 Before I get into the project itself, I want to tell you a little bit about who I am and who Kyle is as well. So myself and Kyle Vogt, you may recognize his name. He's one of the founders of Justin.tv. That's what he does these days. But at the time, two years ago, we were both students at MIT and clearly we had too much free time on our hands because we kept doing all kinds of bizarre projects. One you might have seen online, it got written up a few places, was our safe cracking device that we built that cracked high security safes. But this is entirely different. But another good question to answer when dealing with something like this is who were we not? We are not stock market experts or spammers or get rich, quick scam people. We are none of these things. So everything that I say involving the stock market or spam or anything involving that, there's really no credentials to back any of that up. So don't take any of that as fact, okay? First off, quick spoiler alert. Everything will be seen through a soda straw. There's, because all of this was what we were looking at at the time while we were doing this project, it was a very limited set of information that we were looking at. We didn't have all the information about what botnets were operating, what spammers were sending, what we were literally just looking at the spam that was coming out and extrapolating from that. So it's all from our point of view at the time. We couldn't see the forces behind anything and everything is guesses and hypotheticals, but you'll see that. But from that, what you're gonna see, and I'll give it away right now, the moral of the story is a lot can be determined without the underlying information. It's all about how you look at the information, even the information that everyone else all has. So how it all started, October 2006. That was a time when everyone in the world was getting all of these ridiculous stock spam messages. I'm sure everyone here seen it. Yeah, everyone, everyone seen these? Yeah, absolutely. Everyone was getting these stock spam messages. October, my friend Kyle, he saw one of these. Advertised GDKI, Gold Mark Industries. Fantastic sounding stock. And he said the stupidest thing I'd ever heard, ever. He said, there must be a way for us to make money on all these people trying to sell spam stocks. It was like, no, I think it's the exact opposite. And I said the obvious answer, which was, you're an idiot. And that normally would have been that, except for the fact that a lot of these discussions that we had that sounded ridiculous at first, one of us would go off and actually take a look at these things and kind of do the research and would come back and find out it's not so ridiculous. So that's actually what happened. And that's why I have a whole talk here. So this is why I thought that Kyle must have been wrong, right? Profit in markets is driven by asymmetric information. If there was a perfectly liquid market and everyone knew every piece of possible information about it, there would really be nothing to trade on. No one would be selling anything for a price that other people would want to buy it at, right? Everything would be just fixed price. It's all about that I might know something that you might not or you might know something that I don't. Everything that everyone already knows is already priced into the value of any given commodity. And in this case, my problem with it was, well, everyone gets the spam, right? What can we possibly know that everyone in the world, hundreds of millions of people, don't already see in their inbox every single morning? What is the special information that we have that other people don't? But first, we'll get back to that in a second. The real big question a lot of people ask is, what are these people actually trying to do with all this spam? It tells you to buy a stock or something and then, I guess, they do something? I mean, the plan is obvious, right? They phase one, send stock spam, phase three, profit, right? They must be doing that. So there's a little more details to it. It's what's called a classic pump and dump scheme. And these things are as old as markets themselves, you know? We'll go through it step by step. So basically, let's say that I am the scammer in question and I own 100 shares of worthless ink at $1 a share, okay? And we're right where the blue line is in time. So I own this worthless stuff. It's the kind of stock that nobody's buying this. I mean, it's garbage, right? So then I go on message boards and I tell everyone, oh my God, this stock is gonna go through the roof. It's gonna be the greatest thing ever. You shall buy it. And so then, being the good people they are, they go and they buy the stock. And as they buy the stock, I'm selling them little bits and the price of the stock, the value is going up, going up and up and over over time. The price goes up as the demand increases and then I go and I sell all my 100 shares or whatever I have left at $2 a share. So I just hit step three, profit, and I'm gone. I have no more exposure to that stock whatsoever. And then the next step happens, which is that surge in demand that occurred was totally artificial. There are no new buyers anymore. People that did buy in when I sold them all those shares, they now try to sell the shares back because they wanna make a profit too. But there's no one to sell them to. There's no more people hearing that it's great. There's nobody to sell it to. So the price starts to drop and drop and drop and drop until eventually it ends up that they're getting rid of their shares at a price below what the shares started at. Back when I bought them when nobody was interested in the company. And eventually, as you see in this diagram, and this graph is an actual stock from 2006. It was a JNNE over the course of one day. You see it starts at market open, it ends at market close. And you can see in the end it oscillates and ends up closing close to what it