.

MACD EA

Search results for "MACD EA" and "MACD robot" have shown up in the website statistics lately. While I coded some MACD EA's in the past, none of them were very successful.

Out of interest, I attempted to code another MACD robot last night using all the coding techniques I've learned over the last 3 years, but to no avail - MACD doesn't appear to lend itself to successful automated trading.

Even using it just to determine trend on higher time frames and using other methods for trade entries were failures.

While it's always possible someone has coded a successful MACD based EA, or while it may be useful for trading manually, it's not something I've ever had any luck with.

xTDI Cross EA

This EA uses the free Traders Dynamic Index indicator which seems to be quite popular lately.

You can learn more about the Traders Dynamic Index indicator here, and the EA was downloaded from this forum post on Forex Factory.

There's several trading methods described in the forums using this indicator, and even some more EA's.

This EA was downloaded from a trading method forum that didn't look overly complicated, but I'm not sure if
the EA follows that particular method exactly or not.

After installing the indicator, the EA was run through a back test with it's default settings. After having no luck with the 60 minute EURUSD chart, I tried the 4 hour chart, with these results:

(Back test on 4 hour EURUSD chart, 1-1-2011 to 1-1-2012)

The equity curve is a little jagged, but better than a lot of my EA's. The lot size is a constant 0.10 and this EA returns almost 13% over the year. Only a 43% win rate, but the average win is $70 compared to the average loss of $45.

This EA places quite a few trades for the 4 hour time frame, and the back test reveals almost 5600 mismatched chart errors, which means the simulation wasn't quite so accurate (which must be very poor, considering accurate simulations are pretty bad to start with!)

Still, considering no setting changes or optimizations were made, these are reasonably impressive results for a free EA downloaded off a forum.

Comments in the EA say it is copyright of Xaphod, http://www.xaphod.com

danw2 - np - next stop robot

This robot I coded according to the rules from a forum post on Baby Pips, called "Next Stop".

It was tested, and the reviewer had this to say about it: 

"In fact, out of ALL the systems I have tested, Next Stop ranks the highest."

The review can be found here. 

(Back test on 60 minute EURUSD chart from 1-1-2011 to 1-1-2012)

The results are positive, but they're not great. The equity curve shows slight profit, a dip into negative territory, then regains some to close with the same profit it had earlier on. 

57% win rate, average profit on a win is $34.30 vs average loss $40.92. Averages 3 wins to 2 losses. Those aren't terrible statistics, but with the lot size set at 1% of the equity balance, it's only achieving a 1% return over the year. 

I can't quite make out the Baby Pips reviewers statistics, but at first glance they don't look all that impressive. 29% win rate compared to 57% on my test, but their average win compared to loss looks to be a lot better than my system. 

The systems are different, as mine tries to follow the original rules more closely, whereas the reviewer mentions leaving out the daily trend detection and not using RSI. However I'm not completely clear on the Ichimoku settings either, so I left them at the default in my code, except for the Kijunsen which was set at 3 bars back.

I tried running it through the back test optimizer, trying different values for Ichimoku and RSI, taking RSI out, adding an over brought/over sold filter, and several other things, but none with better results that what I got here.

It also bombed on GBPUSD and AUDUSD.

The instructions for trading it manually include two additional triggers for entering trades which I didn't code, as they tend to be unreliable in the code. 

So, doesn't look to be a very good system when automated. 

danw2 - mtf adx again EA

A while ago the web site stats for this site started showing increasing hits on my ADX based robots, and searches for "multi time frame ADX robot or EA", so I vowed to review and post more of them.

Alas, interest fell, and I discontinued, but you can still download all of the ADX robots in a zip file here, as well as search for ADX on this website at the top left to see the ADX related posts.

Interest is picking up again, so I threw together another ADX based robot using multiple time frames, ran it through the optimizer, and here are the results, using a modest lot size:

(Back test on 60 minute EURUSD chart, 1-1-2011 to 1-1-2012)

A reasonable equity curve, 46% return over the year, 90% win rate, and averages 11 consecutive wins to 1 loss.

