[00:02] dream. We wish we could predict where the market is going. And today, more and more of us look to AI to help us do this. But can AI really predict the
[00:14] market direction? >> When people think of AI, they think about large language models, chat GPT, Claude. And this is something that is used to predict the next token, the next letter, but not necessarily
[00:28] used to predict a chain of of numbers. So, it's very hard to predict the So, it's very hard to predict the market. AI itself, uh, especially LLMs, cannot do that. They're not built They're not built to do that. But AI
[00:42] connected to live market data and some prediction pipeline could become a very interesting solution for traders. That's my answer to to it, and we can, uh, also use AI,
[00:55] uh, I would say, shift the game from prediction to behavioral execution because, uh, that's where traders win because behavioral finance points time prediction game. It's traders lose because of a behavioral
[01:09] leakage. >> So, if AI is not exceptionally good at predicting a market, where can it actually help traders today? That's what we are going to dig into in, uh, this, uh, video. I'm joined by Andrey Raya
[01:23] Oberlina, which is the founder of windy.trade, windy.trade, an AI company that is built to help traders and individual investors. And, uh, full disclosure, this video is
[01:38] produced in cooperation with the windy, which is an affiliate partner of Theta Profits. So, if traders stop thinking about AI primarily as a market predictor, how should we think about it instead? You've been working a lot with
[01:53] AI solutions for traders and should have the background to answer this. >> So, I've been working with price prediction and hedging for almost 10 years before, you know, moving to
[02:06] uh do the same thing for our space. But, uh to answer your question, why why is bad at prediction, not bad at prediction, but it's not built for prediction. It's it's because of the nature of of the technology itself. It's
[02:22] built to predict the next letter, but not you know, a chain of of values. Now, AI may not be best at that, but behavioral finance
[02:34] points time and time again that traders actually don't suffer from lack of information. They they actually need a better decision-making process. And that's where technology like LLMs and AI can
[02:47] excel because it can summarize a wide range of data for you and help the trader themselves like figure out a couple of things, pre-trade, post-trade, or even during
[03:01] the trade to make them more consistent. And that's how we built our CLEAR And that's how we built our CLEAR framework, C L E A R framework, which is based on behavioral finance. It stands for cutting the noise, learning faster,
[03:15] evaluating objectively, and acting consistently, and recognizing pattern. go through each of the element in this framework is step by step. And the first framework is step by step. And the first you said the C was cut noise. So, how
[03:30] can we use AI to cut the noise, so to speak? >> One way that AI is really good at is aggregating information, John. So, traders don't suffer from from lack of information, right? They have charts,
[03:44] they have tweets, they have news articles. How do you make sense of all of that? So, one thing that we do here at Wendy, is uh aggregating all this information. So,
[04:00] when you look at Wendy, for example, uh yesterday, the markets were, you know, going up a little bit, I just asked like, "Markets are pumping. What is going on right now?" So, the tool compiles gamma exposure,
[04:13] computes support and resistance levels, and then also you will see that it goes on Google, it goes on Perplexity, it goes on X, too. X is one of the fastest source of information when it comes to the trading because
[04:28] even though it's a little noisy, uh even the president himself sometimes tweets there. And so, it it does that by it aggregates information and it it gives me exactly what I need here, such
[04:41] "There's no news right now. There's no new magical catalyst. Here are the levels. Please stay with the levels." So, this is one way that you can use the kind of cut the noise and really focus on what is it that it should be
[04:54] important for you as a trader. >> The second point in your framework is learn faster. What do you mean by that? And why is learning faster a challenge? >> This is post trade, right? And uh
[05:10] trading is a wicked learning environment, as opposed to like a kind learning environment. Uh wicked learning environment means the feedback sometimes is slow and noisy. And in an environment like that, experience alone doesn't
[05:24] necessarily guarantee skill building. So, you kind of need to really sit down and and face uh your demons as a trader uh every day, and LLMs and AI can actually do that. An example is uh after each
[05:39] do that. An example is uh after each trade, a trader can upload a screenshot for For with their entry marked, their exit marked, and have the LLM and the AI review
[05:53] their entry timing. They can also add information as like their sizing, for example, and and so on and so forth. And the LLM can derive from that some sort of journal. It's It's basically journaling,
[06:06] but LLM makes it simpler because if you have to write down I know people who do that, but writing down your notes every day after a trade trading day sometimes hard to uh remember everything. So,
[06:20] uh AI here uh in the learn faster is a more like a a tool to really name the behavior, for example, chasing an extension already, and and get the that feedback in in minutes and not months.