opened at, but not a lot. And when you look at these things over time, eventually they recover to the initial kind of actual market value price. But in the meantime, I just created this huge fluctuation really from nothing. Just out of my own imagination and telling people that it was cool. So when you look at the graph, there's the size of my profits. There's the size of everyone else's losses. It's pretty straightforward. That's a pump and dump. All it is is it's touting a stock. It's just shouting from the rooftops, the stock is awesome and making sure that people buy that. Concept is old, word of mouth, boiler rooms, the classic people calling up old folks in Florida telling them about amazing investment opportunity. Forums are big, but spam really opened it up. Spam provided this huge inexpensive way to send that information to millions and millions of people. And, you know, so as you saw, profits are determined by when the tout sells out their mass block of shares. The losses for the suckers are determined by how late they bought into the scam and when they try to sell out again. Obviously in that graph, the people who bought in when it was near its peak, they're screwed. I mean, they are dropping all the way down and they're losing, you know, probably 10%, 20%, 50% of what they bought in for. What kind of stocks are these? You know, what does this work on? Well, they're what's called penny stocks. They're not traded on big exchanges like NASDAQ or the New York Stock Exchange. They're traded on little exchanges, we use the term traded very loosely. They're barely traded. A lot of these stocks, you look on a day-to-day basis and their volume is zero, practically. They're not on major exchanges. They're on things like the, they're called over-the-counter. They're traded on exchanges like OTC, BB, OTC Bulletin Board and a market called Pink Sheets. They're thinly traded, like I said, near zero volume on most days. And they're high volatility. The values of these stocks, they're generally under like $5. Many are under like $1. So we're talking maybe like a 20 cent stock, right? So if that 20 cent stock goes up by 10 cents, that's a 50% gain in value, right? So it's huge. The volatility of these things just jump all over the place. Because of that, that's what makes these stocks particularly, you know, particularly possible for you to mess with them. You could spam all day about a big New York Stock Exchange spam. You know, I could tell you all to go buy Apple Computer tomorrow if the Stock Exchange were open tomorrow. But I could tell you all to go buy Apple Computer and all of you could all go buy Apple Computer and we would all collectively probably have no effect on that stock, right? Because there's such a high volume of trading that already occurs on it. So the bigger point here is it's very illegal. You know, the pump and dump scheme, it's illegal. The all changes in supply and demand of what's going on in the market, it's all artificial, it's all made up. There's no actual information about the company that's changing. The SEC frowns upon this. So, but really, the real question is who is dumb enough to buy a stock because someone emailed you telling you to? And this was my problem in the very beginning. I looked at Kyle and I was like, nobody buys these. This is entirely impossible, right? I delete them, you all in this room, I'm sure deleted them. But the problem is, apparently somebody didn't delete them because we looked. And what we found was a lot of people, a lot of people bought these. GDKI, in that week that we first saw this, we looked at it and it spiked 60% in value and the volume was over 600,000 shares. So if you do the math for the spammer that was probably a quarter of a million dollar profit. You know, a quarter of a million dollar profit on like a $1 a share stock. And we were kind of impressed by this. This was peculiar. If you look at this graph that I've got here, you'll see, and I can, I'll highlight this for you. But you'll see here, this day right here, this is the 20th of October, this is a Friday. That's when the spamming started. You'll see that the starting value of that stock, the opening value, the low was about a dollar, a little over a dollar. The closing value was like $1.30. You know, it started to go up and then Monday, huge spike in volume, huge spike in price. And then over the course of the rest of the week it started to just go back down over time and that value was lost. But the interesting thing here is that the email started on Friday. They started before the market closed on Friday. So there was an opportunity to buy shares early. And that intrigued us. We kind of saw something there. So, but even looking at that graph, right? Quarter of a million dollars. Wow, quarter of a million dollars. That's a lot. It turned out that was actually tiny, tiny in these shares. Same stock, couple months later. This was a 300% increase over five days, 10 million shares, so possibly a $30 million profit. Absolutely immense. And these are the sorts of things we were seeing, like on a regular basis. And we were like, wow, this is ridiculous. You know, I don't know who's buying these, but they're buying a lot of it. So, we looked at more of this and actually remember this particular chart later because I'll talk about an SEC investigation that involved this particular week of trading on this particular stock. So we had our answer. It did actually happen. It did actually produce results for the spammers, which was incredible. So, the next part, what could we do about it? What could we, how could we participate, I guess, in a legal way? But wait, right? So we look at this first week, October 20th, 27th of 2006, and we say, you know, not all of them were winners. We received spam emails for 20 different stocks that week. We followed all of the stocks and three of them produced profits. The other 17 produced nearly nothing. I mean, practically no shares bought or sold in that