Thread: Statistically testing every indicator with my Genetic Algoritm

Heh, never really had a problem with quick profits Also would be useful if all outcomes are set in first post after completing

EDIT: Just a thought, would you put too average monthly profits along with maximum monthly and cheapest...

Originally Posted by ;

Hehnever had a issue with quick profits Also would be good if all results are put in first article after completing

Sure I will record them and put links to every place where I detailed them

6) DeMarker (Above 0.3 BUY / Beneath 0.7 SELL)

Pair: EURUSD
Timeframe: M30
Years analyzed: 12
Net pipettes acumulated (win-lose): 48139
Complete transactions: 93
Accuracy rate: 49.462 percent
Minimum parameters/price cluster: Above 100 atleast
Profitable: Yes
Remarks: So far the best indior, defeats the MACD also


7) Bollinger Bands, Force Index, ATR,Momentum, and Envelopes will be skipped. The BB and Envelopes will be skipped because the price isn't normally distributed, therefore taking a look at the standard deviation of them is pointless, and likely wont be accurate.

Others are not signal indiors, they measure other stuff.

8) Ichimoku (Kijun Tenkan Cross)

Pair: EURUSD
Timeframe: M30
Years tested: 12
Web pipettes acumulated (win-lose): Factory 18404
Total trades: 1093
Accuracy rate: 37.45percent
Minimum parameters/price cluster: Averaging around defaults, looks like the ichimoku's defaults are fairly great
Profitable: Yes
Comments: Looks just like the default configurations may be acceptable, although the profit is not much, but atleast its accurate.

9) Ichimoku (Kumo Cloud egy)

Skipped, because it requires 3 parameters.

Additionally we'll bypass the OSMA also, its like MACD, but it's also 3 parameters accordingly, theres a lot of to compute with.

Addition: Honestly I'm very dissapointed about its performance I thought the ichimoku would perform far better, I mean each of the 300 decades of japanese rice trader tradition and intellect is worth almost nothing nowadays is sad. It works better in rice futures, but sadly the ichimoku is probably an overhyped myth in forex.

Originally Posted by ;

EDIT: Just a thought, would you put also typical annual profits and highest monthly and cheapest...

Sorry that the AI doesnt work like this, it segments the whole historical data into arbitrary packets, or clusters, also calculates the best parameters, then chooses the very best and moves , and it gets better and better with much more data, just as a little one gets smarter by studying more, its a artificial inteligence using a heuristic algoritm to find out.

I have no method of knowing the intra-week or intra-month profit, I understand the overall stats at the conclusion

that's why its best to have as many trades as possible not to have an entire month in without, that would s*ck very poorly.

That's why I say that 189 is the best setting of the RSI, since it isnt, in Various times different configurations are great, but I can quantify the minimal setting, therefore its plausible that if I say anything below 100 is awful, is since the algoritm never picked any value below 100 so they must all be shedding parameters, but that still doesnt imply That Each parameter over 100 is great, You Need to optimize, but you atleast have a benchmark

10) RVI =Relative Vigor Index (Above 30 BUY / Below 70 SELL)

Pair: EURUSD
Timeframe: M30
Years tested: 12
Web pipettes acumulated (win-lose): Factory 42040
Total trades: 1529
Accuracy rate: 34.401percent
Minimum parameters/price bunch: Above 150 to BE
Profitable: Yes
Comments: A bit bad at filtering, it requires high settings but the second greatest indior still after DeMarker

11) WPR = Willi Percent Range (Above -80 buy / below -20 sell)

Pair: EURUSD
Timeframe: M30
Years tested: 12
Web pipettes acumulated (win-lose): -33321
Complete trades: 3351
Accuracy rate: 57.20%
Minimum parameters/price bunch: None,atleast not for this particular TF
Profitable: Nope
Comments: The first crappy indior up to now, it was designed for greater timeframes or it doesnt work at all.A bit surprising because its accuracy is very high, yet it didnt make profit, whereas the rest of the indiors up to now, no matter how bad but made profit, this 1 didnt.

It's supported, on AUDUSD it created -69862, its not working, sorry.Chaos theory egy might be obsolete aswell.

11) MFI = Cash Flow index, dont confuse it with market facilitation index (Above 20 buy / under 80 Market )

Pair: EURUSD
Timeframe: M30
Years tested: 12
Web pipettes acumulated (win-lose): 47802
Total Transactions: 46
Truth rate: 42.3%
Minimum parameters/price Bunch: Close to default
Profitable: Yes
Remarks: At first glance it looks like it made to 2nd place after the Demarker, but it didnt, the sample of Commerce is too small and probably unreliable too long trades with huge DD.So 2nd place is still reserved for RVI

12) Accelerator Oscillator (Buy/Sell above/below 0)

Pair: EURUSD
Timeframe: M30
Years tested: 12
Web pipettes acumulated (win-lose): -352618
Complete Transactions: 16134
Truth rate: 38.94%
Minimum parameters/price Bunch: It's no parameters
Profitable: Nope
Remarks: Total disaster and waste of time, since it has no parameters, you cant adjust it to market price, and the initial formula itself Only doenst follow the price, its Quite unprofitable indior

13) (NOT)Awesome Oscillator (Buy/Sell above/below 0)

Pair: EURUSD
Timeframe: M30
Years tested: 12
Web pipettes acumulated (win-lose): -102321
Complete Transactions: 6075
Accuracy rate: 33.33(3)%
Minimum parameters/price Bunch: It's no parameters
Profitable: Nope
Remarks: Total Reverse to its name, its not Amazing in Any Way, its a bullsh*t Rather

Let's recap a little bit.

We completed all groups of oscillators and volume indiors, only the trending group abandoned, and of course the habit indiors.

We jumped a couple of indiors here and there, by way of instance that the Gator which had 6 parameters OMG, although with 3 parameters that the total permutations will be like 500^3 since my maximum parameter cap is 500.

It is reasonable because if an indior needs over 500 then it was likely not designed for M30 interval, so why bother calculate further.

So with 3 parameters my 8 core chip PC has to calculate 125000000 combinations of 12 years of M30 data, like 200.000 candlesticks the individual chip work that an indior needs for each calculation the CPU that the AI needs to learn and unite the clusters.

Currently with two parameters usually a calculation takes 5-15 minutes, nevertheless with 3 parameters it takes 8 hours .

With 6 parameters it could take 300 years

Hence the Gator Oscillator is out of question

The CPU requirement develops quicker than exponential should you add more parameters...

I wish I had a PC like this for all these tasks...



Along with the Fractals indior just laggs like hell, I have a couple of custom ones, but just not match for AI trading, but I never understand why use them if they lagg 2-3 bars.

Wow. Thanks for the outcome. Must be dull to work but would encourage your work. Thanks

Proximus, superb work And when you say such as for DeMarker or RSI there Minimum parameters/price bunch: Above 100 at least. . What exactly does that mean just a setting for Period of at least 100 candles?

Great work, thanks! I use 9-14 parameter which stands in timeframe for 72-112 and trade on 4h interval.

Can you check Commodity Channel Index I wonder how will it perform