40 ms·

The general thrust is spot on. Models are approximations. Don't confuse your models with reality. Know what your assumptions are, and have a plan for what to

The general thrust is spot on. Models are approximations. Don't confuse your models with reality. Know what your assumptions are, and have a plan for what to do when those assumptions no longer hold. Know why your models work, so that you know when they are likely to stop working. Get deep domain knowledge. Know that the market can stay irrational longer than you can stay solvent. Experiment ruthlessly. Apply the scientific method. Start simple and iterate. This probably applies to more than just finance.

scotty79 14y ago
Basic error is that you think you can predict fully random process. You develop a model that seems to work excellent until it doesn't. It doesn't matter whether it's sophisticated algorithm running on big iron or hunch running on wetware. You will fail because nobody knows the future. Trying to reliably squeeze money out of the market by better predicting its behavior is like trying to build perpetum mobile by trying more and more convoluted designs. All the smarts of traders ... they do nothing: http://finance.fortune.cnn.com/2013/01/24/buffett-hedge-fund-bet/ http://finance.fortune.cnn.com/2013/01/24/buffett-hedge-fund... There is money that can be reliably extracted from investing though. If you can trade faster than others, you can drain some money from them. That's why algorithmic trading yields returns. Your strength comes not from predicting future better, but from acting and reacting faster.
- niggler 14y ago
  "If you can trade faster than others, you can drain some resources from them" What's embarrassing is how many intellectual resources are focused on this problem. For those not in finance, compare it to focusing talent on finding better advertising techniques.
- crntaylor 14y ago
  Certainly you need to protect against over fitting your models, and you need to consider tail events. But to say that the entire market is fully random is to claim that the efficient market hypothesis holds, which it clearly doesn't - not even in weak form. Anyone who wants to claim that the EMH is true needs to explain the equity premium puzzle, the success of value investing and the persistent outperformance of momentum investing - among many other price anomalies that I could name.
  - lutusp 14y ago
    > But to say that the entire market is fully random is to claim that the efficient market hypothesis holds, which it clearly doesn't - not even in weak form. Of course it does. All evidence supports the Efficient Market Hypothesis (EMH), and note my use of the word "evidence". A contradiction would have to be a series of funds that applied a given, predefined (internally) strategy and (very important) didn't reveal its picks to the public (to avoid the announcement effect). Such an experiment hasn't been done, and those cases in which a specific strategy was applied have been universal failures. Let's say for the sake of argument that there is a surefire winning strategy as is so often claimed in worthless investment books. If it were to be recognized and then applied, it would become the new market norm, everyone would apply it, and its differential value would evaporate overnight. That hypothetical experiment merely supports the EMH. > Anyone who wants to claim that the EMH is true needs to explain the equity premium puzzle, the success of value investing and the persistent outperformance of momentum investing - among many other price anomalies that I could name. Those result from a combination of the announcement effect and insider trading. Berkshire-Hathaway is successful because people expect it to be successful, and who invest in the same stocks when they're announced. It's a self-fulfilling prophecy. The only fair, scientific test would be a strategy that is applied in secret, not in public like Berkshire-Hathaway, and in a way that doesn't tweak the market itself by way of large purchases that create their own following of fanboys. In other words, a pure computer model that is defined in advance, not tweaked enroute, that doesn't actually invest but only models hypothetical investments, and that shows its value over other systems in perpetuity. There is no such thing, and the EMH is true. Remember the Null Hypothesis, the basis of modern scientific thinking. The Null Hypothesis assumes something without evidence is false. All evidence supports the EMH.
    - yarou 14y ago
      Only in academic circles does EMH stay true. Time and time again, reality has shown that EMH is not applicable, or exists in a very weak form. One need only look to the LBO bubble and the dotcom bubble to see counter-examples of EMH.
      - 14y ago

scotty79 14y ago

Basic error is that you think you can predict fully random process. You develop a model that seems to work excellent until it doesn't. It doesn't matter whether it's sophisticated algorithm running on big iron or hunch running on wetware. You will fail because nobody knows the future. Trying to reliably squeeze money out of the market by better predicting its behavior is like trying to build perpetum mobile by trying more and more convoluted designs. All the smarts of traders ... they do nothing: http://finance.fortune.cnn.com/2013/01/24/buffett-hedge-fund-bet/ http://finance.fortune.cnn.com/2013/01/24/buffett-hedge-fund... There is money that can be reliably extracted from investing though. If you can trade faster than others, you can drain some money from them. That's why algorithmic trading yields returns. Your strength comes not from predicting future better, but from acting and reacting faster.

niggler 14y ago
"If you can trade faster than others, you can drain some resources from them" What's embarrassing is how many intellectual resources are focused on this problem. For those not in finance, compare it to focusing talent on finding better advertising techniques.
crntaylor 14y ago
Certainly you need to protect against over fitting your models, and you need to consider tail events. But to say that the entire market is fully random is to claim that the efficient market hypothesis holds, which it clearly doesn't - not even in weak form. Anyone who wants to claim that the EMH is true needs to explain the equity premium puzzle, the success of value investing and the persistent outperformance of momentum investing - among many other price anomalies that I could name.
- lutusp 14y ago
  > But to say that the entire market is fully random is to claim that the efficient market hypothesis holds, which it clearly doesn't - not even in weak form. Of course it does. All evidence supports the Efficient Market Hypothesis (EMH), and note my use of the word "evidence". A contradiction would have to be a series of funds that applied a given, predefined (internally) strategy and (very important) didn't reveal its picks to the public (to avoid the announcement effect). Such an experiment hasn't been done, and those cases in which a specific strategy was applied have been universal failures. Let's say for the sake of argument that there is a surefire winning strategy as is so often claimed in worthless investment books. If it were to be recognized and then applied, it would become the new market norm, everyone would apply it, and its differential value would evaporate overnight. That hypothetical experiment merely supports the EMH. > Anyone who wants to claim that the EMH is true needs to explain the equity premium puzzle, the success of value investing and the persistent outperformance of momentum investing - among many other price anomalies that I could name. Those result from a combination of the announcement effect and insider trading. Berkshire-Hathaway is successful because people expect it to be successful, and who invest in the same stocks when they're announced. It's a self-fulfilling prophecy. The only fair, scientific test would be a strategy that is applied in secret, not in public like Berkshire-Hathaway, and in a way that doesn't tweak the market itself by way of large purchases that create their own following of fanboys. In other words, a pure computer model that is defined in advance, not tweaked enroute, that doesn't actually invest but only models hypothetical investments, and that shows its value over other systems in perpetuity. There is no such thing, and the EMH is true. Remember the Null Hypothesis, the basis of modern scientific thinking. The Null Hypothesis assumes something without evidence is false. All evidence supports the EMH.
  - yarou 14y ago
    Only in academic circles does EMH stay true. Time and time again, reality has shown that EMH is not applicable, or exists in a very weak form. One need only look to the LBO bubble and the dotcom bubble to see counter-examples of EMH.
    - 14y ago