week. So, a lot of people spamming, a lot of people talking about a lot of stocks, but only a couple actually being profitable. And those couple that are profitable are suddenly wildly profitable. So, why is that? What are the different characteristics here that allow us to decide which one is successful and which one isn't? I don't know, that's what we wanted to know. So we looked at what we had, right? What data do we have? Well, we have a lot of spam and that's mostly it. We had about 1,000 stock spam messages every single week. And then other than that, we were just looking at the freely available stock market information that was online, bigcharts.com or whatever, and just looking at what happened. So, we could look at the results of the previous week and we could in some way compare it to the huge, massive spam that we received about all these stocks. So, that's all we had. I mean, that's not anything special. It's the same information that everyone else in the world had. So, what did other researchers see at this time? There are two big papers that you'll find online by two different sets of researchers, Frieder and Zitrain and Hank and Hauser. Frieder and Zitrain is the big one that everyone mentions. It says, you know, spam works and it's a very good paper. You should actually check it out. It was based on data from 2004, 2005. So, it was a little dated by the time that it got released, you know, at the speed that spam changes. But it showed some interesting correlations. It basically said that the more spam email that a spammer was able to put out, the more effect it would have on that particular stock. So, this was pretty straightforward. Hank and Hauser found a very similar correlation. The funny thing was though, we were watching this in fall 2006 into 2007 and there were a couple articles that came up about that time which said stock spam is dead and they said there was no longer any effect on the market due to stock spam. And this really set us to scratching our heads because we were staring at this ourselves and we saw these huge effects. And then we looked at the data sets that all of the various researchers were using. And what they were doing was they were going through, you know, these large collections of noted spam email and they were searching for stock symbols. They were searching by text, right? The problem is by fall 2006, any spam filter of any sort was good enough to filter out text, right? I mean, text, it's easy. You know, you look for three or four letters that are all capitals and something that says buy, boom, filter it out, right? By 2006, there were a lot of spammers that were getting huge results in the market. The problem was the only ones that were getting any results were all graphical based spam. They were all pictures. So there were a lot of researchers that were missing huge portions of what was going on in the market at the time. And here I explain exactly what I just said. Quote from Frieder and Zittrain's paper basically saying how they found their data, which was they went through and they searched for stock symbols in the text of emails, which for the data set they were looking at 2004, 2005, worked, it totally worked. Everything was still text back then. Graphics were just starting to come out and spam. So no problem, they found that correlation. But others who followed their methods started to see nothing happening anymore. So that becomes the question, right? You've got all this graphical spam. You know that the graphical spam is what produces results in the market, but you also know that as far as sorting it, using any automated means, the whole point of them doing the graphics, the whole point of how they're formatting it is to prevent that. It's to prevent spam filters from reading what it is in the message. So how do you sort graphical spam? Anyone? Anyone? By hand. That's how you do it. So we did. We decided, you know, we played around with a couple image recognition type programs a little bit, and we decided to do it by hand. So we did. That's how you do it. That's, you can see on the left side, that's on my computer, there's a folder which says week three, and there's a whole bunch of folders that have stock symbols, and in each folder, there's hundreds and hundreds of emails. And that's how we did it. So we sorted it. We followed it for 14 weeks. We, Kyle and I hand sorted about 50,000 spam emails, and resulted in 12,168 stock spam emails. Pretty good. So this allowed us to sort it in real time as we got them. So what could we get out of this? Obviously we already said we know previous results of what happened, but there's two things that we can infer from the data as we collected. We can figure out the relative botnet power, and we can identify the spammer's unique signature. And here's where the fun starts. So let's talk about botnet power, okay? You sort the messages by the actual stock symbol, and then you plot the total emails over time for each symbol. Pretty simple. Let's look at a graph. So this is what we're looking at, right? Now I understand it's a little hard to read, but let me explain. The y-axis here on the left side, that's the total emails received for a particular stock symbol, okay? For a particular touted stock. So each line, each line that I've graphed represents one particular symbol. You'll see I've identified GDKI, that one that we tracked in the first week. The x-axis is time. The yellow lines are the divisions between days. We start at market close on Friday, and we end at market close on Friday. Again, the other side. So I've got green dotted lines that are vertical that show when the market opens, red dotted lines that show when the market closed, and you start to see some interesting things. So you'll notice that when we already said, we already saw from the previous results that GDKI had the biggest impact on the market that week. And then we graph it, and we see, oh, they also sent out the most email that