Metatrader expert advisors dissected.
NeoTradEx live forward test
Trading style: uses a neural network to enter the market when a trend is identified. Its positions have a stop loss, take profit and trailing stop calculated based on the volatility (the default settings don’t include a TP; the positions are closed by the trailing stop).
Currency pairs: pretrained for EURUSD and USDCHF but it can be trained for any other pair
Timeframe: pretrained for H1; it can also be trained for other timeframes
Price: €199
License: currently not limited
NFA compliance: yes (with MaxOrders set to 1, the default setting)
Refund policy: 60 days
Read more at the NeoTradEx website
Buy NeoTradEx
Note: this is an EA that is based on a neural network. Generally, neural networks have to be trained to obtain good results and NeoTradEx comes pretrained for H1 EURUSD, USDCHF, EURJPY, EURGBP and USDJPY. The robot also caters to those who like to fiddle with stuff and the user has the option to train the neural network on his own. See the user guide on the NeoTradEx website for more info on that and the theory & mechanism behind it.
Birt’s forward test
Settings: default (using the provided set files with Fixed off and Capital 0.05)
Started: 09.04.2012
Broker: PrivateFx
Account type: live, micro
Starting balance: $300
Current version: 1.5
Currency pairs: EURUSD, other pairs removed temporarily due to some issues.
Note: the account is using the neural network with the training files supplied by the vendor (available out of the box). The first three trades were taken with the Fixed parameter set to false and the Capital parameter set to 0.1 so they had a rather high lot size; that seemed too risky so the next 6 trades were executed with Fixed enabled and a lot size of 0.1; I finally decided to run it with Fixed off and Capital 0.05 because I wanted to take advantage of the fact that the EA risks a fixed percent of the available capital – this setting was chosen based on the drawdown exhibited in the backtest results starting from mid 2010.
Official account
Settings: unspecified
Started: 02.05.2012
Broker: Alpari UK
Account type: real
Initial balance: $529
Currency pairs: EURUSD
Backtests
I’m stressing this a bit more: please bear in mind that this is a neural network EA and it is trained on past data because… well, that’s the only way to do it. This means that the backtests are somewhat meaningless since the EA is already prepared to deal with that data. Nonetheless, I ran a few backtests for those interested and as you will see below, it’s evident that the neural network was trained on the past couple of years. As a side note, I’ve followed the vendor’s forward test for a while before starting this one to ensure that its live performance matches the one seen in the past couple of years from the backtests.
All the backtests were performed using Dukascopy tick data with variable spread.
NeoTradEx v1.4 EURUSD H1 2007 2012 tick data backtest, real spread, default settings (Fixed on)
| Symbol | EURUSD (Euro vs US Dollar) | ||||
| Period | 1 Hour (H1) 2007.03.30 16:01 - 2012.03.26 12:59 (2007.04.10 - 1970.01.01) | ||||
| Model | Every tick (the most precise method based on all available least timeframes) | ||||
| Parameters | learn=false; Verbosity=0; Depth=10000; ValidationLength=200; DeltaOpen=0.8; DeltaStop=2; VolatilityStop=96; VolatilityTake=0; VolatilityTrail=2; Fixed=true; Capital=0.1; Lots=0.1; MaxOrders=1; MaxTraining=100000; StopMSE=0; | ||||
| Bars in test | 31157 | Ticks modelled | 71124282 | Modelling quality | 99.00% |
| Mismatched charts errors | 0 | ||||
| Initial deposit | 1000.00 | ||||
| Total net profit | 2283.88 | Gross profit | 5358.95 | Gross loss | -3075.07 |
| Profit factor | 1.74 | Expected payoff | 1.31 | ||
| Absolute drawdown | 435.22 | Maximal drawdown | 447.96 (44.23%) | Relative drawdown | 44.23% (447.96) |
