So, let me pick Open because I know a lot of people care about Open. And so, this is the last... still kind of a work in progress. Bear with me. I'm sure I'm getting some kind of error or something.
Okay, so this is Open. Okay, so this is Open, which is Open Door. Let's check actually to see if this is the exact shape. Yeah, so this is the last 30 minutes. And you can see that that's what that chart looks like. Do you see on the bottom, like kind of scrunched? It's the same pattern. Yes. No, this is the intraday chart of Open. You see that same shape, right?
So, it's like superimposed, right? Everybody see that? Great. So, what this does is it looks in history of the stock market and it says what are the closest Euclidean distance or just the closest pattern. And you see all these little superimposed patterns are the closest pattern. What's interesting is it's telling you that the probability of the stock being up in the next 10 minutes is 50/50.
Why is it 50/50? Well, the mean return of the 20 closest neighbors, so these are the 20 closest matches. And if somebody looked at this and said, well, I really think this is bullish, well, the average 20 patterns that are the closest actually did not have a positive return. It's actually a negative return.
And you can see all the statistics here. And in fact, if you click here, you can see all the matches. Some of the closest matches go back 25 years ago. A company called Myriad Genetics had the same match. And that stock was down 1% at 10 minutes. It's one of these lines. Now, if I want to see the 200 closest neighbors, I can do that as well. I just click run and it returns really quickly. And again, the mean expected return is negative.
So this is like an automated statistical arbitrage tool. It's a vector search. I think that you can look at median, there's standard deviation there. It gives you the probability distribution. You can make a lot of different adjustments, like I have softmax here. You might want to try uniform instead.
You know, how many neighbors do you want to... I don't know, graph. There's a lot of different, you know, ways to do it. There's 60 neighbors. You can see it's kind of a real thick, thick line there. So in any event, you know, there's a lot of ways to do this analysis. Like maybe you want to see 200 closest. The funny thing about this is, guess how many vectors are in the database? Guess how many shapes are in there?
Yeah, we're working on longer time frames and shorter time frames, including second-to-second time frames, which are going to be really fun.
Yeah, you can use this tool and go to godelterminal.com. Yeah, you could do SPY as well. Yeah, it's six billion vectors and they're all in RAM, which is really wild. So, proud of my little engineering. So, we're going to update that to hopefully a little under a trillion vectors, and it actually sorts in, you know, 20, 30 milliseconds.
So, I think that the backtest idea here is that this isn't yet a backtesting platform. It will be shortly. This is just a vector search. So this forms the basis for a good backtesting platform. So the question on backtesting is, what is your strategy, right? This is just data. So there's not actually a strategy implicit here. The question is, what strategy do you want to make out of it? And that's where PAT comes in.
So what PRT is doing, the reason I started with this, this is really interesting. So I took 700 stocks, right? And what PRT does is it's running all 700 at once, right? So PRT is behind the scenes doing PAT on 700 different stocks. Okay? And it's ranking the ones that have the highest probabilities and the highest likelihoods of success.
It won't be free for long, but if you try it on Godel, you can try it right now for free at godelterminal.com. Like I said, I'm planning on selling this to the big quant firms who I've already talked to. But give it a shot. It's very computationally intensive. Yes, it runs on a cluster of H100s.
You can backtest by actually using this tool, but you're not going to use the real time. So if you want to go backwards in time, you can backtest that way. It's just, you know, H100 rig, Python.
Well, I can't give it to you for free if Citadel's willing to pay millions for it. So, but regardless, so yeah, so if you run, you know, put in your... if you just care about, like, let's say you just care about five stocks: Palantir, Open, Joby, these stocks. Kidding, Palantir is a good stock actually. Nvidia. I don't know what other memes are there. And you run.
It'll run really fast and we'll just say, okay, short Nvidia, short Open, short... long Palantir, short... but the edge, it's five times more, or it's like three times more confident on him than any other ones. Short him.
Right, so you can kind of decide what search space to use. I use this 700-stock list and it's, like I said, it's running PAT on every single one of these and it's very, very large amounts of calculations. It's searching through billions of combinations.
Yeah, I mean, you can just copy and paste an S&P 500 list here, but yeah, I'll make it a dropdown where you can just paste it. But yeah, it likes Wingstop a lot here. So you can obviously make an automated trading strategy out of this. Like, so for example, one of the things I've considered doing is turning it into like an API. So you can decide... 27382.