week. Okay, interesting. Makes sense. A lot of these other stocks you see at the bottom, the little tiny lines that kind of pop out and disappear and pop out and disappear all week long, they got nothing, right? Nobody bought their stocks, nobody cared. But another relevant fact you see, if you look in the sections between market open and market close, a lot of times you see that there's no emails sent out during that time. Right there, you see just these beautiful flat spots. Every time the market is open. Presumably, the guys that are sending emails are probably busy trading at the time. They're probably busy selling those stocks. That may explain why they're not busy sending out emails at that time. So that's interesting, right? GDKI, that's, and the real take home from these, you start to look at these a lot, and you look at a lot of these graphs, and you figure out that the slope of those lines, the slope of those lines, that's giving you the rate at which they're sending out these spam emails. What you find is that the rate that you see on any particular spam email stays constant for a particular botnet that's sending spam. What does this allow you to do? It allows you to identify botnets or spammers, whatever you, you know, one organization sending out email, shall we say. It allows you to identify them just by the slope of the graph between weeks. And that's what we did. Next, let me talk again about, next, the spammer signature. Spammer signature is just a fancy way of saying one guy's email kind of looks the same from week to week. He's got his own particular style that he does. Everyone's got their own bag of tricks for evading the spam filters, right? And those are very kind of closely guarded style kind of secrets, right? So there's the layout, the encoding of the email, any type of capture type obfuscation, you know, little squiggly things to make sure that a spam filter can't read the text. And just generally their style. It's kind of hard to quantify, but it's really, really easy when you're looking at 50,000 spam emails all the time. It turns out you just recognize stuff really quickly. So let's play a little game, right? Here's an email from someone and it tells me to buy this stock, S-R-R-L, okay? And we're in week N of watching these things and it turns out that S-R-R-L ends up doing fantastically that week. Oh, great, you know, but now the week's over. So what does that do for us next week? Well, next week we receive a bunch of spam emails and in week N plus one, which of these emails might be from the same spammer? It's pretty obvious, you know? Maybe it would take some work to make an algorithm that instantly identifies this, but if you're a human and you're staring at these, you see it, you see it instantly. You see it the very first email you get from this spammer. You know who that spammer is and you know what his results were the previous week and the week before that and the week before that. You become kind of intimately connected with these spammers. You've been watching every single email that you get from them and you kind of feel like you know them after a while. It's a little scary, but so that's their style, right? So then we connect the two. We connect the style and the botnet power and then we look at the graphs. Now, this is the graph you already saw of week one. We have GDKI and then we look at week two, okay? Here's week two. Can you spot the same spammer? Yeah, he's like the exact same line, right? His, the slope of his line is exactly the same. It's unchanged, but it's a different stock symbol, right? It's SBNS, okay? Now let's, we go into our spam and we look at what SBNS that email looks like. Hey, it looks exactly like the GDKI email. Okay, great. So we know that this is the same guy sending us this email. So what does that tell us? Well, we know that the week before he sent out a whole ton of emails and he got a huge result on the market. So it doesn't take a rocket scientist at this point to kind of extrapolate. And the results that we actually found, because we watched this for a number of weeks, you know, without even considering buying any stock, we looked at this and we found that the relative strength of any given botnet stayed relatively constant over time. They were trying to send out at the maximum rate that they were able to achieve on their botnet. And because of that, they achieved a relatively fixed slope each time. And then their style of their email was gonna stay the same because they wanted to use their same set of tricks to evade the spam filters. So another example, right? This is week three now. We see SRL there. He's got the same slope. He's the same spammer. You go to this email. Oh, yeah, he's the exact same guy. Okay, but now we've got two other guys. Two other stocks popped up kind of out of nowhere, week three, and they've both got pretty good slopes. The first one kind of rises somewhat slowly, but the second one is very, very specific slope, right? It's very recognizable. It's got a very fixed value. The problem is you look at these stocks week three and you look at the results on the market and what do you think results they had? I mean, they were sending out more stocks than the first guy. What do you think they had? Nothing, nothing at all, like zero volume, right? Why was that? Okay, well, first you've got that SRL. Okay, looks like the same one. Now here's the other two. What's the problem there? It's text, it's text, everybody filters out text. So what do you have here? These guys were pushing enormously powerful botnet. The strongest one we could track on any of our stuff, but they were getting no results because they were just sending text and they didn't understand that because we kept seeing them week after week. This week, week four, they start to mix it up. We've got our same dependable guy pushing MPRG, same results in the market every single week, and then we've got these really, really steep-sloped lines right