| Total trades | 1747 | Short positions (won %) | 906 (86.09%) | Long positions (won %) | 841 (89.77%) |
| Profit trades (% of total) | 1535 (87.86%) | Loss trades (% of total) | 212 (12.14%) | ||
| Largest | profit trade | 35.78 | loss trade | -41.96 | |
| Average | profit trade | 3.49 | loss trade | -14.51 | |
| Maximum | consecutive wins (profit in money) | 149 (573.55) | consecutive losses (loss in money) | 5 (-49.83) | |
| Maximal | consecutive profit (count of wins) | 573.55 (149) | consecutive loss (count of losses) | -83.43 (2) | |
| Average | consecutive wins | 9 | consecutive losses | 1 | |
NeoTradEx v1.4 EURUSD H1 2007 2012 tick data backtest, real spread, default settings, Fixed off, Capital 0.01
| Symbol | EURUSD (Euro vs US Dollar) | ||||
| Period | 1 Hour (H1) 2007.03.30 16:01 - 2012.03.26 12:59 (2007.04.10 - 1970.01.01) | ||||
| Model | Every tick (the most precise method based on all available least timeframes) | ||||
| Parameters | learn=false; Verbosity=0; Depth=10000; ValidationLength=200; DeltaOpen=0.8; DeltaStop=2; VolatilityStop=96; VolatilityTake=0; VolatilityTrail=2; Fixed=false; Capital=0.01; Lots=0.1; MaxOrders=1; MaxTraining=100000; StopMSE=0; | ||||
| Bars in test | 31157 | Ticks modelled | 71124282 | Modelling quality | 99.00% |
| Mismatched charts errors | 0 | ||||
| Initial deposit | 1000.00 | ||||
| Total net profit | 3218.55 | Gross profit | 5415.02 | Gross loss | -2196.47 |
| Profit factor | 2.47 | Expected payoff | 1.84 | ||
| Absolute drawdown | 205.96 | Maximal drawdown | 213.06 (21.16%) | Relative drawdown | 21.16% (213.06) |
| Total trades | 1749 | Short positions (won %) | 909 (86.14%) | Long positions (won %) | 840 (89.76%) |
| Profit trades (% of total) | 1537 (87.88%) | Loss trades (% of total) | 212 (12.12%) | ||
| Largest | profit trade | 49.53 | loss trade | -44.13 | |
| Average | profit trade | 3.52 | loss trade | -10.36 | |
| Maximum | consecutive wins (profit in money) | 149 (678.51) | consecutive losses (loss in money) | 5 (-43.79) | |
| Maximal | consecutive profit (count of wins) | 1130.02 (138) | consecutive loss (count of losses) | -72.89 (3) | |
| Average | consecutive wins | 9 | consecutive losses | 1 | |
NeoTradEx v1.4 EURUSD H1 2007 2012 tick data backtest, real spread, default settings, Fixed off, Capital 0.1
| Symbol | EURUSD (Euro vs US Dollar) | ||||
| Period | 1 Hour (H1) 2007.03.30 16:01 - 2012.03.26 12:59 (2007.04.10 - 1970.01.01) | ||||
| Model | Every tick (the most precise method based on all available least timeframes) | ||||
| Parameters | learn=false; Verbosity=0; Depth=10000; ValidationLength=200; DeltaOpen=0.8; DeltaStop=2; VolatilityStop=96; VolatilityTake=0; VolatilityTrail=2; Fixed=false; Capital=0.1; Lots=0.1; MaxOrders=1; MaxTraining=100000; StopMSE=0; | ||||
| Bars in test | 31157 | Ticks modelled | 71124282 | Modelling quality | 99.00% |
| Mismatched charts errors | 0 | ||||
| Initial deposit | 1000.00 | ||||
| Total net profit | 18547722.54 | Gross profit | 22323392.36 | Gross loss | -3775669.81 |
| Profit factor | 5.91 | Expected payoff | 11621.38 | ||
| Absolute drawdown | 955.02 | Maximal drawdown | 1494983.42 (14.72%) | Relative drawdown | 95.65% (988.06) |
| Total trades | 1596 | Short positions (won %) | 835 (85.27%) | Long positions (won %) | 761 (88.96%) |
| Profit trades (% of total) | 1389 (87.03%) | Loss trades (% of total) | 207 (12.97%) | ||
| Largest | profit trade | 1460666.16 | loss trade | -1014488.74 | |
| Average | profit trade | 16071.56 | loss trade | -18239.95 | |
| Maximum | consecutive wins (profit in money) | 149 (507427.11) | consecutive losses (loss in money) | 5 (-42.55) | |
| Maximal | consecutive profit (count of wins) | 9586560.40 (29) | consecutive loss (count of losses) | -1014488.74 (1) | |
