Statistically testing every indicator with my Genetic Algoritm | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

codification of the market / Genetic Algorithms 36 replies

S/R Area Algoritm 17 replies

What Constitutes a Statistically Significant Sample for Back Testing? 4 replies

Has anyone had luck with high frequency using either neural nets or genetic Pr? 3 replies

Genetic Algorithm for Metatrader 1 reply

Attachments: Statistically testing every indicator with my Genetic Algoritm
Exit Attachments

Statistically testing every indicator with my Genetic Algoritm

  • #41
  • Jun 8, 2014 1:13am Jun 8, 2014 1:13am
Quoting Proximus
{quote} It has 3 parameters, its not good.
You mean it is not good to test or it is not a good indicator? If the later, is there any data to support this?
If you are good at something, never do it for free--Joker
  • #42
  • Jun 8, 2014 2:33am Jun 8, 2014 2:33am
Quoting jmn5611
{quote} You mean it is not good to test or it is not a good indicator? If the later, is there any data to support this?
Neither, it would simply take too long.

Quoting Proximus
More than 2 parameters take ages to compute, if 1X1 has 200x200 values, which is 40.000 combinations are usually calculated in less than 5 minutes, then a 3rd parameter would have 3 dimensions 200x200x200, which is 8 million combination which would take 16 hours to calculate, and i wont do that
"There's a sucker born every minute" - P.T. Barnum
  • #43
  • Jun 8, 2014 5:42am Jun 8, 2014 5:42am
Quoting Proximus
(NOT)Awesome Oscillator (Buy/Sell above/below 0) Pair: EURUSD Timeframe: M30 Years tested: 12 Net pipettes acumulated (win-lose): -102321 Total trades: 6075 Accuracy rate: 33.33(3)% Minimum parameters/price cluster: It has no parameters Profitable: Nope Comments: Total opposite to its name, its not awesome at all, its a bullsh*t instead
Try this, it has settings

When considering Stoch, can't you pick 2 parameters and leave the third one out of the optimization? Like slowing, for instance.

Best of luck!
Attached File(s)
File Type: mq4 AOwInputs.mq4   3 KB | 508 downloads
  • #44
  • Jun 8, 2014 7:10am Jun 8, 2014 7:10am
Quoting candero
{quote} Try this, it has settings When considering Stoch, can't you pick 2 parameters and leave the third one out of the optimization? Like slowing, for instance. Best of luck! {file}
Can't, with the MACD i used the first 2 parameters and ignored the signal period of the MACD, because we only need 2 parameters for the histogram, however the stochastic uses all 3 of them for each line, so i cant just fix 1 and use the other 2, that would not be dynamic.

Ok i will test your indicator.Thanks for uploading.

Other guys can upload too, all indicators that you have and curious about!
"There's a sucker born every minute" - P.T. Barnum
  • #45
  • Jun 8, 2014 7:32am Jun 8, 2014 7:32am
Hi,
a very interesting thread. Am curious if its possible to test the TDI - Traders Dynamic Index?

cheers
Attached File(s)
File Type: ex4 Synergy_TDI.ex4   22 KB | 389 downloads
  • #46
  • Jun 8, 2014 7:41am Jun 8, 2014 7:41am
Quoting OuterWest
Hi, a very interesting thread. Am curious if its possible to test the TDI - Traders Dynamic Index? cheers {file}
Sorry no ex4 files.Only .mq4
"There's a sucker born every minute" - P.T. Barnum
  • #47
  • Jun 8, 2014 7:56am Jun 8, 2014 7:56am
Stochastic 7,2,2 smma, close/close
If you are good at something, never do it for free--Joker
  • #48
  • Jun 8, 2014 7:59am Jun 8, 2014 7:59am
Quoting candero
{quote} Try this, it has settings When considering Stoch, can't you pick 2 parameters and leave the third one out of the optimization? Like slowing, for instance. Best of luck! {file}
Name: AOwInputs.mq4
Net pipettes acumulated (win-lose): -9201
Total trades: 2638
Accuracy rate: 47.49
Minimum parameters/price cluster: None good
Profitable: No
Comments: Althought i only tested a 100x 100 version, it looks pretty lame, other indicators had many good outcomes in the 100x100, so probably no point testing it further
"There's a sucker born every minute" - P.T. Barnum
  • #49
  • Jun 8, 2014 8:00am Jun 8, 2014 8:00am
Quoting jmn5611
Stochastic 7,2,2 smma, close/close
Whats with that?
"There's a sucker born every minute" - P.T. Barnum
  • #50
  • Jun 8, 2014 9:29am Jun 8, 2014 9:29am
Hey very cool setup, I have an indicator for you!! I hope this indicator meets your requirements to test, it is a Volume zone oscillator, based on the OBV with a parameter that allows for time to be screened, unlike the OBV.
Not sure if your able to test this ? The Mq4 file I have for this indicator, but doesnt give signals, I had one created to give signals but I only have the EX4 file.

I would love to get the VZO tested, if I cant with the current file attached, let me know what I can do, to get it test ?
Attached File(s)
File Type: mq4 VZO.MQ4   3 KB | 743 downloads
Learn, a forex trader must, unlearn and relearn he will.
  • #51
  • Edited 12:49pm Jun 8, 2014 12:30pm | Edited 12:49pm
Quoting dkrock
{quote} What moving average indicator do you have that allows for decimal point settings? This information is fairly accurate. In the old days I used SMA/LWMA of the same period setting and it was pretty good.
okat my fastest was 8ma, so i moved the next to 10 ma, which is pretty close to 1.2. I change the background color on the window of the pair according to the condition of these 2 moving avearges and a 96ma. (the slow only plays a role in that it is telling be to close a trade in loss if it gets down there.