here. As you can see, blue, orange, pink, blue, they're all over the place. They've got a huge steep slope. They're sending tons and tons of emails really quick, as fast as they can, and they're getting no results. And they try some different strategies, right? One week you see them, and this week you can count it up right there. They touted five different stocks that week, five different stocks, and they got nothing on all five, right? They didn't quite understand what was going on. But the other guy, just touting one, doing dependable, simple botnet, boom, he's good. Once again, there's our same guy doing the same dependable thing. Once again, same botnet. You can see the dependable guy going on and on, and you can see these huge slope guys that they're super powerful botnets just crushing everyone with text-based email that no one is seeing, and they got nothing, right? So these are the sorts of conclusions we were able to reach just from looking at the spam we were getting, and that's it, right? Nothing. That's pretty much that, right? Week five, same thing, the text-based guys are getting nothing. Week six, the text-based guys, they go all out, right? They're done with switching stock symbols, they're done with everything else. Man, they are gonna hit this one like nobody's business, and they do. You'll see the scale on my graph has changed. It's no longer up to 300, it's now up to 900. They're crushing like everyone else in the world with this spam, and this week they still got no return, just nothing. The other guy, still the same results. Now, two interesting things happened in week six, and this is where things actually got interesting for us, because we had been watching the five previous weeks, we had been seeing these ridiculously dependable returns, and we had been able to predict them every single time, right? So we were like, oh man, let's get in this market as quick as possible, this is awesome. And then week six comes along, and week six, that curve on WEXC, he sent out the same dependable amount of emails. He, they seemed to be the exact same thing going on, and the results in the market were like nothing. They were like nothing at all. It was very weird, and the problem is, we saw how much this guy bought into it. It was about a $2 million buy-in, right? Because whenever we received these emails, we would look at the stock, and you would look at the few days before you received any emails about the stock. And you could see, and you could roughly approximate how much money that spammer had put into the stock to kind of buy-in before trying to sell it off. We estimated it was about $2 million. The problem is they weren't able to let go of like any of it. I think, I mean they let go of some of it, right? But it was at practically nothing. So we estimated they lost about $1.2 million that week. And then week seven, they disappeared. They were gone, just gone. We got like maybe one or two emails from that botnet the whole week. And that was it, poof, you know? You'll see on the scale of the graph nobody's going up to 300 anymore. Nobody's doing anything. It's just the kind of little spammers that weren't really producing any results right now. And the one guy that was producing results, he's gone. And just disappeared. So as conflicted as we were about it, we were a little sad, actually. It was kind of weird. I'd never like felt sad about a spammer not sending spam. But it turned out that was the case. Then week seven, week eight, he was gone, week nine, he was just gone forever. I think around week 10, I saw some emails from that same botnet that was, you know, he had moved on to pushing some sweepstakes scam or something, you know, I guess, greener pastures. But in the meantime, the text-based people that were just crushing the world with their text-based spam that nobody saw, they kept spamming like crazy. And they kept going and going and hyping emails week after week. But in general, you know, here's them again, pushing out 900, you know, from our perspective, 900 emails in a week. Everyone else is like below 100. They're nine times the power of everyone else, and they're getting your results. So then the next week, I think they split it up again. They, you know, the usual. They flip between hyping one stock, hyping five stocks. Week 12, once again, just nothing. The whole market kind of fell out of the spam business in about January 2007. And we didn't really know why from our perspective. And once again, we were a little sad. Week 13, the text-based people kind of lost their minds. They spammed 15 different stocks that week. Just, I mean, I don't have the graph of that one because we kind of stopped graphing stuff at that point we gave up. But I mean, you're just talking about 15 little spikes. Boom, boom, boom, boom, boom, the whole week long and obviously no results. I don't know what happened, but they couldn't seem to understand why they weren't producing results. But we could because we saw everything, right? So what, right? What did all this analysis get? Well, what it gave us was that it meant that when we saw a spam email hyping a stock, we could predict the results in the future. We knew what was going to happen. So instead of waiting to see all the various researchers that you would have seen that were talking about stock spam in 2006, 2007, they were all talking after the fact. It was all hindsight. It was all this game of like, well, these results happened and these were the emails that they sent out. So, oh, great, this is what happened. None of them were predicting anything. But what we were doing was predicting it every single week. So any given week, we would see one email come out. We would see the very first email for a stock symbol and we would know what the result of that would be in the week. So we could be basically the first ones to buy it. So where does that put us? We buy the stock there, like right after the spammer bought his, you know, depending on the