Steve Hopwood's Fib Target-level retracement auto-trader: compliments Simone

Bing!

Thanks, mate.

I'm a little late, but I have to wish my boy a happy birthday.

Happy Birthday Steve!!!

Hope all is well.

Happy birthday Steve!

Guys, I know Steve in his relentless industry coded up a Stop and Reverse trade variant of the FLTR EA based on my suggestion. For this I am truly grateful. Actually, the idea was originally for NB's 4HR system because I noticed that his original EA (for the 4th version, based on EMA3(open) vs EMA3(close) crosses in the correct direction for a reversal trade would occasionally enter on a minor retrace in a strong trend, and the resulting drawdown would be very large. NB himself uses a 2 4 2 L2, L3 system to re-enter progressively as the trade goes against him on the premise that the trend will eventually reverse. This approach is obviously successful as NB's trading demonstrates - if you have the steel balls to wear the drawdowns and does seem to suffer during major crisis such Greece/Euro in early May when a strong trend is created and continues.

NB's system is fantastic for identifying the regions in which price is likely to reverse direction. This is its key strength.

The variation I was proposing to tackle the drawdown issue (which I personally find scary) was to initially enter a trade with trend once these regions are hit (setup), and then use a Stop and Reverse trade when the EA determines it's time to enter for the trend reversal (trigger). This would mean a number of short trades (hopefully with small profits but possibly small losses) with the occasionally big winner before the counter trend reversal trade begins. The advantage here is that instead of drawn downs you get profits for any long runners - assuming of course that that conditions for entering the reversal trade are sufficiently robust that the EA doesn't enter on any minor retrace. Tweaking this balance (around trigger entry conditions) is the key issue, and perhaps difficult to code. As I understand the NB system, the sixths and bollinger bands alert you to price being in the right zone for a retrace (setup) and the script based EA is then dropped on the chart to wait until conditions are correct for entry (trigger). This used to be the SMA6 cross that Squalou mentioned In version 4), but this has been superceded in version 5 (i'm not sure by what yet). The variation I'm proposing is enter a trade with trend at setup, then Stop and Reverse at trigger.

I'm quite certain that a variation using SAR will prevent large drawdowns and may give traders the option to not to use NB's L2 and L3 re-entry system (which as stated does work, but requires BOS to use). NB himself defended his system (as he should) and urged me to see the futility of my suggested variant by providing the attached chart, which shows ticks for NB v5 entries. I'm not certain he's right, so I have marked this chart up indicating where I would expect initial with trend entries (red circles) which might be small profits prior to the Stop and Reverse (which would be the ticks as per NB v5). The red crosses indicate probable losers or super short with trend trades (mark ups are hard to see - expand the chart).

I will attempt to code up this variation based on the current experimental NB v 5 EA, but I am time poor, and not an exceptional mql coder, so if anyone wants to assist for this variant I'd appreciate it. The goal would be to create a backtest chart like that submitted by macman which did not have a big negative spike in early may, and in general to reduce the drawdown on the EA (measure in backtest), while maintaining profitability. There may be a tradeoff there (between profits and drawdown), but IMHO having the option to avoid the larger drawdown with NB's system would be be a good thing.

cheers
f451

Just to let you know that I have not forgotten this. It is just rather difficult for me as it involves working out a lot of logic with which I struggle.

Also, I wanted to draw a line on the chart to mark the start of the week, to make sure that I have sq's code working correctly. I had trouble with this because I forgot to change the tf to H4 and couldn't make the code work.

I have finally managed to locate the start of the week on any sub-D1 tf, and so can move things forward.

Cheers sq; your code works brilliantly.

Steve, check out the attached for me, please.

Drag the ea onto a 4H chart and see if it looks correct.

The yellow line marks the start of the week.

The first turquoise line marks my understanding of the 3rd calculation point in your instructions.

The second turquoise line marks my understanding of the 5th calculation point in your instructions.

Does this look about right?

Another question. Doesn't this limit trading to the final two days of the week?

Hey Steve,

This is really great! Having my name on an EA is a bit surreal.
I checked the line trajectories on the UJ, and it matches my manual measurements perfectly well. To answer your second question, yes originally I had it in mind to work on the 4hr, so last two days of the week trading, but it may indeed have applications on lower TF's. I will be investigating that right away.

There also certain exception to buying/selling deviations, namely when the trajectory established is too steep, and you have radical mid-week trend change, and price breaks in the opposite direction as a runaway. That type of scenario can be solved in a couple of ways. One would be disallowing trades when the angle is above or below a certain degree, or another would be simply changing the trading parameters in such instances.

I'll attach a chart to show you what I mean. This is a relatively rare type of week, maybe 1 in 7 as far I can tell.

But this is already an amazingly cool and useful indicator! Great work!

After some experimenting on the Lower TF's here are my findings:

1) This works beautifully on lower TF's.
2) This has a lot more value as a breakout mechanism than a retrace mechanism, or countermove mechanism.

It seems to work equally well on the M30 as H1 (actually on closer review 30 min seems better). So, on the M30 you'd be ready to go after just 9 hours of collecting data.

My idea is this:

First of all its pretty obvious that this indy does a great job of finding the probable direction that price is going to break during the week.

If price is lingering above the average line, you have a high probability of a high breakout, and vice versa.

But you want to verify and you want to find best position to trade it. Using hourly ATR is tough IMO. So lets ditch the hourly ATR idea.

Using an RSI at standard settings, there seems to be a great solution. Simple, if price is above average bar, and RSI is low say less than 37, then buy.

If price is below average bar and RSI is high, above 63, sell.

This fulfills two functions:
1)Gives you superior buy/sell price, and entry near likely breakout point.
2)Shows that price can be put into oversold/overbought and still maintain above or below average line, thus verifying trend direction. It will also prevent triggering trades when ranging around average bar.


I'm not sure about the exact settings for the RSI(ie trigger at 60, 63, or 65-if you include as variable input can be tweaked in backtesting), but I'm pretty sure that this would be a very profitable direction, and seems to solve all the problems I've encountered so far.

One more thing. If you get a situation where price stabs average line when RSI is oversold, or overbought(as with the UJ example below), you are concerned with the close of the bar. Ex. for potential buy: RSI 33, if bar closes above, buy even if price stabbed line. If bar closes below-no trade.

Charts attached.

Here are some examples on the 15min, which could also work. Higher and Lower RSI settings on this TF. Great examples on the EU and GU, NZDJPY shows things can be going less than perfectly but still work out very well.
M30 is still probably your best bet though. Going lower than M30 gives more sensitivity but will probably occasionally misjudge trend, as in NZD example.

You will be amazed at what you are about to get.

Just ironing out some wrinkles.

Ok Steve, here it is.

You can trade either method - the break of the trendline in the direction of the break, or the retrace from the Atr/2 trendline.

Backtesting on EU 15M to make sure the code is working, I noticed this trading the breakout method: a trade would hit its tp then another would immediately be taken. The robot would see there is no trade open, that the market is above/below the trendline and take the trade.

In the backtests I ran, this was still profitable.

I added a filter to make sure the robot did not enter a trade when the market is more than MaxPipsDistanceFromTrend. This killed profitability immediately.

MaxTrendSlopeAllowed should turn the robot off if the distance between P3 and P5 creates too-steep a slope.

Play with all these settings and see what you can arrive at that is successful.

The StopTrading input allows you to turn off the robot's trading capability and use the ea as a trendline drawer for manual trading.

Would you like the usual trading time filter?

Have fun

Hi

First, thanks to all those who have worked so hard on these EAs - it's really great stuff to read and learn from such knowledgeable people. I'm so lucky!

I loaded up the Muncher on a few pairs on a 15M timeframe and I'm getting the
order send failed with error (130):invalid stops.

Thanks again for all your great work.

Just letting you know.

Good Morning Steve,
This thing just kicked out my global TP at +750 pips so my test ends for this week. This script is really high on the fun meter!!!
Started it on Monday at the start of the Tokyo session.......

whip

P.S. Scoobs (as normal) looks to be right on....Thursday is the day

Wow! These are really promising first results. I am indeed amazed that this bot can backtest so positive on first version.

Just a note: If backtesting make sure to start your period on a Sunday, otherwise I think results might get skewed.

The big problem I'm seeing is not with the order placement, which seems to be getting carried out according to its program.

About this: You see clusters of losses in the same minute or a minute apart. This is happening because price is sliding so hard, it buys, gets stopped out, and rebuys in the same minute. So, this can be reduced with a minimum time delay between buys. 2 minutes should do the trick, but as a variable input it could be tweaked.

The big problem I get is that it won't issue any sells on any pairs, just buys. Even with these two problems bot is making money in backtests, and some results are really very impressive.