I set a net of pending trades beyond the nearest S/R line, with the reasoning that if price gets there, the ma.s are aligned( I draw them so that this must be so.

I have pages of rules and explanations for them but one thing tends to throw them out the window. If the 8 bar ma is moving at a good angle and price is moving even better, I usually take the plunge. many of my better trades have happened because of this: I am wondering What sort details our friend is willing to share about the single M.A. system. since I may actually be conforming more to that one than any other he describes. Or that I describe myself as doing...

but take a look at the 8bar m.a. and the ten bar m.a. see how closely their crossings line up with the candles being above the 8bar M.A. or below them!
  • #52
  • Jun 8, 2014 12:44pm Jun 8, 2014 12:44pm
Okay proximus, If I use the 8 bar m.a. with price cross over begin the entry signal , and price recrossing to the other side of it being the close signal ( profit or loss, does this seem reasonably profitable?


It sounds that some indicators are income generators with positive expectancies.
It sounds as though if you combine one of these with good anti-martingale monay managment
you should in short order be making an enjoyable living.
  • #53
  • Jun 8, 2014 4:33pm Jun 8, 2014 4:33pm
Quoting Proximus
{quote} Whats with that?
Your genetic algorithm will, or should ask you for the parameter of the stochastic, that is what those are.
If you are good at something, never do it for free--Joker
  • #54
  • Jun 8, 2014 8:15pm Jun 8, 2014 8:15pm
Quoting COGSx86
Hey very cool setup, I have an indicator for you!! I hope this indicator meets your requirements to test, it is a Volume zone oscillator, based on the OBV with a parameter that allows for time to be screened, unlike the OBV. Not sure if your able to test this ? The Mq4 file I have for this indicator, but doesnt give signals, I had one created to give signals but I only have the EX4 file. I would love to get the VZO tested, if I cant with the current file attached, let me know what I can do, to get it test ? {file}
Oh sh*t sorry man, my historical data doesnt contain Volume information, all i have is O,H,L,C to be compressed and quick for the GA.
"There's a sucker born every minute" - P.T. Barnum
  • #55
  • Jun 8, 2014 8:19pm Jun 8, 2014 8:19pm
Quoting Rap Skallion
Okay proximus, If I use the 8 bar m.a. with price cross over begin the entry signal , and price recrossing to the other side of it being the close signal ( profit or loss, does this seem reasonably profitable? It sounds that some indicators are income generators with positive expectancies. It sounds as though if you combine one of these with good anti-martingale monay managment you should in short order be making an enjoyable living.
That is how i tested them, at buy signal open buy trade and/or close sell trade, at sell signal open sell trade and/pr close buy trade.

But i didnt put a TP and SL, i dont believe in them, so the intra trade DD could be high, especially on those indicators where the parameters are big.

For example if you would use this strategy on simple MA 300 crossing price, then a trade could last 2-3 months, if its a winner, then you lose swaps on EUR/USD, if its a loser, then its a psychological terror.

So that is why i emphasized to find the indicator where with the smallest parameter to make the best profit, so that we would be in and out of the trade in maximum 2 days.

So the PSAR got the 1st place with 3604 trades over 12 years thats about 0.82 trade /day, where in average a trade lasts 1-3 days or so, which fits our goals.

Because the price cant move so much against you in 3 days to wipe you out if you size your position correctly 1-2% risk or account.

Quote
It sounds as though if you combine one of these with good anti-martingale monay managment

No its not anti martingale, it is just scaling the order to be always 1 or 2% of the account balance, so if today i risk 1$ for 100$ account and tomorrow 100$ for 10.000$ account, its the same its not martingale in any shape or form.
"There's a sucker born every minute" - P.T. Barnum
  • #56
  • Jun 8, 2014 9:22pm Jun 8, 2014 9:22pm
Thank you: This has been very informative to me.
  • #57
  • Jun 8, 2014 10:46pm Jun 8, 2014 10:46pm
Quoting Proximus
{quote} Sorry no ex4 files.Only .mq4
Here is TDI.mq4.

Will your algo test Heiken Ashi?
Attached File(s)
File Type: mq4 TDI.mq4   9 KB | 493 downloads
  • #58
  • Jun 8, 2014 11:08pm Jun 8, 2014 11:08pm
Quoting jmn5611
Stochastic 7,2,2 smma, close/close
OK i tested it, but this will be the last time when i test a single parameter set, so please dont ask more, because you must understand that 1 parameter cant hold forever, it's all about adaptation!Thats the primary function of a genetic algorithm, it selects a parameter then as the price evolves, the algorithm learns the best parameter in parallel with the price, so it will give out the best parameter for each season, but it will rarely be the same, as the price never repeats itself identically.

So you must understand that i`m not testing the whatever parameter of the indicator, but the basic formula of the indicator and the adaptability of the indicator, and the predicting ability of it.The more similar the indicator is with the price flow, the better it can predict price.Its logical

The results with Stochastic 7,2,2 ,smoothed ma, close price , M30 ,EUR/USD is:

-251351 pipettes lost
12305 trades made
and 11.568% accuracy
"There's a sucker born every minute" - P.T. Barnum
  • #59
  • Jun 8, 2014 11:39pm Jun 8, 2014 11:39pm
Are you testing Pinhead's TDI.mq4?

PS. I've got spare CP cycles on my server.
  • #60
  • Jun 8, 2014 11:57pm Jun 8, 2014 11:57pm
Quoting nondisclosure00
Are you testing Pinhead's TDI.mq4? PS. I've got spare CP cycles on my server.
Not now, i`ve put some trades on, i will test it in an hour.
"There's a sucker born every minute" - P.T. Barnum
0 traders viewing now