[06:34] >> So, what you are saying, we all know that we should really evaluate our trades at the end of the day and review what we did right, what we did wrong, but we are running out of time and a lot of us tend to skip it, but this is where
[06:49] AI can really help us to do this faster and more efficiently and maybe even at a better quality than our own evaluation. Is that so? >> Yes. Journaling is hard for everybody. Uh staying consistent over many, many
[07:02] years when you trade, only a very few people can do that. Obviously, there's journaling tools out there, but you still have to do a lot of work. Uh AI is able to do this proactively
[07:15] um by by fetching your data and even doing it in a more interactive way as opposed to just like a dashboard tool, you know? Uh it makes it more dynamic and also it allows for I would say context behind
[07:31] the performance or lack lack thereof because sometimes something happens in necessarily capture that. It just captures the transaction. But AI can know like, "Oh, I don't know. Something happened today, maybe
[07:46] an accident or something or something that happened in your personal performance." And that is interesting because now we know that when the situation similar is is happening, maybe it's
[07:59] better not to trade as opposed to trying to trying to force a trade. >> And the next two parts of your framework is evaluate objectively and act consistently. What does this mean and why are we
[08:14] combining them? >> So, previously we we discussed post-trade, right? With journaling and so on and so forth. One thing that is important is also is doing pre-trade
[08:27] And um the the damage, you know, we know the damage happens for a trader when there's a gap between knowing the rules and executing under pressure and emotion. So,
[08:40] expert traders, even expert traders actually falls into this uh without naming the name. Um we we speak with hedge [snorts] fund managers that are still like
[08:53] caught up in the emotion, sometimes make mistakes. And these people have been in the market for over 40 years. So, this happens to everybody. It's not just uh it's it's it's just human nature. So, a way of preventing this is
[09:06] having a a process in place pre-trade to um have to do in case bad things happen. And that's where AI can help as well. So, part of our EA in the clear framework is
[09:23] having a checklist. In his book uh Checklist Manifesto, Atul In his book uh Checklist Manifesto, Atul Gawande in uh 2010 described how having a checklist for surgical safety actually decreased errors in
[09:38] decreased errors in and and bad things in in surgery rooms. here. Before entering a trade, the trader can paste their setup. So, Wendy, for example, has your own checklist already
[09:53] ingested through your interaction with her. Also, it it the LLM wants you judge your trade before the trade, it judges against your your criteria this the same criteria, like EMA alignment one and so on and so forth.
[10:06] And can grade that setup and give it back to the trader to make their own decision if they want to enter or not. So, that's what that's the point here about evaluating objectively and acting
[10:22] consistently. John, have you you probably know about disposition effect? Right? How Do Do you think you hold your losers longer than your winners? losers longer than your winners? >> Probably I do because I hate uh losing.
[10:37] With the exception of zero DTE because there I'm very good at setting stop losses. >> Okay. So, this is again, we're all humans, but the disposition effect is a
[10:49] is a thing in behavioral finance that we're more likely to hold on to our losers and cutting our winners short. How do we get rid of that? It's by having a set of rules pre-trade that says,
[11:03] reaching this level, I'm just getting out. And that cuts my losses instead of trying to, you know, holding on and doing diamond hands and and losing my entire portfolio. That an LLM can help you
[11:17] actually build that process pre-trade in order to execute while you're in the emotions because things are already set up before entering a trade. >> But are there any risks here or areas where we should be careful in how we use
[11:30] AI in these areas. >> Yeah, AI should not replace human thinking. At the end of the day, AI is a tool that streamlines the operational process for you as a trader, reading through the
[11:44] news, and so on and so forth. But at the end of the day, you're still the master of your destiny as a trader. Use the AI as a tool like any other but use it Watch out for hallucinations. That's why
[11:58] using like a specialized for example, specialized tool for trading is important because they're less prone to hallucin has hallucinations than, you know, regular LLMs because they have tools. They are
[12:13] trained to to think like a trader and act like a trader. They they have specialized tools connected to them because if you ask ChatGPT, for example, for a key level, they will just regurgitate
[12:28] the training data that they were trained on. So, there's something called the cutoff date in in an LLM training. And have the answer, they will just try to give you an answer but based on that
[12:42] looking at I invite you to do this with with ChatGPT and Claude. They will give you a price, but the price is from 2024 and 2025, not today. So, these are things that you have to watch out for as
[12:58] you work on with LLMs and AI in general. >> The last point in your framework is recognize patterns. And this, to be honest, sounds like technical analysis >> It sounds like technical analysis, but this is actually about oneself. So, the
[13:15] the behavioral pattern for the trader. Again, bringing it back to behavioral finance, we we point to how the trader can generate consistency and and more I guess more more profit through through
[13:30] behavioral edge than trying to predict where the market is going. So, the expert intuition is real. You know, but it only develops in environments that requires that that they can regularly learn from.
