14

Investing in stocks depends on multiple factors. In order to estimate the approximate best time can be done with experience, various calculations, etc.

With the rise of Machine Learning, Deep Learning, etc. there are numerous prediction models where anyone could easily build their own solution to predict prices and the validation data is also quite accurate.

However, I don't see many such commercially available online tools that people could use to predict upcoming values regardless of the liability of the results given. Is there any reason for this ?

Improve this question
7
  • 61
    "In order to estimate the approximate best time can be done with experience, various calculations, etc." No, it can't.
    – Dan C
    Jun 24, 2020 at 2:39
  • 4
    Obviously anyone who has such a tool is doing the opposite of whatever it predicts will happen, until it cancels out the entire prediction and there's nothing left to predict.
    – user253751
    Jun 24, 2020 at 9:59
  • 8
    "With the rise of Machine Learning, Deep Learning, etc. there are numerous prediction models where anyone could easily build their own solution to predict prices and the validation data is also quite accurate." There really isn't any reason to believe this. Deep algorithms and machine learning are good at predicting and building upon hard data - but stocks are driven by human behavior, something computers have always struggled to predict.
    – Zibbobz
    Jun 24, 2020 at 12:42
  • 1
    @Zibbobz Actually if there are patterns in human behaviour, machine learning should be able to find them just fine. The problem is what Allure said.
    – user253751
    Jun 24, 2020 at 12:44
  • 14
    You might want to read about inefficient markets, which is what a prediction tool is trying to take advantage of. The problem is, inefficiencies are eliminated once enough people all start trying to take advantage of it.
    – chepner
    Jun 24, 2020 at 14:29

8 Answers 8

Reset to default
101

If one could build a model for accurately predicting future stock prices then:

  1. Why would anyone share it?

  2. Wouldn't there be some individuals cornering the market, making people like Peter Lynch, Warren Buffet and a host of very successful hedge fund managers look like amateurs?

There are many commercially available tools such as you describe. They are quite successful at taking the money of gullible people who think that simply buying some software is the road to riches.

Improve this answer
21
  • 21
    @Bob Baerker: Yes, you can make mone with such a tool - by selling it to the gullible :-) You also have to consider the Heisenberg effect of the widespread use of such a tool: if a large fraction of investors start using, that in itself changes the behavior of the market, meaning the tool no longer works. (See "stock kiting".)
    – jamesqf
    Jun 23, 2020 at 18:02
  • 4
    @Allure Bob's answer deny that such an algorithm would be available for public use, not that they exists.
    – user3399
    Jun 24, 2020 at 9:33
  • 3
    @Allure You went down the wrong rabbit hole with your inductive logic. The first sentence merely implies that actors capable of bringing the "Commercial Product" into being, would act in a certain way. The second sentence implies that such an actor would not take the steps required to create the "Commercial Product" even though they were capable of doing so. The third piece of the puzzle, is implied, that actors incapable of creating the "Commercial Product" would not create the "Commercial Product". Therefore, no actors would create the "Commercial Product".
    – Aron
    Jun 24, 2020 at 9:57
  • 4
    @BobBaerker I get it, but the self-defeating nature of widely available stock prediction methods is notably absent from the answer. For readers who aren't aware of that, the reasons to not share the method aren't obvious. As far as software goes, it's extremely unusual to have a valuable, useful program that becomes less valuable the wider its user base. It's a rather unintuitive in many ways.
    – Nuclear Hoagie
    Jun 24, 2020 at 20:03
  • 4
    @Chris: You're obviously using a different definition of "work" than I am. A prediction tool that works, according to my definition, should do significantly better than just sticking your money in an index fund.
    – jamesqf
    Jun 25, 2020 at 17:10
59

These tools exist (see Nelson's answer), but they aren't available for purchase because once they become publicly available, they cease to work. The logic goes something like this.

  1. Suppose the stock of company A is currently trading at $100 each.
  2. The software predicts that it will go up to $120. Therefore you buy.
  3. Once you buy, you move the market. You can only buy if someone is willing to sell. In fact the price of $100 is because there's someone willing to sell the stock at $100/each, and someone is willing to buy at $100/each (this simplifies a little; usually there is an "ask" and "bid" price, where the asking price is the minimum price the seller is willing to sell for, and the bid price is the maximum price a buyer is willing to pay). Once you buy, there are fewer sellers who still own the stock + are willing to sell for $100/each. Eventually there are no more of them, and the stock price goes up.
  4. Since the stock price has gone up, you make less profits. For example if you buy at $100 and sell at $120, you make 20% profits. If you buy at $101 and sell at $120, you make less than 20%.