The wins are small, however, and the losses large - about 5 times the average win. But the win rate makes this robot profitable over all.

Always be aware of robots with high accuracy rates - you may find they have small win rates also. Or perhaps this is just something that occurs with mine!

The stop loss is set at 80, which is quite large - I generally prefer my robots with smaller stop losses, although back test optimization's usually suggest higher values - I'm guessing to allow rogue trades to come back into profit.

Alas, it fails miserably on GBPUSD and AUDUSD. This indicates that it is probably not a great strategy over all, and will fail if you demo test it.

The strategy is uses is to look for a trend using ADX on 3 time frames (by default the hourly, 4 hour, and daily time frames), then open trades on pull backs to the trend defined by the stochastic indicator.

Absolutely do not run this robot (or any of my robots) on a live account!

Download it here.

EURUSD - 2012-05-12

News from Europe lit a fire under the EURUSD which had an exciting open and started trending down, before consolidating later in the week and going nowhere.

Let's look at charts!

(Hourly EURUSD chart with 100 & 200 period moving averages)

The price is well below the 100 period moving average on hourly, 4 hour, and daily time frames.

(Daily EURUSD chart with Ichimoku Kinko Hyo)

Looking good here - price has jumped out of the cloud and is below the indicator lines as well.

(Hourly EURUSD chart)

Whoa there! We see some signs of stalling about the 1.29 price level, and after the big jump Monday which finally pushed price below the 1.30 barrier which has held for so long, we have another descending triangle, similar to the one which formed on the daily chart using 1.30 as support.

While several other currency pairs (not shown) may be showing signs of a pending reversal on higher time frames, USD has been gaining strength across many of them this last week. So EURUSD should continue pushing down, but we may see a retrace upwards first.

Also, the bearish trend line on the weekly chart is still intact, which is encouraging. We'll just keep an eye on price, and in particular wait for a retrace back to the 100 period moving average on at least the hourly chart before initiating some bearish trades, keeping in mind what other pairs are doing.

Robot Results 2012-05-07 - 2012-05-12

An exciting start to the week, which then fizzled out somewhat as traders took their profits and probably went on vacation. Let's see how our regularly reviewed demo bots did!



Inside Bar

Last week made $403, this week $513, most of the profit coming from a trade left over from Friday, which benefited from the market open gap on Monday. Balance is a little over $6,000. Interestingly, no EURUSD trades this week.

This robot has made profit several weeks in a row. Will it hold? History tells us no, but we'll enjoy the spectacle while it lasts.



Pending Reversal Candle

Last week made $83, this week makes $515, benefiting from the dropping AUDUSD. Balance is $5548. This robot gives mixed results, often surprising with wins when other bots fail. I really thought this would have failed by now.



Daily Close Breakout

Late last week I switched this to H4 time frame (from my vague memory), and adjusted the stop loss settings. Lost $706 last week, this week is up $202, but has one order currently open which is down $216. Lots of small wins, but a few big losses wipe them all out. Balance is $3934.

I suspect this robot needs further settings adjustments.



Weekly Breakout 2

Up $79 last week, but some open orders carried over the weekend posted a good gain Monday morning, which it then proceeded to squander during the week. Closed up $57, one open order for +$168, balance $3905.

Still feel this one has potential, but needs adjustments somewhere. 

Open-Close Bot H4

Lastly, our oldest bot, running unmodified since November last year, posts it's first profit in a while, catching the falling EURUSD. Lost $220 last week, this week up $243, squandering many of it's early gains.

I added an over-brought/over-sold stochastic based filter to this robot, because I felt sorry for it, and wanted to see how the filter might perform, having gotten some positive results on back tests from other robots.

Balance is $4363.

Weekly Close + ATR Breakout

Wait, what's this? You said we were done! Just for fun and for comparison, I present a modified version of the Weekly Breakout 2 robot. This uses the same general principals, but instead of using the first H4 candle of the week to draw the breakout box, it uses last weeks closing price plus an ATR based box.

Current balance is $4143, so like it's parent bot, it's not done so great during it's life. I believe I adjusted stop loss last week and also changed this to H4 time frame.