Settings Analysis:

GJ, EJ, like tighter(larger) settings on RSI, wider Stops and TPs, as you might expect with volatile pairs:

GJ EX:

Test from 2/28/10-7/12/10, all based on 15min

Settings:
SL: 1200
TP: 2200
RSI: 73, 27
Max Dist from Trend: 1500
Lot Size: .01, 1 penny
Max Slope: 100000 (shouldn't matter when trading only breakout so I set it way up)

Results:
Net Profit: +17.98
Max DD: 5.98! (profit to Max DD 3 to 1!)
Gross Profit: 27.58
Gross Loss: 9.61

EJ at similiar settings:

Net Profit: +10.67
Max DD: 8.71

And I'm sure this could be better if it would take sell positions on weeks when trend points down.

UJ:
Settings:
RSI: 65, 35
SL: 1000
TP: 1300

Results:

Net Profit: 3.79$
Max DD: 3.61$

All Pairs seem to like this bot if the settings are tweaked for their ATR and volatility. Smaller ranging, less volatile pairs need smaller stops, tps, and looser RSI settings. Larger ranging more volatile vice versa.

Only pair that doesn't seem to like it is EU somehow. I got it to show a profit, but 1.50$ max. (Actually I think this is because testing period I'm using is last 6 months which is mostly bear for EU, and the bot won't make sells.)

Thats about it, only fixes I see from my end is:

1) won't take any sells
2) needs a time delay option as variable input to reduce loss clusters

Once again, great great work Steve! Don't know how you do this stuff so fast...

Here is a pic from my control demo at 1:30 PM GMT.

It has been interesting. I used the script to send the trades in this order:

 

1. managed by mptm, to close Thursday mid-day
2. unmanaged, to close Thursday mid-day
3. unmanaged, to close Friday mid-day

The few minutes in between sending baskets 1 & 2 made a huge difference to the pips count. Both were losers: 1) by about -650; 2) by about -90.

The Friday-close basket is still open and hoping for a change of trend to help. It could do as the big losers at the bottom of the list out-weigh the winners at the top by a large margin.

I noticed some of the bad trades have simply got worse and worse. Next week, I am goingo to adopt a SAR approach to the unmanaged Thursday basket. I shall have a look at the individual chart and see what the overall weekly direction has been, and perhaps do a sar at -50 to -75.

1. I saw a bug in the sell-order send, but it still is not taking sells. Can any fresh minds see why not (void LookForTrendBreakTrades())?
2. Added SleepMinsAfterTradeSent. This is a bit of a blunt instrument because it sends the bot to sleep after a successful trade-send, not after one closes.

One thing I notice in the backtest I just ran is this: a trade can hit sl but the robot instantly takes a fresh trade that often moves to tp. This means the original trade closed even though the market was still on the correct side of the trendline.

Shalll I modify the sl so that it becomes StopLoss pips on the other side of the trendline, instead of a hard stop?

The bug fix in the sell code has solved the no-sell thingy. The bot is happily taking sells in a GJ backtest right now.

Hi Steve

Thanks for the new bot.

I have been testing a lot of different combinations of settings for the Fib Retracer SAR bot, but alas nothing is looking good in backtesting on a number of TFs and currencies. The tests on the original FTLR bot are still running, although they have seen / are seeing some quite scary floating DD.

Thanks for your efforts. Sorry I don't have really good news about these bots.

Cheers

Invisible

Not to worry. We can but try these things.

Hey Steve,

Glad to hear about the bug fix for the sell. Trying to figure a few things out. Check out this report for the EJ. Shows 6 repeat losers starting buy at 117.55. Sl's 100 pips 6 times in a row down to 111. In the report I've got max pips from trendline set at 1500, but then I changed to 100 and got same result.

(Bot still comes out very profitable )

That aside though, something isn't right there. If tradeable max distance from trendline is 100 then should only be able to take one loss. Even without that, no way it can trade 600 points over the average line as a first trade of the week. I guess in some sense its possible but it doesn't seem likely.

Maybe a glitch with sell thingy? Anyway check it out.
As far as the idea of using the average line as a stop point instead of hard stop I think it could work providing:

1) the open trade at max distance from trendline function is working properly.
2) you put a min SL on also, maybe 20 points, to prevent trades opened really close from the average line from getting stopped out just by brushing the line.

Otherwise, the sleeper function seems to be working for the positive in most cases. Sometimes it doesn't make a difference. Sometimes it reduces DD a lot, and sometimes it reduces profit by a very small fraction. 95% positive results. 5 min setting.

I got a 12.06$ profit to 3.45$ max DD on EJ using 5 min sleeper on some settings or the other.

Thats my whole report. Thanks!

PS, can you post the version with bugfix thats taking sells? TY.

------------------------------------------------