spammer. But, and then, you know, we could kind of just ride the whole game exactly like he would and, you know, just kind of take a piece of the pie of the touting that he had done without, you know, dirtying our hands in the world of illicit spam and whatnot. So that's what we did, right? The overall method that we did. We sorted each week's worth of spam by ticker symbol. We identified the spammer by the email style. We matched them to their past results. We identified the top spammer of that week. And then when the first email from that top spammer arrived for the next week's spam, we would buy that stock and then sell it eventually and, you know, step three profit, et cetera. So did it work? Yes and no. No, it didn't. The method worked for a few weeks. Like I said, we were watching it. It was amazing. We all had incredible dreams of millions of dollars and Lamborghinis and whatnot. But then, you know, that week six, seven, the whole thing fell apart. All these spammers just stopped getting results. Nothing was happening and, you know, that's it. Those guys, whoever they were, lost $2 million or something and then disappeared, never to be seen in that kind of game again. And there was some speculation. I read a Google article at the time that they were saying that the entire quantity of spam at that time had taken some huge hit. Nobody was quite sure whether it was some mix of, you know, botnet takedowns, ISP shutdowns, or software patches of various sorts. But whatever the case, the whole spam world was kind of took a hit at that time. And so the stock spam people stopped making money. Also, there was a major SEC crackdown at that time which they called cleverly Operation Spamalot. Operation Spamalot in March of 2007, they started out by suspending trading on 35 of these penny stocks. Almost all 35 of which we had looked at and we had seen, because, you know, these were the stocks happening. Two guys in Texas, they were indicted for securities fraud. They eventually reached a $3.5 million settlement. It was estimated that they had made a total of $4.6 million in profits. But, you know, I mean, all these are so speculative and even from the SEC's standpoint, it was hard for them to actually determine which stocks those people had been involved with. You know, they nailed them on, I think, like 13 different stocks. But, you know, who knows? Who knows what they were actually doing? And as it says here, the operation was actually started because an SEC attorney was getting that spam. You know, so that's the one thing you wanna not do if you're a spammer, I suppose. So the big question here, you know, maybe it didn't quite work out for us then, but could it work again, right? I mean, in theory, if the botnets had still been able to push out their same level of spam and enough suckers had still been buying stocks, spam stocks for some reason, our method would have continued to work. So could it happen again? Could this whole thing come back? Maybe? I don't know, right? So I went ahead and I looked before this talk, you know, back in April of 2009, just a little while ago. I sorted through all my spam once again. I sorted through, you know, maybe 1,000, 2,000, 3,000 spam emails, and I sorted them by what type of spam they were. And there was not a single stock spam. I was actually a little hurt, but we have drugs, scams, watches, diplomas, sex books, jobs, and gambling, but no stocks, no stocks. So clearly, at least from what I can see of the spam and really what I can, you know, some people have asked me, you know, well, you can't really say that there's no stock spam happening. There's no touting of these stocks. Well, yeah, that's true, but if I can't see it happening, then you all can't see it happening. And, you know, Joe Schmo out who wants to buy these stocks, he can't see it happening. And in that case, it's not working. So, you know, clearly whether or not there might be a couple of these guys trying to do it, it's not working. It's gone. So will it happen again? Maybe, I don't know. Spammers have given up on stock manipulation for now. If it starts again, our method will probably work again. Or will it? The question here now is, you all know about this. Okay, let's play this game. So you know how to do this. If you all go out and do this, if spammers start setting these again, and they now have a pool of people who understand what's going on and who will clearly buy into this, that'll produce increased liquidity, which will probably increase the spammer's initial profit. But the other problem will be the stocks they'll tank faster. Because all of you, you know that you need to get out as soon as possible. And you know that you need to get out before the spammers, and you know that you need to get out before other people like you. So you're all gonna buy in and get out really, really fast. So that timeframe on that little blip of what happens to the stock, it starts getting shorter and shorter and shorter until eventually there's a tiny little blip and maybe it just crashes like crazy at that point. And there's literally nobody buying it because the only suckers are you. There's nobody else buying the stock. So it's kind of an artificial demand created by my talk. Okay, okay. So we start there, maybe it'll still work, or maybe I've created a new meta strategy to act on the fact that you know, that I know, that you know about this. Okay, okay. Maybe, maybe. Only time will tell. So now, I'm open for questions. Any questions? Sorry, go ahead. Good question. He was asking whether we checked the other markets, European, Asian, et cetera. No, we didn't at the time. We didn't look at any of those. The biggest question there for us was that we considered these issues, but we didn't see that stock spam, right? And we don't read Korean or whatever. So even if we were seeing them, we don't know what that means. But for instance, I know, I've got on my notes there was a German group that was doing this on the Frankfurt Stock Exchange in like late 2007. They had actually evolved from graphical-based emails to PDF-based emails. That was their new way to get around it. And they had big success. They were also caught and prosecuted and all that good stuff. But it does evolve and it does move to other markets. I have no doubt that the Asian markets could see this. I don't know if they've seen it yet so far. Next question right here. Right. She's asking, do we know anything about why it crashed there in week seven? Why those guys lost a ton of money? Did the spam filters magically catch all their graphics? I have no idea. I really wish I knew better. Like I said, the scope of what we did is fairly limited to the view that we had at the time. I've looked since then and I've tried to look back and find the various articles and SEC investigations to piece together which exact groups and the names of the people that were doing the various different botnets and stuff. And you can find a couple. But the coverage of penny stock fraud is pretty limited, as you might imagine. Next question right there. Right. Okay, I'll answer that question real quick first. He said, what gave us the confidence to attempt to get a piece of this without thinking that maybe the SEC might investigate us too, thinking that we were a part of it? And yeah, we didn't know anything. That's, I mean, yeah, that could have probably happened. I don't know. The biggest reason there that I think we, yeah, well, the biggest reason was because the quantity of shares we would have been moving was minuscule compared to these large buyers, right? That was the biggest thing. So they probably wouldn't care about us. For college students, it would have been like a ton of money, but for these guys, not much. He asked if we think of riding the curve the other way and shorting it at the high. The thing is, these are thinly traded stocks, so shorting doesn't really exist in that sense. Shorting does exist, but it's, we can't do it basically. Next person, you're right there. Ah, good, that's a fantastic question. The question was, do we know why they picked particular stocks? There's a lot of possibilities, and I've read about a number of possibilities. In some cases, they were inside jobs, right? The spammer basically offered to the company to take a portion of their stock at a low rate and hype it up and make profit for themselves and the company. That happened, there were other companies that just had no idea this was happening. Just one day they woke up and their stock was flying all over the place, and they had no idea what was going on. I think all types of this thing occurred. Actually, one interesting thing, when I talked to earlier about how we could look at a stock and look at the previous few days of the previous week or so before they had spammed about it and determined their size of their buy-in, sometimes you would look at a stock and you would see no buy-in, hmm, right? What does that mean? Inside job, right? So you could see both types, and to some extent you could guess at which type it was from looking at the previous week leading up to the spam. Next question, in the back. Right, right. So he's asking about what scale would we have to trade at from our little college student side to make up for the transaction costs of trading these shares. Because he's astutely pointing out that when you trade in these thinly traded stocks, most brokers, they're gonna charge you some kind of larger fee than normal. We found it wasn't a problem. The even buying a few thousand shares, it totally made up for it. Because normally you have that problem with trading these sorts of shares because trading these sorts of shares in a legitimate context, you would expect the changes in price to be like, what, like 5%, you know, something like that. So you start getting close to your margin there. But in this case, we were looking at changes of like 50%, or doubling our money, or tripling our money, right? In which case the, you know, 5% premium and the amount of charge on those shares didn't matter anymore. Yes. Yeah, he's asking about, did we look at more of the content of the emails? For instance, the content that these spammers were claiming, you look at these emails and they say things like target prices and target dates, like they've got all this cool insider information. Really, we found that for the most part, we looked a little bit at it, but it was all noise to the best of our judgment. The target prices they did, they were just arbitrary. I mean, even the example you can see right here on the screen, the right side, WEXE, that was that week where those guys, more the target dates. Yeah, the target dates didn't mean anything either. The only thing that correlated with anything was the volume that they were able to send out. You know, because to the targeted user group, right? You expect that their view of it is not about what the date range is. Excuse me. They just wanna get into it as soon as possible, right? So everybody rushed in as soon as they decided that. We didn't see any correlation to these various things. You know, because it's not acting on real information, right, there is no press release coming out, there's no product. Yeah, from what we saw, I don't think the duration of their spamming campaign ever actually correlated to what was going on. Sometimes you would see, especially earlier on in 2006, you would see the spamming campaigns kind of, the trading would last more than one day. Later on, as you approach December 2006, January 2007, the trading was all done within one day. The demand just couldn't put up with five days of trading. Next question, right there. Right, yeah, okay, so he's actually asking about repeat hits. That would be a stock that's touted one week. It does that full up and down like you saw, and then it's touted again like the next week or a few weeks down the road. Yeah, and we did see that. We did see that with a