scooby-doo BOJ JPY pairs trading robot

It just closed the basket in profit Small.. but profit nonetheless!



You guys are amazing!

Scooby original BOJ basket closed in overall profit - approx2 hours

Original 2.3 hit profit took about 2 hours
worth the wait

FXDD Live

Maybe we should post broker and live or demo so others can see

Snap. My first live trade and first profit. OK so far

One hour seven minutes here. 250 bucks...1000 so far this week. As scoob said- Like takin' candy from a baby

Well, I did code it without prompting, so it would be a bit strange if not.

Default retracement appears to kick trades in a bit too soon and started losing money immediately, so now I am demoing defaults except for UseH1ATRrt = true. This has started out much better - several open trades, no losses, a couple with 50% closure and stop moved to be, a decent amount of draw up; early days.

Cheers Nick.

Scoobs, great start to THG V6. It started with two humungous losers, then spent the rest of yesterday and today recovering to net +111 pips.

Yes, not doing so hot today at all. Big DD and actually losing money. Hopefully some further tweaking in it's future, but will keep it running to see if it can redeem itself.

impressive!

but can't wait to see a loss :-)

just want to see how a loss looks like, you know...

Bear with me on the length of this....

Here's a suggestion on how to get the TP > SL and to turn this, and maybe others, into robust positive expectancy systems where the average winner ($ size) > average loser ($).

With Zoogle's sys, we are much more likely to have a successful trade if the entry price is near the current day's ADR (ADR of last 10 days) for the pair. This is a far better and objective measure of OB/OS than arbitrary MTF parameters for RSI, which are always a crapshoot (arbitrary parameter game, which never ends). Indicator parameters are always a crapshoot game.

Could use this ADR calculation to filter all trades (say, only take the trades within a 10% threshold of 10 day ADR, i.e., enter only at 90%+ of that day's ADR already). This should be made user customizeable, reflecting how conservative or aggressive one wants to be. And perhaps make it customizeable per pair.

I don't know what indi you would want to use for this. I have one done by Hanover a while ago.

With this method, one can have a shot at realistically adjusting the TP up from 0.25 of ADR and SL down from 0.75 of ADR without killing the system with:

a) random noise (EAs of course always degrade and are unworkable with tight SL's)

b) SL's three times the TP, which will kill dead any system (e.g., Turtles had very wide stops, but their TP's were still a huge multiple of those wide SL's).

There's no way any system can be accurate enough to make up for a SL three times the TP. No way, jose. http://cdn.forexfactory.com/images/smilies/yim/sad.gif Average losers will always be larger than average winners = negative expectancy system, big time.

Goal is to realistically get the TP > SL, and to do this by only taking trades that are objectively OB/OS, meaning the current entry price is very close to or even above the pair's 10 day ADR. Let the user customize the percentage, and allow the percentage also to be set to values above 100%.

If a pair is OB near 100% of its ADR, many folks would be very comfortable taking a sell there, and putting a SL at 30-50% above that (means the stop is at 130%+ of that pair's already reached ADR).

The odds are good there that you have S/R stacked behind the back of the trade, so the odds of hitting a 0.50 of ADR TP are good (if price is at 100%+ of ADR already, it's pretty likely one can at least get a 30-50% of ADR retrace back, ie. 0.30-0.50 ADR, and that's our trade back with the 200 ema trend). Re one person's retrace is another's trend. All the while you can have a realistic relatively tight SL of say 0.30 of ADR above that.

The point is to avoid these big losers, the account killers, that are 3 times the size of the winners, and to hit high probability 0.5 ADR "retraces"/trends (>SL size) back in the direction of the 20/200ema overall trend on pairs already at 100% of ADR http://cdn.forexfactory.com/images/s...m/thumbsup.gif

This is the only way I see of turning this system into a robust positive expectancy one.

Doesn't matter if trades are infrequent per pair if one is running 15 pairs 24 hours a day. (Nothing will work till we can get the avg. winner size in $s > avg loser in $s = realistic TP > SL). If someone wants lots of trades on a negative expectancy system, go ahead: just kills your account faster and builds massive commissions for broker. The latter and their bank clearers will both be "laughing all the way to the bank"/other banks, whatever.

Regards

Thanks, Steve,

and I think that there is a logical reason to make time expiration of an order by default.

That is, we should forfeit the postions if it has taken strange-long time.

the logic goes like this:

an EA is try to gain an edge in fitting the market (in fact, predicting).

if the prediction is valid this time, then usually, in 1 or 2 hours, the market should move towards the direction EA predicts.

if the market did so, we TP. great!

if the market instantly go against us, we SL. it's ok. However, due to the set-up the EA applies to enter, this SL situation seldom happens. very RARE! if it does happen, ok, we accept, but this is not important to the EA performance.

a more common situation is:
if the market did not move, or slowly move a little against the positions after quite a few hours, this in fact announce the set-up for the entries ARE NO LONGER VALID. or at least, very likely not valid. now these positions are very likely liablilities!

now, these positions, after a few hours, not going to EA's predicted direction is in a coin-flip situation. 50% chance to go up, and 50% chance to go down. However, we know the original BOJ bot's TP is much smaller than SL (for example: 25tp, 100sl). if we let these positions do complete coin-flip situation, in the long run, we lose 4 vs 1 win.

To sum-up, the positions not moving to EA's favored direction in 1~2 hours are milk that turning sour. it might be still nutritious, but not worthy the risk. The order expiration is defintely needed.

someone, if interested, may go back to see how those SLed positions developped. in my limited testing experience (1 week), those SLed positions seems take quite a few hours (even 1 day) to exit.

i may be wrong, just for discussion.