If the software is always accurate, and the seller also has access to that software, then they are not going to want to sell for $100. They will only sell for $120. You will not be able to profit off it.

Another way of looking at it is that, by making the prediction known, you also make the prediction no longer useful. Your predictor is great as long as it doesn't influence the system, but if you sell the predictor, then the system reacts to the predictor's predictions, and you can't make money anymore (see this question I asked on the Physics SE for another example).

Therefore, it does not make sense to sell them. If you invent one, you are better off keeping it to yourself and yourself only.

Improve this answer
16
  • 46
    @alephzero that's also known as "pump and dump" and is illegal ...
    – Allure
    Jun 24, 2020 at 11:36
  • 7
    The people who got rich during the Gold Rush were the ones who sold provisions and services to the prospectors, not so much the prospectors themselves. Unless you have lots of money to invest, you can probably make more money selling your software than trying to use it alone.
    – Barmar
    Jun 24, 2020 at 15:30
  • 10
    This is the correct answer. The system is fundamentally self-correcting. There have in fact been many successful attempts in the past to develop algorithms which perform significantly better than average at predicting the market, but as soon as they become known to other traders, they stop working.
    – BlueRaja - Danny Pflughoeft
    Jun 24, 2020 at 22:02
  • 4
    @NuclearWang they exist, and they exist at ginourmous hedge funds that are snarfing $0.01 per share 1 millisecond ahead of their peers, not spotting a 20% misprice. And no one else has the capital to do that. We pour more of our collective effort and intelligence into short term securities pricing than literally any other endeavor.
    – Jared Smith
    Jun 24, 2020 at 22:38
  • 4
    @vsz: The market can be modeled as random, but it's actually fairly predictable in specific ways. This is why pump-and-dump, insider trading, and many other schemes are so successful, and need to be enforced using laws instead. Dismissing all successful attempts at finding other ways it's predictable as pure luck is disingenuous.
    – BlueRaja - Danny Pflughoeft
    Jun 25, 2020 at 16:58
25

They exist, but it's not for you to buy.

Here is a story about Ke Xu, quantitative analyst.

Basically, the algorithms exist, but they are closely guarded secrets. Someone tried to steal them and got sued into the ground, along with travel bans, extradition, repeat jail time, private detectives tracking their parents in CHINA (Company was in UK). The legal team couldn't get the government to keep the "target" locked up, so the legal team started suing the government.

Very impressive article... a bit extreme, but if it is worth hundreds of millions of dollars, I would do the same thing.

Just to give you an idea of how much money these algorithms are worth, the analyst got £400,000 as a bonus. He got upset because he was expecting a million... and the algorithm earned his boss an undisclosed amount of millions of dollars.

Improve this answer
5
  • 3
    The impressive thing is, the guy fled from London to Hong Kong, bound for Mainland China, and the legal team in UK acted so fast, the dude got stopped in Hong Kong and extradited back to the UK. This is a PRIVATE COMPANY that did this.
    – Nelson
    Jun 24, 2020 at 7:45
  • 3
    Interesting read. The firm acted with lightning speed or he would be back in China and there would be nothing that they could do.
    – stackoverblown
    Jun 25, 2020 at 0:43
  • 4
    Had he left on a Friday, not say anything till he was back in China and then resigned by email, he would not be caught.
    – stackoverblown
    Jun 25, 2020 at 0:50
  • 1
    @stackoverblown To be honest, I think he would've been assassinated or outright kidnapped. It's not like the guy had special protection, and with enough money anything can happen in China.
    – Nelson
    Jun 25, 2020 at 8:21
  • I am not sure the boss of G-Research wants to cross that line. Otherwise some goons would have already raided and ransacked Xu's mother's flat/apartment for the laptops, memory sticks etc. Or maybe they've already done. Do we even know?
    – stackoverblown
    Jun 25, 2020 at 18:00
4
However, I don't see many such commercially available online tools that people could use to predict upcoming values regardless of the liability of the results given. Is there any reason for this ?

Because every stock investor has a simple, free tool to predict upcoming values of stocks. It's called compound interest and the formula is:

P = P0 * k^t

Where P is the future price, P0 is the initial price, t is the time and k is the compounded interest. Typical values is k = 1.08. To arrive at a more accurate value for k, one can look at dividend yield, inflation and GDP growth.

Why would I pay for a commercial tool, if I know this free tool is exactly as accurate as any offered commercial tool? That is, the tool is not very accurate in the short run, but let's say over a period of 50 years, it starts to have some value.

