Data mining with Asirikuy's OpenKantu to discover trading strategies

Over the course of the last three years I've been studying trading as a method of portfolio diversification. One promising looking option for strategy discovery seems to be data mining, and I've had a look at OpenKantu from the Asirikuy stable.

OpenKantu is billed as the "first free and open source price action based trading system generator".  All of which sounds very promising.  The author, Daniel Fernandez sounds like he knows what he's doing.  OpenKantu has been superseded by pKantu, which forms part of the Asirikuy paid-for service.  Daniel has released OpenKantu for general use, and because he signposts the user towards Asirikuy, I think it reasonable to assume he's using OpenKantu as a "try-before-buy" system.  Indeed, he specifically states that there are no trials available for the Asirikuy service.

With that all in mind, I embarked on a trial of OpenKantu.

My impressions are:

• it's easy to use
• it's fast
• the documentation is OK
• it can produce systems

But:

• the MetaTrader4 (MT4) ExpertAdviser (EA) included with OpenKantu (RecordBars_MT4.mq4) to export data for testing direct from MT4 produced a vast multi-gigabyte file of the 4 hour data I fed it, creating huge numbers of multiple entries
• it's no trivial matter to import data and then correctly run tests outside of the RecordBars_MT4 EA, the manual is short on information, and Daniel politely made it clear he would not answer my query surrounding data digit depth (0.0001 is 4 digit, 0.00001 is 5), and how to correctly set OpenKantu up to deal with this
• I could not get any strategies that it produced to replicate on MT4

Here's what I did:

• Read the OpenKantu manual
• Found a source of 4 digit data from Forextester.com
• Using CSVed, edited the data to import into MT4
• Created a .csv of the data, 4 hour timeframe
• Edited this with CSVed, and following the manual's instructions, imported this into OpenKantu
• Ran a basic simulation in OpenKantu
• Used OpenKantu's "export to MQL4" function to create an EA
• Ran this in MT4

Remember, I'd used the same data both in OpenKantu and MT4, same date range, etc., so should have had (almost) identical results.

But no.  Even a cursory glance at the following balance graphs shows there are no similarities.  Even the number of trades do not match.

The following chart and table is the output of the results generated in OpenKantu for a trial strategy, from which the MT4 EA was generated: 

Balance - OpenKantu - from which eurusd EA was generated

P	No.	Symbol	Abs. Profit	Profit/trade	Profit Longs	Profit Shorts	No. Longs	No. Shorts	Total Trades
1	66	EURUSD240	28160	461.63	891	27269	30	31	61
Max DD	Ideal R	R^2	Ulcer Index	Max DD Length	Cons.Loss	Cons.Wins	Profit:Max DD	Win %	Reward:Risk
3.29	0	0.939	1.25	129	4	9	7.94	59.02	0.69
Skewness	Kurtosis	PF	Std.Dev	Std.Dev Breach	Total ME	SQN	Mod. Sharpe Ratio	Custom Criteria	Bars out
8.4	1.79	2.61	1534.56	2.35	17.9509	2.35	0.3	0	0
OSP	OSP/trade	OS_ProfitLongs	OS_ProfitShorts	OS_TotalLongs	OS_TotalShorts	OS_TotalTrades	OS_MaxDD	OS_UlcerIndex	OS_MaxDDLength
0	0
OS_Cons.Loss	OS_Cons.Wins	OS_ProfitToMaxDD	OS_WinningPercentage	OS_RewardToRisk	OS_Skewness	OS_Kurtosis	OS_PF	OS_StdDev	OS_StdDevBreach
OS_TotalME	OS_R2	OS_SQN	OS_ModifiedSharpeRatio	OS_CustomCriteria	OS_BarsOut	Lowest Lag

The following chart is the output from MT4, using the same data and strategy from OpenKantu:

Balance and Report - MT4 using eurusd EA

So, what do I conclude?  Not a lot, really.  OpenKantu, and therefore I assume (but can't tell) by association pKantu and the Asirikuy system works and seems to produce results.

But in order to test this, I'd have to buy into the product, taking a leap of faith - that's $300 worth of leap that I'll get a return.  I'm not ready to take that leap yet.