| Average | consecutive wins | 8 | consecutive losses | 1 | |
NeoTradEx v1.4 USDCHF H1 2007 2012 tick data backtest, real spread, default settings (Fixed on)
| Symbol | USDCHF (US Dollar vs Swiss Franc) | ||||
| Period | 1 Hour (H1) 2007.03.30 16:01 - 2012.03.26 12:59 (2007.04.10 - 1970.01.01) | ||||
| Model | Every tick (the most precise method based on all available least timeframes) | ||||
| Parameters | learn=false; Verbosity=0; Depth=10000; ValidationLength=200; DeltaOpen=0.8; DeltaStop=2; VolatilityStop=96; VolatilityTake=0; VolatilityTrail=2; Fixed=true; Capital=0.1; Lots=0.1; MaxOrders=1; MaxTraining=100000; StopMSE=0; | ||||
| Bars in test | 31143 | Ticks modelled | 55473743 | Modelling quality | 99.00% |
| Mismatched charts errors | 0 | ||||
| Initial deposit | 1000.00 | ||||
| Total net profit | 1271.66 | Gross profit | 3492.12 | Gross loss | -2220.46 |
| Profit factor | 1.57 | Expected payoff | 0.83 | ||
| Absolute drawdown | 178.96 | Maximal drawdown | 335.25 (28.99%) | Relative drawdown | 28.99% (335.25) |
| Total trades | 1529 | Short positions (won %) | 802 (88.15%) | Long positions (won %) | 727 (86.52%) |
| Profit trades (% of total) | 1336 (87.38%) | Loss trades (% of total) | 193 (12.62%) | ||
| Largest | profit trade | 21.75 | loss trade | -33.80 | |
| Average | profit trade | 2.61 | loss trade | -11.50 | |
| Maximum | consecutive wins (profit in money) | 93 (221.67) | consecutive losses (loss in money) | 3 (-75.99) | |
| Maximal | consecutive profit (count of wins) | 231.80 (61) | consecutive loss (count of losses) | -75.99 (3) | |
| Average | consecutive wins | 8 | consecutive losses | 1 | |
NeoTradEx v1.4 USDCHF H1 2007 2012 tick data backtest, real spread, default settings, Fixed off, Capital 0.01
| Symbol | USDCHF (US Dollar vs Swiss Franc) | ||||
| Period | 1 Hour (H1) 2007.03.30 16:01 - 2012.03.26 12:59 (2007.04.10 - 1970.01.01) | ||||
| Model | Every tick (the most precise method based on all available least timeframes) | ||||
| Parameters | learn=false; Verbosity=0; Depth=10000; ValidationLength=200; DeltaOpen=0.8; DeltaStop=2; VolatilityStop=96; VolatilityTake=0; VolatilityTrail=2; Fixed=false; Capital=0.01; Lots=0.1; MaxOrders=1; MaxTraining=100000; StopMSE=0; | ||||
| Bars in test | 31143 | Ticks modelled | 55473743 | Modelling quality | 99.00% |
| Mismatched charts errors | 0 | ||||
| Initial deposit | 1000.00 | ||||
| Total net profit | 2571.63 | Gross profit | 4956.43 | Gross loss | -2384.80 |
| Profit factor | 2.08 | Expected payoff | 1.68 | ||
| Absolute drawdown | 62.34 | Maximal drawdown | 230.25 (19.71%) | Relative drawdown | 19.71% (230.25) |
| Total trades | 1529 | Short positions (won %) | 802 (88.15%) | Long positions (won %) | 727 (86.38%) |
| Profit trades (% of total) | 1335 (87.31%) | Loss trades (% of total) | 194 (12.69%) | ||
| Largest | profit trade | 38.17 | loss trade | -37.44 | |
| Average | profit trade | 3.71 | loss trade | -12.29 | |
| Maximum | consecutive wins (profit in money) | 91 (285.54) | consecutive losses (loss in money) | 3 (-32.75) | |
| Maximal | consecutive profit (count of wins) | 579.04 (72) | consecutive loss (count of losses) | -43.27 (2) | |
| Average | consecutive wins | 8 | consecutive losses | 1 | |
NeoTradEx v1.4 USDCHF H1 2007 2012 tick data backtest, real spread, default settings, Fixed off, Capital 0.1
| Symbol | USDCHF (US Dollar vs Swiss Franc) | ||||
| Period | 1 Hour (H1) 2007.03.30 16:01 - 2012.03.26 12:59 (2007.04.10 - 1970.01.01) | ||||
| Model | Every tick (the most precise method based on all available least timeframes) | ||||
| Parameters | learn=false; Verbosity=0; Depth=10000; ValidationLength=200; DeltaOpen=0.8; DeltaStop=2; VolatilityStop=96; VolatilityTake=0; VolatilityTrail=2; Fixed=false; Capital=0.1; Lots=0.1; MaxOrders=1; MaxTraining=100000; StopMSE=0; | ||||
