I gave Jev $10,000 and let it trade

Jev + Money = Jevv?

😂

most of the time LLMs fall prey to forward bias you should test in a live env even with only a small amount of money and small position size almost 90% of the time if you give full reign to the LLMs to trade on historical data it will always show profitable due to these kind of biases

yeah that’s a good point. live testing with small positions is definitely next, historical replay is mostly just to test the decision loop first.

Yeah Would love to see the results of live testing

btw i tried to explain in simple words how jev works:

You didn't spend $10k why do you lie.

is simulation

@opita04 Oh no….expect a petulant reader or some stupid bot to CN this. It’s very tiring.

@abolbuild Why not say things as they are, not petulant, just want more truth in social sites. Tired of clicking on clickbait.

Plenty of folks will do the same. Try inverse jev instead

haha inverse Jev could actually be a fun test too

No you didn't :) click bait, what you have is to claude to build this for video :D

ok

What's the alpha? How long have you run? How does it do bet sizing?

It’s just ~21h of historical replay, not live trading yet Right now it basically goes full size based on which side has the higher probability, so better sizing is probably the next thing to improve Mostly just testing the decision loop here, not calling it a strategy yet

Yeah, that's very cool. There's a lot written about when you have a trading signal what to do with it. So that includes things like how confident your signal is be sizing and direction and existing portfolio. Merging signals has a lot of literature as well.

yeah exactly next i want to improve sizing and also add news. I think Jev can help make the probabilities better with real-time events.

Sentiment analysis on News is a very crowded trade and one that is very old. Not to say that there is an opportunity here but it's very well trodden ground. See RavenPack

fair point my thought is more that Jev could use the news together with market data to make better trading decisions, not just do sentiment analysis

This is the perfect model for it

El problema son las comisiones, si no fuera por las comisiones podría llegar a funcionar apoyado con otros modelos

Jev is not where the edge is

It’s just simulation!

Yes You can, with the help of Binance's public API, snapshot the data and test on it and improve

Interesting baseline, but without dynamic bet sizing the EV will be volatile, a modest position could reveal true edge

yeah. smaller positions should make it easier to see if there’s actually an edge

continuously trading stocks every second is how you lose, actually

yeah, every second would probably just be trading noise 😂 the idea is to let it decide when there’s actually enough signal to act

THIS IS INSANE. We’re so close to achieving human reasoning in these models, it’s insane how much potential good we have for the future!!!

Have a feeling the smaller timeframe it will do better...

yeah could be🤔

Performing similar to math random lol, but to be fair forex exchange markets are not data driven nowadays

fair that’s why i’m more interested in testing it with more context like news + market data and seeing if it can be better over time

right, connecting the right data sources are like half of the solution here

Can I start with him from 10$?

not yet, it’s still just a simulation setup for now

trade: 1. Already showing one at the beginning of the video. What are we supposed to see?

the point was more to show how Jev is making the decisions, not just the trade count.

losing money now

Is that $10,000 of real money traded on an exchange, or a simulated portfolio? The playback controls suggest a replay, but “I gave Jev $10,000” reads like live trading. Can you share exchange fills if it’s real, or explain how you kept future price data out of Jev’s inputs if it’s a backtest?

looks pretty random to me...

Can i bet on this?

If anyone does this for real, just be prepared to get a bombastic side eye from your accountant

I belive the most important thing is what data you are feeding into Jev and what outputs you have (type and question referencing the data). Raw price is probably not enough for the classification. Do you plan to open source the code or share it?

May I ask what your input schema looks like in such a scenario? Could you share an example?

then everyone uses JEV who win? lol morons

🥹🥹🥹

@RanYEET

Good let's see you make your first 1 Million and report back the progress here

It seems like you forget to tell buy low sell high

How did you hook this UP?

At last i can lose money FASTER