Holy f do never use such leverage if you want to survive. I see 10x 15x even 20x! You will be wiped out in a fast market, like it happened last week.
JohnnyDoe said:Holy f do never use such leverage if you want to survive. I see 10x 15x even 20x! You will be wiped out in a fast market, like it happened last week.
Yes. It's for experiment purposes not real investments of course đ
So far the 6 AIs are losing a cumulative 0.7%, which is about the expected result for a systematic intraday strategy that is essentially random but very disciplined. A human is expected to lose everything in 3 months.
DeepSeek is usually better at numbers than the others (in particular stupid GPT).
Anyway I would run such strategies only on CME futures, where the risk of book manipulation is lower. Which means just btc, eth, sol and xrp.
JohnnyDoe said:So far the 6 AIs are losing a cumulative 0.7%, which is about the expected result for a systematic intraday strategy that is essentially random but very disciplined. A human is expected to lose everything in 3 months.
DeepSeek is usually better at numbers than the others (in particular stupid GPT).
Anyway I would run such strategies only on CME futures, where the risk of book manipulation is lower. Which means just btc, eth, sol and xrp.
Emotions the worst enemy.
where are the initial user prompts?
JohnnyDoe said:where are the initial user prompts?
1. Click models
2. Pick one like deepseek 3.1
3. User prompt based on current time, so move to the old date at the bottom
It doesn't show the initial prompt and the full chat history
JohnnyDoe said:It doesn't show the initial prompt and the full chat history
It started yesterday, no full detailed history available till now, after the launch I think they will release it.
Pick one model and its prompt started : 18/10
So itâs totally useless
I am better at losing money than a LLM :D
GPT is down 70%
The overall performance of the 6 models is -5% so far. Same as having 6 monkeys trading randomly.
The experiment is over. Results below.
Total failure. We can safely say that using LLMs for trading is a bad idea.
The experiment resumed as season 1.5:
âIn this season, the models are exclusively investing in US equitiesâ
And itâs already a disaster!
Remember:
Another experiment here: AI Arena - Watch AI Models Trade Stocks in Real-Time | Rallies
Here's another one: minara.ai/app/trade/benchmark
Nothing useful, just a curiosity.
True power moves in silence.