| Bars in test | 31143 | Ticks modelled | 55473743 | Modelling quality | 99.00% |
| Mismatched charts errors | 0 | ||||
| Initial deposit | 1000.00 | ||||
| Total net profit | 11871491.80 | Gross profit | 13159207.02 | Gross loss | -1287715.22 |
| Profit factor | 10.22 | Expected payoff | 8301.74 | ||
| Absolute drawdown | 742.27 | Maximal drawdown | 941885.63 (9.12%) | Relative drawdown | 91.89% (2920.84) |
| Total trades | 1430 | Short positions (won %) | 749 (87.45%) | Long positions (won %) | 681 (85.90%) |
| Profit trades (% of total) | 1240 (86.71%) | Loss trades (% of total) | 190 (13.29%) | ||
| Largest | profit trade | 910621.32 | loss trade | -412938.23 | |
| Average | profit trade | 10612.26 | loss trade | -6777.45 | |
| Maximum | consecutive wins (profit in money) | 93 (20197.60) | consecutive losses (loss in money) | 3 (-389.28) | |
| Maximal | consecutive profit (count of wins) | 8167664.89 (38) | consecutive loss (count of losses) | -412938.23 (1) | |
| Average | consecutive wins | 8 | consecutive losses | 1 | |
-
-
#2 written by birt April 16, 2012 (1 year ago)
The whole thing with “selling versus keeping for their own and making a ridiculous amount of money” has been discussed several times. Generally, if you have a good system on your hands you’re better off selling it even if it is just to fund your trading account. That is, unless you have hundreds of thousands in trading capital already.
Unfortunately, I lack the time to test the learning feature and will leave that to those who want to experiment with it.
-
-
#3 written by socalengineer April 16, 2012 (1 year ago)
-
#4 written by birt April 16, 2012 (1 year ago)
-
#5 written by socalengineer April 16, 2012 (1 year ago)
Yes, but in theory if you fed it tick data from 2007-2012 it should learn the previous years and provide even better results on the backtest than is depicted here. Because the more experience it has the better it gets, just like a human.
But I understand if you don’t have time.
Also I would go ahead and test myself but the refund policy stated on eareview is no where to be seen on their website. I contacted the vendor regarding their policy, and no comment was made but they have responded to my other questions.
-
#6 written by birt April 16, 2012 (1 year ago)
It’s sold via Click2Sell and all vendors selling stuff via this platform are forced to use the Click2Sell refund policy, see the Terms of Purchase for more info. Actually, I seem to have been wrong: it’s 90 days, not 60 days, have to update it above.
-
#7 written by socalengineer April 16, 2012 (1 year ago)
-
#8 written by birt April 16, 2012 (1 year ago)
-
#9 written by socalengineer April 19, 2012 (1 year ago)
Just for everyone’s records. I contacted click2sell:
“Click2Sell is an assistant merchant system used by the seller of your purchased product and we do not process payments ourselves, so we are not able to make refunds.
If you will buy their product and you won’t like it you will need to send a message to the vendor requesting a refund. In most cases, if a buyer is reasonable, vendors refund money without any problem.
In case you do not get any response from the seller, please log in to your PayPal account, find the payment to the seller and in its details click a button/link to raise a claim. It is an official way to get a refund from the seller.”
But the EA Vendor for Neotradex has made a statement below regarding their policy…
-
-
-
-
-
-
-
#10 written by socalengineer April 16, 2012 (1 year ago)
Also a few links for others to other live accounts (demo although):
http://www.mt4i.com/users/neotradex
http://www.forexpeacearmy.com/metatrader_expert_advisor/neo_trad_ex/demo -
#11 written by scls19fr April 16, 2012 (1 year ago)
Hello,
a good way to avoid the learning bias you are talking is to split your backtest data in two parts.