Improve this answer
3
  • 4
    The OP believes that stock picking and market timing work. This answer doesn't really help the OP, because it doesn't offer any evidence for its assertion that they don't work. Nor is the assertion necessarily true, e.g., there's something called the weekend effect, which some economists argue is a real bias that just doesn't get efficiently arbitraged away for small stocks. What would be more accurate and helpful would be to explain that if there were some market anomaly that was easy to exploit, it would get arbitraged away.
    – user13722
    Jun 24, 2020 at 2:55
  • 50 years? Most people don't want to wait that long.
    – stackoverblown
    Jun 25, 2020 at 18:03
  • 2
    A compound interest formula doesn't predict the price of stocks.
    – Bob Baerker
    Jun 26, 2020 at 19:49
3

As other answers points out: if this exists - it's not shared. The more widely used they are, the less valuable they become.

Additionally, nobody really knows what's going to happen with any given stock - and if they did (and acted on that information) there are other problems.

Oh, and because people like Elon Musk can tweet that Tesla is over priced and have it drop by 20%.

There's so many variables, that if someone actually worked out a model that worked, why would they want to share it (instead of benefiting directly)*? And even if they did share it - it's usefulness would diminish.

*Remember that wealth is relative. If I make $1mil, and I show everyone else how to make the same amount, now we all have an extra $1mil - thereby removing my additional financial power from having more money.

Improve this answer
2

These tools simply don't exist. Many will try to sell you them, but in reality they don't work in real life. I worked in a hedge fund that spent years making these and the reality is that the vast majority of tools simply don't make money in real life. They don't take into account multiple risk factors, trading costs, legal issues, etc etc.

Improve this answer
2

Imagine the stock market is a big chess tournament, and every time you win a chess match, you win money, and every time you lose a match, you lose money. Many humans are playing in the tournament, but some people are using software to play instead, and those with software are winning against those without. Once the software is shared, eventually everyone uses the same "winning" software, but as soon as that happens everyone just breaks even. As soon as one person gets better software, they start winning again until everyone else gets the new software too, and then everyone is just breaking even again. Unless someone keeps their special software secret, over time the end results collapse to a break even for everyone.

Moral of this analogy: Using AI for stock trading is widely in use today, and if it truly is "winning", then eventually the more people that use it, the less profitable it is compared to any set of random strategies. If you happen to have a secret winning tool or strategy, you probably won't want to share it. Therefore, if someone is trying to sell a winning tool or strategy, it probably isn't a winning tool or strategy .

Improve this answer
0

They are called Robo-Advisors and their market share is getting bigger every day.

What Is a Robo-Advisor?

Robo-advisors (also spelled robo-adviser or roboadvisor) are digital platforms that provide automated, algorithm-driven financial planning services with little to no human supervision.

[...] The industry has experienced explosive growth as a result; client assets managed by robo-advisors hit $60 billion at year-end 2015 and are projected to reach US$2 trillion by 2020 and $7 trillion worldwide by 2025.

Writing a algorithm to outperform S&P500 is possible. There are a lot of quantitative analysis communities, which are sharing their results. Still you need investors and lawyers to build a successful enterprise. Therefore the entry into the financial market is a little bit more difficult than building yet another social start-up. As the industry is growing, we can expect to see more companies entering that market and making it more popular.

Improve this answer
5
  • 3
    A robo-adviser is not a stock market prediction tool
    – AakashM
    Jun 24, 2020 at 21:31
  • @AakashM "Prediction" obviously doesn't exist as long as there is no time machine. We are talking about "projections" and also obviously Robo-Advisors have to estimate the potential growth to make a decision. OP's question wasn't specific enough if he wants to have the full details, e.g. target price, or if it's enough to select the stocks and a timeframe.
    – user27411
    Jun 26, 2020 at 7:11
  • @AakashM How can a robo-adviser recommend one possible investment (or set of investments) over another without predicting what the results of taking the two options will be?
    – David Schwartz
    Jun 26, 2020 at 15:30
  • @DavidSchwartz I'm pretty sure saying 'this equity market will typically return 8% pa over time' - or even a more complicated stochastic model - doesn't count as "a stock market prediction tool" for the purposes of this question. Now, technically such a statement is of course "a prediction about the stock market", but looking at the question body rather than just the title, we see the asker is looking for something that can "predict prices". Roboadvisers don't do that.
    – AakashM
    Jun 29, 2020 at 7:48
  • @AakashM They do that to the extent it is possible for a widely-used tool to do it. The risk/reward trade-off is fundamental.
    – David Schwartz
    Jun 29, 2020 at 17:24

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .