The only way you can consistently make money is by knowing things others do not (e.g., insider trading) or by taking advantage of structural problems or inefficiencies in the trading platform (e.g., high frequency trading).
I do believe that insider trading is rampant. I've seen many occasions where a large corporate announcement sends a stock up or down, but hours before, you can see the price of the stock slowly slide in the direction. Obviously a certain amount of that may be random/explainable, but I've seen it a lot and my intuition says more is going on. This also seems relatively easy to quantitatively measure, I'd love to see a real analysis of it.
I find the market for knowledge which others do not have but which aren't actually illegal to be quite fascinating.
There is a company who will break down a brand new car into its component parts and by knowledge of materials and production processes can tell to a very narrow margin what the cost of producing that car is - and therefore the profit potential of each sale. The information about the number of cars sold is either public or more easily obtained than the actual company accounts (public in many jurisdictions due to car registrations).
There are other companies who will buy a bunch of iPhones for example and by carefully analysing the serial numbers can deduce to within a reasonable margin exactly how many have been sold.
Using information like this I'm sure most large trading firms have very good models of the likely performance of each publicly listed company.
I wonder how far down this rabbit hole things go. Do they have people standing outside factories counting how many cars leave on the back of the truck? Do they have access to systems like Galileo in the travel industry where they could figure out how many seats had been booked on a plane, or how many rooms were booked in a hotel? They can then take a good guess at staffing / fixed costs and arrive very much in the correct ball park.
> There are other companies who will buy a bunch of iPhones for example and by carefully analysing the serial numbers can deduce to within a reasonable margin exactly how many have been sold.
Except it only works when serial numbers are sequential, I believe the common practice today is to have serial numbers to be random alphanumeric strings.
Not usually, and especially not for physical products. An 8 digit alphanumeric serial contains 2.82E13 possible values (which is more than any one company will sell of anything, ever), and it'd be crazy not to use that to store some kind of information like models, series, bins, etc. Two characters allow almost 1300 possibilities, and just three give over 46,000. Our POs have 6 digits, and after over two decades of multiple weekly deliveries, we're still in the 500000s.
To make that random would be a huge waste and also destroy the potential for creating visible patterns helpful within a lot of industries.
Edit: I just thought about redemption codes, which yes, do aim to be randomized have a huge space relative to the issued codes. In those cases, I see what you mean. Serials of digital media are often and intentionally randomized.
Bought 3 identical Sony cameras recently 1 a month ago and a pair this week, serial numbers are numerical and the most recent pair are two digits apart.
In the movie "Wall Street" (1987) Gordon Gecko has his lackey Bud Fox follow a rival around to try and figure out what he's going to invest in, so they can undercut him.
He follows him to a private jet and finds out it's flying to mid-sized city that happens to be the corporate headquarters of a particular company. In the movie it's sort of implied that this activity is illegal.
I wonder if it actually was, and if so, why that would be different than what those hedge fund guys you reference were doing?
A company I interviewed at was finding people who lived near power plants and paying them to set up cameras. The cameras would look at the smoke/steam coming out of the plant and based on that and the type and age of the plant they were estimating the power being generated. They'd then sell that data to energy traders.
>> Do they have people standing outside factories counting how many cars leave on the back of the truck?
Yes. They do. I took a car ride with someone a few years ago who claimed to work for a consulting firm that the big banks+others hire to gather on the ground intel from companies through things like factory tours and shooting the sh*t with employees at bars near the warehouses etc.
The problem with that is that it could be insider information (or, at least, it isn't obviously not insider information, the way info about the serial numbers from iPhones that anyone could buy is public information).
The gist I've read in the past is that basically as long as it's not obtained illegally (e.g. breaking in and stealing) and there's no benefit to the person providing the information, it's not insider trading.
Matt Levine writes on the topic pretty regularly. Here's an example with a number of interesting situations.
This is famously the Eddie Murphy rule - it was a crime to steal the briefcase, but not to trade on the knowledge gained from the contents of said briefcase.
iirc 'Trading Places' is on netflix, and Planet Money[0] did an entertaining episode about this.
I've never thought about that before, but it would certainly make for an interesting news story!
"Meals on Wheels is accused of receiving insider information about big market movements from dozens of investment bankers. The Meals on Wheels endowment has seen 10x growth in the past 18 months. It looks like the group they provide for will be fed forever, even if we went through another Great Depression-level financial catastrophe. Should they be allowed to keep these funds? It certainly looks bad for the government to fine them hundreds of millions of dollars with the recent cuts in government-provided social programs."
I wonder if the contract with the firm is setup so that the information they gather is public, if anyone can find the application in the back of the drawer in the basement if you show up in person and request the form from the admin who is only there 2 hours a week.
For something to be insider trading, one of the people involved has to have breached their fiduciary responsibilities. Low-level employees typically don't have any fiduciary responsibilities to their company, so the firm that collects information from them is under no obligation to make it public.
I'm not a lawyer, but this definition can't possibly be correct, can it? If you're a janitor and you come across a piece of paper or a conversation between bigwigs discussing a merger and sell it to a trader, the trader and the janitor can surely both be sent up the river for insider trading, right?
>I wonder how far down this rabbit hole things go. Do they have people standing outside factories counting how many cars leave on the back of the truck? Do they have people standing outside factories counting how many cars leave on the back of the truck?
Mostly as an exercise for myself, I've been curious to make a system that would do this. Cameras on all the entrances, license plate tracking, and maintaining an "inventory" of vehicles in the facility.
You could track shift change, number of consumer vehicles (estimating number of people working), number of semis, etc.
Throw things into mechanical turk and you could categorize semis. Maybe that factory that makes big things needs one big component for every unit, and it's super obvious when it comes in on a semi. Boom, that's going to be a strongly correlated factor to production.
Consumer vehicles versus true production over time could help find increases in efficiency.
There are tons of ways to leverage little bits of data, and I think a vehicle in/out tracker would get you a LOT of information.
Funny enough, in the last couple of days I've been wondering whether you could use computer vision in combination with sports betting to generate more accurate odds given the current state of the game.
Of course the problem with that is getting up to the second video feeds (as apposed to delayed video feeds from SKY etc) from a fixed view point. I'm sure you could pay some money to get essentially a security camera feed at the game but that would be expensive.
Which made me wonder whether someone could turn up the night before a big game, park a drone on the roof broadcasting the video to a van with the computers processing the results while evading detection...
>I've been wondering whether you could use computer vision in combination with sports betting to generate more accurate odds given the current state of the game.
Totally unfamiliar with the gambling side of things. Do they set odds at the start of the game and maintain them throughout? There are already companies who will do live "coding" of game events to create a stream of data about the game. Players on the field, time on the field, location of the ball, location of the players, who has possession, possession by player.
There are some other trackers out there targeted for coaches. E.g. track player running speed. If they are going 10% slower than typical after the start of a play in American football, it's a good sign that they need to come out.
The common problem with your idea and mine are getting high enough quality for relatively modest prices. A standard dash cam isn't enough to grab a license plate, it'd have to be at least somewhat better tuned. Drone video from a distance is going to be tough to use to discern who is who and to get valuable info.
I haven't seen this far, but one summer I joined a friend working for a commodity trading firm. We went into corn fields, ripped off a couple ears and made some counts to help them estimate yield. it was definitely trespassing, one time he got caught and got an "earful" from the farmer.
> I do believe that insider trading is rampant. I've seen many occasions where a large corporate announcement sends a stock up or down, but hours before, you can see the price of the stock slowly slide in the direction.
I can tell you from professional experience that is most often not insider trading. You might be surprised at the amount of meaningful information that can be legally obtained through research.
I used to (very legally) find and analyze data that would lead to better conclusions than street analysts. The results of equity earnings are basically an open secret among institutional investors before they're announced.
Yeah, along the same lines, the actual definition of insider trading is a lot narrower than people think, there's interactions that would be insider trading if it came from the CEO, but is not insider trading if it comes from a guy who overheard the CEO on the phone in a coffee shop or something. Multiply that by thousands of employees with "sensitive" information having unguarded conversations that they're perfectly allowed to have, and there's a lot of information out there.
Could also be confirmation bias. If you see the stock going up before a good news you notice it and think this is insider dealing. If you see the stock going down you don’t notice it because it is just a random movement.
One trick insiders use is when they know a stock will go down, they invest in competitors within the sector. They don’t have the same reporting requirements for those shares and are much less likely to get caught.
Your two ways fall a bit short. The goal of markets is to integrate any kind of information into an assets price which is relevant for its value (so called price discovery). So the ideal way to make money would be to discover some new source of information, like for example the data from social media (which was pretty new some years ago). Another example: Some decades ago quant funds started using weather data for trading commodity futures, etc.
I used to visit fuckedcompany.com to short sell companies based on information I found there. It worked for awhile until that site became too popular. That was a long time back, 1999 or so.
Which get's into the short vs long term problem. In the short term strategy X works, in the long term it failed. How much of that success is random chance and how much is the value of the underlying data? And how do you avoid losing all benefit when the strategy eventually fails?
The reality is this kind of short term incite based investing does not really scale. Over time the money you spend on paying people to do this is less than your risk adjusted long term benefit.
"Knowing things others do not" can be perfectly legal, because a lot of times the crucial knowledge is recognizing something as important among all the chaff.
There was somebody who recently demonstrated that you can derive useful information from simply reading SEC filings, which orthodox theorists would presume should already be reflected in stock prices.
I do believe that insider trading is rampant. I've seen many occasions where a large corporate announcement sends a stock up or down, but hours before, you can see the price of the stock slowly slide in the direction. Obviously a certain amount of that may be random/explainable, but I've seen it a lot and my intuition says more is going on. This also seems relatively easy to quantitatively measure, I'd love to see a real analysis of it.