number of stocks pretty well across the board. The second, third, you know, and number of times they would do it. Each successive try would have worse results. One thing that was kind of interesting though that you would start to see is some of these stocks, like GDKI, LITL, a couple of these other things, you would see it spammed about one week, and not that it was spammed the previous week or the week before, but if you looked, some of these were hit like the year before, or they were hit like eight months before, a year and a half before, some of these stocks were getting pounded like on a regular basis. And I don't know whether they were inside jobs or whether they were just really unlucky, but some of these were really, really getting thrown around all the time. Yeah. Yeah, so he's asking about basically data analysis of all of the OTC or pink sheet stocks in order to see these things without even looking at the spam, just seeing that one is starting to spike or that there's an abnormally large volume of buy on a particular stock, and then assuming that that will become a spammed stock. And you're right, that's doable. And you can do the data mining to try to look at the whole market to find these hits. You do run into problems though, because like I mentioned, not all of the buy-ins from these people were successful. So you might see an enormous volume buy-in and then you join in too and nothing happens, right? Maybe they didn't even get any emails out that week. And we actually, we did a little bit of looking at this market where we looked at some stocks and we tried to, we basically did exactly that. We would look at the volumes and we saw some where it looked like there were buy-ins and then we never saw a single email about that stock. So it was kind of weird. There were also some of the stocks, I mean, once again, so to straw of view, we weren't seeing all of the stocks that were touted. Some of the other websites around that were kind of tracking this thing and building little fake portfolios around them, they occasionally popped up with stock symbols that we didn't even see. I mean, we didn't even see a single email about them. So we were still our very limited view, so who knows? Yeah. Good question. Okay, there's two questions there. In that limited view, where did our sample set come from? It came from my email address at MIT. That's where it came from. Myself and Kyle's actually. At that time, we were getting tons and tons of tons of spam email. Part of the reason for that and I'll kind of let everyone know this now. MIT has a lot of mailing lists. The place runs on mailing lists. So the spammers would get the emails of these mailing lists and then so you would get multiple copies of emails. You'd get it through all sorts of different sources. How representative was our sample? I don't know. It's really tough for me to say. It seemed representative enough to the sense that I mean, we saw all this happening, but like I said, there were certainly things that we were missing, definitely. All the way in the back there. Uh-huh, true. Yeah, yeah, sure. Yeah, so he's asking if we considered other factors such as hacked accounts and things like that, buying some of these and moving up values like that. Yeah, maybe. If you look up this stuff, there are examples of those sorts of things happening. Ours was mostly focused on the spam effect specifically because that was what we saw the most of all the time in our inbox. But the other stuff did happen. There were a number of, there was some Russian organized crime group that had a botnet that was stealing passwords to whatever trading accounts and buying up stocks through those zombie accounts. And all this stuff was happening, but I think the biggest effect was just from spam and normal actual people buying it. Yes, I'm sorry, can you repeat that? Of what? Gray listing, yeah, yeah, yeah. That's actually a good question. So gray listing, right? And I assume what you mean by that is when it gets reported through to these various brokers and it says like, hey, stop having people trade these stocks, right? And that happened. In fact, that happened almost every time. Every time that we went to go buy a stock, I had to call up whatever online broker I was using and tell them over the phone to let me trade that stock because they wouldn't even let me trade it through my web interface. It would say it was locked and that was an off-limit stock that they didn't wanna sell me any of. So yeah, but despite that fact, despite the fact that I had to jump through extra hoops to even touch any of these stocks, a lot of people still bought them. So that didn't really seem to have an effect. And when we talk about things like operation spam a lot there where the SEC locked down 35 of these stocks, they locked those down in March of 07, right? The spam on these stocks had stopped in January of 07. Their lock down of these, their ceasing of trading in these had no effect on anything. So take that for what it's worth. I could take one more question right there. Great question. Everyone's covering every little extra thing that I wanted to cover, which is great. He asked, did we look at other methods of hyping these stocks like forums and whatnot? And yeah, we did. We actually saw forums, people discussing these in forums as we were doing this. In fact, let me, I got one right here. There it is. So yeah, that's actually a forum in October to December of 2006 talking about exactly the stock we were looking at, GDKI. And that's people talking about, oh, this seems kind of too good to be true. And other people, there's one that's like, what is it, need a nice close today. Like, okay. So yeah, people were talking about these in a couple of those posts. It was clear that these people were, they were kind of pushing them themselves. So that effect was there too. But I think the massive amount of spam was the bigger driver of everything. So with that, that's all. Thank you.