From getdown71:

"My suggestions for timers;

Buy = 150 minutes
Sell = 300 minutes"

Looks good to me.


Above comments are right on bull's eye, to me personally.

This morning, sell basket opened and closed with 1% profit, within 150 minutes with time expiry setting on, on my live account with MXG (the so called first Swiss forex bank).

Can't wait till closing session of Tokyo!

Short CAD/CHF in Zoogle's version of the 20-200-ema RSI sys., etc., at 1.0279. @ red vertical line.

That's nearly at 44 pips of average daily range of 101 pips (ie. at about 44% of ADR), ie. it's circa 44 pips above the day's low so far of 1.0245.

Not a real high odds place, though note it's equivalent to the R1 pivot, so it's not a bad place. Also, if we assume price will range down below the day's open an equal amount of approx. 40-50 pips to equal the 10 day ADR avg. (text at bottom left of chart) of nearly 100 pips, this is a good entry point: it's near 50% on one side of the day's open (so de facto 100% of ADR, if we assume a bi-modal distribution of prices around the day's open).

100 pips of ADR up from the day's low so far would up at about 1.0345 (near the R3 pivot, by coincidence). Seems unlikely we'll reach that, so far, so maybe an entry point filter for 50% of ADR, about where we are now, is more realistic, assuming a bi-modal distribution of prices in the average day around the open.

Another approach: set the filter so entry point always has to be within 5% etc. of any classic pivot (R1,2,3, or S1,2,3) OR at least 50% of the ADR.



http://i50.tinypic.com/inqtd0.gif

Thanks a lot!

Look, this ADR application is golden, in my humble opinion.

ADR concept may also be found for those breakout system. It is always NOT appropriate to fix the TP/SL.

Only 1 thing, there is still one "crapshoot" about your system -- ADR(10). This 10 is a fixed number without a legit reason. However, since there is no holy grail, a profitable EA is the one with least "crapshoot" issues, here you have only 1, seemingly.

Thanks.

On fixing the TP and SL, the problem with EA's is that they alway fix these by definition because they are mechanical systems. At least the ADR concept, which is already in the system (that's good), is adaptive to recent market volatility.

Yes, 10 ADR is a somewhat arbitrary number, but it's not really a crapshoot. I think you will find that once we get up to ADR for two trade weeks, the ADR per pair will be remarkably stable through most market conditions (i.e. 5-15 day ADR etc., +/- 5 days from 10 days, will be very similar values). It's a robust value, reliable over a range of parameters/lengths once one gets up to a week or more. This is a good basis for a variable in an EA.

For example, I just checked the values for CAD/CHF (daily pip range averaged over these number of days):

5 days - 88 pips
10 - 101 pips
15 - 101!
20 - 99! (are you getting the picture now?)
30 - 93
100 - 94
300 days - 123

Yes, it was more volatile if you average it over a year or so/300 trade days etc., whatever, but who cares: those are older conditions. In the last month, it's pretty well 90-100 pips. 10 pip difference is "peanuts". The point is the value is stable at about 10 days, and I think this is the case for most pairs.

RSI, MACD, ADX etc., even EMA's, any indicators, are all much more arbitrary and will vary enormously, NO matter what period one chooses. ADR will not move that much once one gets above a week or so, because each pair basically has a characteristic stable volatility (eg. GJ will always be much more volatile than EURCHF, etc.).

This is stable/robust, reliable basis on which to make OB/OS decisions for entry points and exits.

To address the unfavourable R:R ratio I've lowered my sl to 50% and left tp at 25.

I like your idea bit9ret, let's see if I have it down pat:

- user defined ADR threshold percentage (l'll use 50)
if current price < today's open + (0.50 * ADR) >>> allow buy

if current price > today's open - (0.50 * ADR) >>> allow sell

As for your cad/chf position, I don't see how that could have been a valid sell since the h01 ema has been above the h04 ema since jan. 26th meaning we only buy.

What I've found to be the biggest problem personally, and I'm surprised no one has brought it up (maybe it hasn't happened to anyone) is the unreliability of the Economic News indicator. I've had it return false on 2 major news events so far.

Interesting idea indeed, and on my own, i can't trust any calendar news over there, pure wasted time to look at them.

Yes, I´ve noticed it as well, and think Scooby mentioned it a while ago. The problem seems to rely on the fact that at high-impact news times, the FF Calendar page gets overloaded and it cannot be accessed, so the indicator returns a false value.
A possible solution is to make the indi check the news every, say, 12 hours, and register the scheduled release times for the whole next 12 hours. The EA should then have to have a function that automatically disables trading from x minutes before and until y minutes after the scheduled news release times, without a need to check them when they happen. I don´t know how and when it checks the news right now, though. It may as well close all open trades, even if they are at a loss (all the logic of the EA is wrecked by the the price movements caused by red news, so if a trade reaches such a point still open, leaving it open is just as tossing a coin while swimming surrounded by sharks).

One thing I have noticed with THG v6 and Zoogle´s EA is that the losses occur in streaks. I´m waiting for the weekend to do some analysis on their performance and will check if the losing rows occurred at news times. Will report on that during the weekend.

Cheers,

Checking it every 12 hours wont work either as it can get updated quickly.

What I would do is change the news filter so that if it cant open the file then assume there is news. Better to be safe than sorry.

Scoobs.