1) The learning data set : you use this data set to train your EA (and only to train it)
2) The testing data set : you use this data set to test your EA (after setting it correctly with 1) )
This advise is given by Ernest P. Chan in his book Quantitative Trading (Wiley Trading).
This problem is called “data-snooping bias”.
But is there a way to clear knowledges ?
Regards
-
#12 written by birt April 16, 2012 (1 year ago)
I agree and it applies pretty much to all EAs. When it’s not a neural network EA, instead of a learning data set it’s an optimization data set. To be honest, I am a little skeptical when it comes to neural network EAs and I was reluctant to add this one to my website, fearing that it was overoptimized on the available backtesting data and that it will fail when trading live. That’s why I’ve waited until I’ve seen that it can also perform nicely in forward testing.
-
-
#14 written by Joao Santos April 17, 2012 (1 year ago)
It is odd that backtest http://eareview.net/wp-content/uploads/2012/04/NeoTradEx-1.4-EURUSD-H1-2007-2012-tick_data-real_spread-defaults-fixed_on.htm had trades exited for low loss and low profit. But on forward test we see higher exit and already a full loss of -108 pips, whereas on backtest from 07/2010 highest loss was only -23 pips.
-
#15 written by birt April 17, 2012 (1 year ago)
-
-
#16 written by Alexander April 17, 2012 (1 year ago)
Hello, but this account here http://cdn8.eareview.net/wp-content/uploads/2012/04/NeoTradEx-1.4-EURUSD-H1-2007-2012-tick_data-real_spread-defaults-fixed_on.htm seems to be a miniforex account,
but we use classic forex account for live tests. That’s the reason for this “discrepancy”Stop loss is around 100 pips, it’s calculated using the current volatility,
so it’s not constant. -
#17 written by onur April 18, 2012 (1 year ago)
regards refund, take a close look at the policy and good luck with that…
Refund Policy
Refunds will be granted to you, the purchaser, under the following conditions ONLY:
You must contact us, within 60 calendar days (“the refund period”) after date of purchase, with the following:
•A written request (by email) for a refund
•Attached proof that you have used, for a minimum period of at least 40 (forty) trading days, the Expert Advisor NeoTradeEx you purchased via the NeoTradEx.com website.
Such proof must be demonstrated within 60 calendar days after date of purchase, to our satisfaction, as follows:
•In the form of your Live account or Demonstration account statement of your broker, AND
•The trades as reflected on the statement must, when examined as against the trading charts (for the same periods) and the similar trades or simulations on similar platforms, show to our satisfaction, that the majority of trades were in fact made using the Expert Advisor you purchased, AND
•The Broker statement shows a loss after a minimum of 40 trading days, AND
•The Expert Advisor NeoTradeEx suffered no interference from another expert advisor, or the user AND
•The Expert Advisor NeoTradeEx was run with default settings, without any modifications or alterations, AND
•The Expert Advisor NeoTradeEx (or update) was declared live and ready to use on a live/real money account, AND
•All logs from the terminal and the Expert Advisor folder for the period are provided to NeoTradEx.com .
You must send a refund request to our support desk and provide all required information as mention above.
Please be advised that when you sent a refund request, your product activation key will be immediately deactivated and you will no longer be able to use your copy of our Expert Advisor.
Should you have elected to have bought this product from us at a “discounted price”, then the discounted price shall be the maximum claimable for refund. Should you purchase at the full purchase price, such price shall be the maximum claimable for refund.
At any and all times the maximum claim that may be made by you against ourselves whether in a single or class action or multiple claim action, including all costs and sundries, shall be limited to the amount of €199.00.
We wish to protect our assets and the interest of our paying customers. If we detect customers are decompiling/modifying/cracking/selling signals/or performing any other acts related copyright infringement, this agreement will be canceled immediately and rendered null and void and refund requests shall not be accepted or honored.
-
#19 written by Alexander April 18, 2012 (1 year ago)
-
#21 written by Alexander April 18, 2012 (1 year ago)
To avoid any misunderstanding we updated our refund policy, please check it here http://neotradex.com/index.php?option=com_content&view=article&id=10&Itemid=9&lang=en
Basically we do our best to make our software work out for you, if we fail to do it you receive a refund.