This week is up $267, but one open order is currently down $113.


--

So that's it for another week. The H4 time frame is proving the most reliable, but I'm still not sure if this is simply because less trades are placed and we don't have enough information on results yet.

Some promising results from a couple of robots over the weeks, but only one has breached the $6,000 barrier, which is a 20% return from the start of the bots, some months ago, and these using unsafe levels of leverage.

I have gotten a few ideas from watching these robots perform, however, and have implemented some of them into other demo robots not shown here. These ideas are mostly based on price momentum and stop loss levels.

More time is required to test them out! So much time. And while testing continues, the market will continue to change and some of these strategies may not work anymore. Funnily, some older ones may start showing positive results again or for the first time.

Back Test Results

As if back testing wasn't hard enough, I get different results using the exact same settings on different passes.

Check out these three graphs:

Three passes of the same EA on the same time frame, same dates, exact same settings, run one after the other. Three different results (although they are similar).

It sure makes testing small variations to the settings difficult. I'm never sure if the settings have made a difference, or it's one of these variations occurring.

danw2 - np + ma cross robot

The idea for this robot came from a forum post on BabyPips.com, an excellent Forex education site. The result is the "danw2 - np + ma cross" robot.

There is a regular series of trade updates from a very similar system called Trend Catcher 3.0. They seem to be making it work, but I never could, especially not with the settings they're suggesting.

To be fair, I get bored of trying such systems and try to automate them, which is where my failures happen :)

Running this robot through the back test optimizer, it suggested settings quite different from the original system - wider moving averages, slightly different stops, and I also added a "Dema" setting, which is actually using simple moving averages, and not exponential as the name suggests (Dual Exponential Moving Averages, or something).

Here's the results:

(Back test on 60 minute EURUSD chart, 1-1-2011 to 1-1-2012)

Equity curve is so-so. The statistics aren't impressive with a mere 36% win rate, but the wins are almost 3 times bigger than the losses. A 13% profit risking 1% of the account on each trade.

I've pretty much given up on moving average based robots. I could never get them to work (a couple I'm demo testing now use pending orders to attempt to catch moves across moving averages, but they're not spectacular).

I'm afraid to run this on GBPUSD, which generally spits my robots out. We'll just be happy and satisfied with these results, shall we?

I also automated their 3 Ducks system, with very poor results. It looks like they're doing a little better, however.

Robot Results 30th April - 5th May

A few days late getting results this week, so the charts have already started showing this weeks activity, but the results are from proper period.

The current balances are from today, however, and include any trades placed last week which closed today. Several robots made large gains with the market gapping in their favor on open.

Also, taking a screen shot of the MetaTrader window now pastes to my local computer paste buffer - it didn't used to do that before. So the charts may be slightly more cropped than before.



Inside Bar

Last week up $217, this week pulls in another $403. It looks like the stop loss on the AUD could use with some reducing, but then again these are the optimized settings from a back test. Current balance is $5506.



Pending Reversal Candle

Last week made $83. This week it performed impressively, balance was up around $5800, but then it suffered two big losses, and ended up $326 down. Balance is $5033.




Daily Close Breakout

Last week lost $706. Yikes! This week I adjusted the stop loss and added two more currency pairs - USDCAD and GBPJPY.

Lost another $477, balance is $3732.



Weekly Breakout 2

Previous favorite system, has suffered a string of losses. Last week loses a staggering $1089. This week I adjusted the stop loss, added USDCAD, and it closes the week up almost $79. Balance is $4367, because several orders carried over the weekend closed with a $500+ profit.



Open Close Bot H4

$196 down last week, down another $220 this week. Did have a few orders in it's favor at market open today, though, so balance is $4664.

Felt so sorry for this robot I attempted to "fix" it using various methods, but none proved fruitful. I can't believe this robot had such a run of good luck at one stage. It really seemed to be "working" at one point.

--

So that's it. A bad week all around, except for Inside Bar, and perhaps Weekly Breakout 2. The breakout bots may need their settings optimized properly, as setting the stop loss lower isn't necessarily a good fix.

