コメント | 自動売買ロボット(エキスパートアドバイザー)

Andrew Lee 2025.03.17 21:16 #21

Thank you for the extremely fast update Inochi and for adding GMT and fixed lots. Really appreciate it

Andrew Lee 2025.03.18 07:12 #22

Is anyone else getting these messages in the Experts tab?

I have multiple currencies using different GPT models, and each window has its own unique magic number.

Unsure of the cause, and whether this would impact trading in any way. May have to wait a couple of days before Inochi can comment.

Andrew Lee 2025.03.18 07:37 #23

Thanks for the fast fix on the above Inochi - amazing support as always!

Andrew Lee 2025.03.18 08:26 #24

Hi Inochi,

It seems the following models are returning a 404 error. Could you please check if the model names are correctly configured in the drop down code?

Not working:

Model_o3-mini
o3-mini-2025-01-31

Model_o1
o1-2024-12-17

Working:

01_mini_2024_09_12 (Default)
o1-mini-2024-09-12

Model_gpt_4_turbo
gpt-4-turbo-2024-04-09

I haven't tested other models - thank you for your help.

LUIS ALBERTO BIANUCCI 2025.03.18 09:44 #25

I think this is the model that's performing best in Forex so far: GPT-3.5-Turbo. I'm using it in another EA, and it's improving its performance. Inochi, it would be important if you could publish the *.sets for each pair in detail, along with the appropriate timeframe.

Akinlolu Afolabi 2025.03.18 10:12 #26

LUIS ALBERTO BIANUCCI #:
GPT-3.5-Turb

how is the model going in terms of result.can you please share your observations so far with the GPT 3.5 TURBO

Lucas Lima Cavalcante 2025.03.18 12:39 #27

Is it possible to share how is ChatGPT being prompted to assist us? What differs us from the rest of so called AI EAs?

Lucas Lima Cavalcante 2025.03.18 18:35 #28

Akinlolu Afolabi #:
how is the model going in terms of result.can you please share your observations so far with the GPT 3.5 TURBO

I had best results with GPT 4.5 Turbo, however costs increased fast. Just today it cost me over $6, while at o1 mini whole week last week cost me near $1.81.

Accuracy is good, but if you dont set cooldown, EA will open new orders after last one closes, even at prices over last TP and not looking for discounted entries.

Akinlolu Afolabi 2025.03.18 19:53 #29

Thank you i did some backtest today and the Gold 5 minutes has a 64 percent winrate which is great for 5 minutes scalping the EU and GU i need to back test it more.i used the setting he gave.

Andrew Lee 2025.03.18 23:01 #30

I'm using GPT_4_turbo and had some good results with GPDUSD and EURUSD on H1 timeframe.

It was more accurate than 01_mini but a little more expensive with the API calls as Lucas mentioned, but given it only checks trades every hour on H1, it didn't cost too much during the NY and London sessions.

Trading with GPT_4_turbo on M5 timeframe would add up API costs quickly with 12 calls per hour, so be careful what model you use with lower timeframes :)

LUIS ALBERTO BIANUCCI 2025.03.18 23:06 #31

My statements are based on costs and profitability.

Lucas Lima Cavalcante 2025.03.18 23:53 #32

we could ask the owner to make a telegram chat group

Yuta Hidaka 2025.03.19 01:46 #33

Hi, can I use another LLM models?

Like Gemini, Claud, etc...

Akinlolu Afolabi 2025.03.19 03:42 #34

Andrew Lee #:

I'm using GPT_4_turbo and had some good results with GPDUSD and EURUSD on H1 timeframe.

It was more accurate than 01_mini but a little more expensive with the API calls as Lucas mentioned, but given it only checks trades every hour on H1, it didn't cost too much during the NY and London sessions.

Trading with GPT_4_turbo on M5 timeframe would add up API costs quickly with 12 calls per hour, so be careful what model you use with lower timeframes :) cool i will try that today.

cool i will try that today.

Akinlolu Afolabi 2025.03.19 03:44 #35

Yuta Hidaka #:

Hi, can I use another LLM models?

Like Gemini, Claud, etc...

Are they cheaper? and how much do these cost.

Akinlolu Afolabi 2025.03.19 03:48 #36

Lucas Lima Cavalcante #:
we could ask the owner to make a telegram chat group

This is a great idea.

trinacria_it 2025.03.19 05:09 #37

Andrew Lee #:

I'm using GPT_4_turbo and had some good results with GPDUSD and EURUSD on H1 timeframe.

It was more accurate than 01_mini but a little more expensive with the API calls as Lucas mentioned, but given it only checks trades every hour on H1, it didn't cost too much during the NY and London sessions.

Trading with GPT_4_turbo on M5 timeframe would add up API costs quickly with 12 calls per hour, so be careful what model you use with lower timeframes :)

The timeframe to set to 1h is this?

thanks

ファイル:

ss_2.png 18 kb

trinacria_it 2025.03.19 05:26 #38

Yesterday I used GPT-4-turbo timeframe 1H, but it did not open any operation, is normal?

what is it the GMT offset broker for activtradeEU? at now is 0 and is right? Thank you

trinacria_it 2025.03.19 05:31 #39

Lucas Lima Cavalcante #:
we could ask the owner to make a telegram chat group

Is a good idea

Andrew Lee 2025.03.19 08:13 #40

trinacria_it #:

The timeframe to set to 1h is this?

thanks

That’s correct. I attach to H1 chart and select analysis timeframe of 1 hour. This will contact the chatGPT api server on the hour.

Follow the settings Inochi has provided on the overview page for more info on how to setup the recommended pairs.

Check what server time your broker uses. You can check yourself by searching for UTC time now in google, then figure out the amount of hours difference between the two times. This will be the offset time. It’ll probably be 2 or 3 with Activetrades depending on their summer time schedule.

Good luck!