Alexander,
Neotradex.com -
#22 written by Johnny Honda April 18, 2012 (1 year ago)
-
#23 written by birt April 18, 2012 (1 year ago)
As posted above your comment by the vendor, the refund policy has been changed to something much more friendly.
Anyway, using “edu” EAs is a very bad idea in most cases – I’ve seen tons of failed “edu” attempts that would leave EAs in a state that would not even resemble the licensed EAs, not to mention the fact that you might as well forget about updates and that you run the risk of ending up with a trojan. If you wish to discuss the “edu” topic further, I would kindly ask you to take it somewhere else (meaning off eareview.net) because it does not belong here.
-
-
#24 written by RG April 18, 2012 (1 year ago)
-
#25 written by Alexander April 18, 2012 (1 year ago)
-
#26 written by socalengineer April 19, 2012 (1 year ago)
Hi Alexander,
It would be nice to have your presence on the other forum regarding your EA to answer questions:
http://www.donnaforex.com/forum/index.php?topic=5672.0-
#27 written by socalengineer April 19, 2012 (1 year ago)
-
-
#30 written by RG April 19, 2012 (1 year ago)
Thanks for your reply. if there is active order that coincides with news release, and suddenly 50 pips spike in a wrong direction, I was thinking it would hit the SL. would be nice to use “news feed filter”… anyway.. if I were to run the EA, and stop it before news (click the EA button with the hat), and then restart the EA after the news, would it mess up the NN learning? Thanks.
-
-
-
#32 written by socalengineer May 14, 2012 (1 year ago)
-
#34 written by Joao Santos May 31, 2012 (11 months ago)
I don’t mean to draw criticism for posting an external link, but rather to elicit genuine feedback on the subject of neural network… What are your views on the backtest and presentation of this neural by Iticsoftware?http://blog.iticsoftware.com/2012/05/24/forex-neural-net-filter.aspx?results=1#SurveyResultsChart
It can be integrated with an EA if source code is provided. I wrote vendor and he said the price is $800 (can use on two computers).
-
#37 written by Joao Santos June 28, 2012 (10 months ago)
-
#38 written by birt June 28, 2012 (10 months ago)
Yeah, we’re in the same boat when it comes to that. It’s been doing rather bad on my account ever since I started it.
I’ve ran into quite a few other neural network EAs before and none of them were profitable in the long run… Looks like this might be no exception. I even waited a while to give it time to get some live results and confirm that it can do good live but as it turns out I should’ve probably waited a couple more months.
Anyway, for some reason, it seems to be trading very different on my account versus the vendor’s account, although lately it didn’t do very good on their account either… I’ll let the forward test run for the time being, might be just a bad period – who knows what the future has in store for it.
-
-
#39 written by alexander June 29, 2012 (10 months ago)
Hello,
we are updating our software at the moment, taking into account differences between brokers’ data. We cannot say that the first version of our software was a total sucess, but most of our customers are satified with the performance, we had only one refund.But certain things will have to be taken into account, unfortunately we couldn’t foresee that before launching the retail version of the software.
Very soon we’ll offer the updated version. -
#40 written by bismillahwd July 19, 2012 (10 months ago)
birt could you suggest me by your own experience for best trend following EA? or at least profitable EA with Stoploss.
i see everywhere on your backtest it was profitable, but on real forward test your account got minus by most of trend follower EA
please suggest.
confuse here watching nice backtest but headache on forward test
anyway your the best, no one ever do this kind heart test for published.
it just you in this world man !!!regards
-
#41 written by birt July 19, 2012 (10 months ago)
Thanks for the appreciative comments. I would suggest taking a look at Forex Growth Bot – although my account seems to be completely haywire, most people report matching the performance of the official account and that one is doing quite good.
-
#42 written by bismillahwd July 19, 2012 (10 months ago)
-
- Comment Feed for this Post
Very very very interesting…. Did you try out the learn feature at all to see if the neural network got better after running dukascopy data through it?
You would think after you ran a few more years data through the network it would change the behavior. For some reason I have my doubts on this because although it is possible, it is a very complex system for someone to implement and for them to sell it after they have implemented seems kind of ridiculous when it can make them insane amounts of money. Don’t you think?
That is why I was hoping you would test the learn feature. Try feeding into it USDJPY and label it as EURUSD to the bot. then run it on EURUSD, it should perform horribly. I want to see if this thing is really “learning”