Increasing the breakout box size seems to help a little with some other robots I'm testing, so we'll see what happens this week and I'll consider further alterations to this robots depending on the outcome.

EURUSD - 2012-05-04

Hark! What have we here? Another doji candle! Goddamn. EURUSD moves one day and causes excitement. Perhaps it will keep going? Perhaps it will DO GODDAMN SOMETHING. But no. Another indecisive day.

(Daily EURUSD chart)

It's amazing how much money some of my robots managed to pull off the big move over one day. If only it could have lasted another day. Oh well. 

Below we might see some explanation as to why it didn't do anything. Price is in the Ichimoku Kinko Hyo cloud, and touched the bottom. Stalled. Stall-age. Going nowhere. 

(Daily EURUSD chart with Ichimoku Kinko Kyo)

EURUSD - 2012-05-03

Charts are interesting. Things have been slow on EURUSD lately, yet the long term charts are still providing useful and valid information. It just doesn't seem like it when you're waiting weeks, even months, for something to happen or to see if price will, indeed, respect various technical levels, or if technical analysis has finally broken down and chaos reins. 

Here is a triangular pattern forming on the daily EURUSD chart, which shows price not wanting to go below 1.30, and now, it's hit a bearish trend line 3 times. 

3 times! That's verification, folks! It feels like magic is working. 

(Daily hour EURUSD chart)

From the above chart, where price has most recently met the bearish trend line, we can see a pin bar form. This is often a clear reversal signal. 

Actually, there are two. Here they are high-lighted:

(Daily hour EURUSD chart)

While things have seemed slow, there was an actual bullish trend on the 4 hour chart, shown below in orange. But see how price last night broke through the trend line? This is after forming the pin bars on the daily chart and hitting the bearish daily trend line. 

(4 hour EURUSD chart)

So exciting! Technical analysis works!

Where price is going next, who knows. But let's not get too excited, as it's entered the "cloud" on the daily Ichimoku Kinko Hyo chart, so it may slow down. Oh boy, the EURUSD slowing down? Never!

(Daily hour EURUSD chart with Ichimoku Kinko Hyo)

I've pretty much decided to start focusing on other currency pairs for my trades, but I'll keep using EURUSD for analysis purposes on this site. But there probably won't be many predictions or trade ideas until the EURUSD does something significant. I certainly don't feel comfortable predicting anything at the moment, except the usual sell below 1.30 and buy if price breaks above the daily bearish trend line. 

"danw2 - np + trend + bb reverse" robot

This robot is loosely based on "GodBot". It works by determining trend direction on a higher time frame using DEMA, then trades pull backs determined by Bollinger Bands on a lower time frame.

It was returning some small gains in back tests, and then I decided to increase the stop loss, and it got better results yet. While not spectacular, the test was run on GBPUSD, which chews up a lot of my robots, and has an 87% accuracy rate and wins on average 8 times in a row. 

(Back test on 5 minute GBPUSD chart from 1-1-2011 to 2-5-2012)

It hasn't yet had it's settings optimized. The downloadable version has the settings that this test was run on already plugged in.

A while ago I read something about management of stops and take profit levels, as well as money management, being much more important than any particular system. Looking over some of my charts, it looks fairly simple to make money, yet incorrect stop loss and take profit levels are eating up what should otherwise be good trades.

Hence the reason for increasing the stop loss with this robot.

Often when optimizing robots, results suggest stop loss settings much larger than the 30 pips or so I usually like to use. In the past, I used much larger stop losses, but the results were very random. Sometimes I would get lucky  and trades would swing back in my direction, but I prefer to eliminate such swings if I could by using the smaller stop losses.

Looking at my charts and seeing the different scales the different currency pairs are in, I can see that 30 pips is nothing to some of them. So perhaps using large stop losses is better. Generally though this means I need to look for potentially bigger trades, which means using higher time frame charts, which means less potential trades and more waiting.

This particular robot seems quite accurate, so we can seemingly get away with making more trades on a lower time frame, but still use a larger stop loss.