Right, let's see how my Weight Watchers is doing. Up since I shorted it. Flat. LNW. What is LNW anyway? Billion-dollar company. So these are all like relatively large-cap stocks.
All right, so we'll see how this goes. Obviously, if you do it over a few trades, it's not necessarily going to be right every time. There's nothing that's right every time. I do think it gives you a small edge and will actually be fairly profitable over time, and I've used it so far to make some money.
But I do think you have to use it for lots and lots and lots and lots of trades at once, which is why I think connecting it to an API or something like that is going to be the main idea, because, you know, you really have thousands of ideas coming at you every second. So you maybe, like I said, take the top few ones that you like. The funds do obviously write their own code, but, you know, software is something where, you know, it's a matter of standing up resources and sometimes it's easier to buy instead of build.
These are, yeah, really, really short-term trades. Yeah, yeah. Weight Watchers. A long time ago, I think WTW was Weight Watchers. WW.
So, anyway, we're releasing lots of really cool tools like this that aren't available, quite frankly, on Bloomberg or anywhere else. So even though Godel is cheaper than Bloomberg, it has tools like that that aren't on Bloomberg. So I think that's pretty cool. Yeah, Panda. I mean, it's one of these things where, you know, the statistics are also... you have to consider the data.
So like if the market's going up, this is why quantitative investing is usually done by professionals, because if you just get that little uptick and you're only short, you know, you're going to see some pretty rough performance. So you got to make sure you do longs and shorts pretty equally.
And like I said, the advantages are small. They're over time. They're not going to be, you know, instant profits all the time, you know, every day. It's across, you know, cross-sections of like lots of minutes. Like if you do, remember there's 390 minutes in a day of trading.
So if one, it'll give you those 700 scored, kind of best scores every minute. So you get 390 trades, portfolios basically a day, and, you know, you'll be right, say, 55% of the time, and that means that, you know, the average is sort of a coin flip, but over the long haul you'll see the results.
No API yet, but yeah, I'm talking a long, long time ago. It doesn't do crypto yet, but that'll be the next version. What I want to do as well is so-called self-similarity. So what you can do is you could just say, just compare it to itself. So PAT doesn't do that, but PRT would do that. So you just say, just compare me to SPY. SPY on SPY.
And that's kind of another neat way to look at it, is like SPY is the only... I just want to compare SPY to itself. And if you're a futures trader, I don't know how you could do it without this.
I think hedge funds would be liquidating their ATYR positions over several days depending on how much you have. Like obviously it's liquid, 109 million shares have traded, so even if you had a $20 million position or more, right, because it's at five bucks, so four million shares, you know, basically get out in an hour.
Yeah, so, you know, I'm also going to make it so it just updates maybe hourly, but yeah, it seems to be bearish on the S&P. Again, we'll keep thinking about it and growing this platform.
We have so many new things coming to Godel. It's really exciting. We're the only platform that has private prices, which is really neat. So this OpenAI stock chart, for example, I think it'll be closer to 600 pretty soon actually. So I like it a lot.
What else? There's SpaceX. SpaceX is just like space. 2870. These are privately held companies that trade kind of hand-to-computer privates. And we're going to try to facilitate trading with them. We'll see.
>> Order filled. Whimo's probably undervalued, I think.
>> Yeah, of course. We're very different products for now. Bloomberg is a very good software platform, but it's not really designed by people that are in the markets like I am, or haven't really been in the markets, quite frankly. It's a lot of people that are like second-tier people, quite frankly, because Mike doesn't really give any equity out. He owns all the equity himself, which is, you know, kind of absurd. You're not going to get talented people if you don't give them some upside.
But we're a new company. They're a 45-year-old company. So give us a little time and we'll make sure we excel. Yeah, Godel is not for new investors or people who are not on Wall Street, quite frankly. So it's really not meant for retail investors. It's meant for people who are really advanced and trade for a living.
>> Yeah, I don't think those things are really real short squeezes. I talked to my friend who is an options expert. He concurred with me. I don't like SLS.
I guess we can look at Hive since one of you guys requested it. Okay, so it's a foreign company technically. Oh, it's Canadian. Hive Digital Technologies, right here in Manhattan. They're actually very close to me.
I don't think there's going to be any recovery in little old... old tire. I don't know why you bought it. Why'd you buy it? Who told you to buy it? Who'd you listen to? What the hell is wrong with you? You think you know biotech? I'm going to scold you.
There's my wing. There's my wing. Stop. Thank you. Cross.