[13:43] This needs conditioning. This needs conditioning, which means recognizing the pattern before it happens. So, for example, like I know happens. So, for example, like I know that I tend to not wait for confirmation
[13:57] before entering a trade. What I can do with AI actually is over time over time build this history of all my behaviors and then surface insights that I may not not even be aware of. That's what I we mean by recognizing patterns.
[14:13] It's basically feeding the LLM your full trade history and ask it to find recurring patterns on your trades in your behavior that may not be
[14:28] easily detectable, you know, without because LLMs have the ability to ingest a large amount of data and summarize it. So, that's why we we put it here. For example, the best win rate by time of time of day, the
[14:43] performance in a low volume condition, losses clustering, and so on and so forth. LLMs are very good at that. And again, this is complementary to what we discussed before with the daily journaling. Like over time, you you
[14:55] what do you make of it? you can derive patterns, and LLMs are very good at that. >> We touched on it, but let's elaborate a little bit. What In what areas should a
[15:09] trader be specifically careful not to or language model? >> Language models are not built to predict like price evolution by themselves. It's it's just built to predict the next
[15:24] token, the next letter, and the next word. So, if if that's how you use the LLM, for example, for summarizing uh a journal or an interaction or summarizing the day,
[15:37] it's excellent. Where LLM may fail is we if we trust it with, "Okay, where do you think this is going?" kind of thing or trying to predict a price. The LLM itself is not built to do that. Where it could be excellent is if
[15:52] you have a a live data source from the market, for example, and some sort of prediction pipeline like a machine learning model into the LLM, then the LLM just acts as an
[16:05] to provide with the insight that you need to make your decision. That That is you can use the LLM. So, that's that's the that's the trade-off here, John. Um you should be careful of directly asking
[16:19] about uh ES or SPX. Where do you think it's going?" If the LLM provides an answer, it's mostly just like bollocks. >> Let's sum up. What would you say would be the two or maybe three most important
[16:36] audience to remember from this interview? >> Use AI to your advantage. Uh AI is a tool just like any other tool for you to to be a best better decision maker. That's where the edge is actually
[16:50] uh you know, time and time again, behavioral finance has said that the game is about behavioral leakage and not necessarily predicting where the market to be consistent. The second is LLM is very good at
[17:04] The second is LLM is very good at summarizing information, not necessarily predicting information unless you bolt onto it specialized tools uh like um uh predicting uh prediction pipelines or live data from the market that you you
[17:20] you you have to build that in the back. And uh lastly, And uh lastly, use the LM to know you know yourself. LM help you cut through the noise, but also help you pre-plan your trade,
[17:37] do an analysis of the post trade uh do a post trade analysis, and over a long period of time as the LM accumulates those information about you as a trader,
[17:49] it actually can surface very interesting insight about your personality because you're a very unique individual and you have a very unique decision-making process in your brain, but it's a very hard to understand and surface that if
[18:04] you don't have like a a a tool that does it consistently for you over time. So, those are my three takeaways. >> I mentioned in the beginning that this video is produced in cooperation with uh Wendy. So, Andre,
[18:19] tell us in 20 seconds what uh Wendy is doing. >> So, Wendy is a caddy. It's a It's a caddy for individual investors. Uh we caddy for individual investors. Uh we baked into Wendy uh several tools. So,
[18:34] difference between that and ChatGPT? I think my answer is it's a it's a tool that yes, it orchestrates been orchestrated by a large language model or multiple large language models, but
[18:46] we have several tools attached to it. You have, for example, you have uh options payout diagram. So, it can help you choose you have a certain vision of where the market is about to go. It can summarize
[19:01] news for you from different sources and combine that with price levels to be able to build a pre-trading plan. And it does journaling for you as well because it will judge at the end of the day how did the trading day go. So,
[19:16] that's when the inner nutshell. >> And if you want to explore this AI trading caddy, Wendy, you will find a link on the screen and in the description to this video. You do get 25% off the annual plan or the first 3
[19:32] months if you sign up using this link and use the code Theta Profits at checkout. But, thank you, Andy, for sharing how we should think about using for trading. >> Well, hopefully this has been useful for
[19:48] your audience, John, and it's always a pleasure to to connect.