3 ms·

>Rather than doing the hard work of modeling fat-tailed probability distributions.... or building new financial models with more realistic assumptions... Unfor

by 1gor 18y ago

>Rather than doing the hard work of modeling fat-tailed probability distributions.... or building new financial models with more realistic assumptions... Unfortunately, you are wrong and Taleb is right. The future of a dynamical system such as market cannot be predicted any better than a weather can be predicted one month from now. You could try to 'model' the temporal structure of the market (trying to find and describe its 'strange attractor'), but this is pointless because this market system has too many dimensions and your measurements of the system are too noisy and incomplete. The best you can do is to run some 'local' predictor. Like simply build a black box model based on the system's past and predict tomorrow based on what you've seen in the past. This works quite well but only for the very short term horizon. And it is not the 'model' you are thinking about. You should dispose of the traditional statistics altogether. It simply does not work in real life and all of the kludges developed to save it (like ARMA models etc) are pointless. As far as his fund is concerned, I would think he has done right by stopping trading in times when his strategy did not work. His strategy was to buy out of money options. Well, we had almost ten years of falling volatility and he would bleed to death in such an environment. A log of smart traders also went out of the market (Soros etc), some are coming back now. Taleb is not even such an innovative thinker. All of the stuff he is talking about has been an active field of scientific study, but he is valuable for being a popularizer of these ideas.

ctkrohn 18y ago
I'm not saying that you should try to predict the future of the market. I would agree with you that in many cases, this is a fool's errand. Mathematical finance is about a lot more than prediction, however. Risk management is even more important, since there are so many market participants who find themselves exposed to risks that they must hedge. Even your neighborhood mortgage banker is exposed to interest rate risk -- both the level of rates and the volatility of rates. Any multinational corporation is exposed to foreign exchange risk. An option trading desk will often find itself exposed to volatility and correlation risks. It would be foolish to leave these unhedged, and without some sort of a model you're forced to. The right way to approach this isn't to treat markets as some inscrutable thing, but to propose the best method you can and be aware of the risks you remain exposed to.
- eugenejen 18y ago
  I don't think Taleb disagrees with this part. What he found annoyed is the modelers frequently forget what we have is a model! While in physics, any elegant theory or model will be "disapproved" by experiments. Financial modelers may just stubbornly refuse to consider the model is wrong! Unfortunately whether string theory is right or wrong may not affect out daily life. But a blow up in millions people's retirement money is in fact a tangible disaster that affect many people's life. What I get after reading NNT's books is "being humble, prepare to be humiliated" I think it also the same like what PG said for startups, "get users feedback asap, and iterate it fast" An arrogant developer or hedge fund manager is just a disaster waiting to happen.
- 1gor 18y ago
  The only true hedging is - you got an asset - you sell it forward. You are hedged. What passes for hedging and 'risk management' today in the financial industry is something different. You measure how one instrument has correlated with another instrument (or a basket) in the past build an offsetting portfolio based on these correlation matrices -- and voila! But your assumptions that those correlations between assets will stay the same in the future (including volatility etc) are baseless. As you see, it is also predicting the future. And it is wrong.
  - ctkrohn 18y ago
    It's often impossible to sell an asset forward, especially an illiquid one. Suppose you are a trading desk that makes markets in corporate bonds. If a client comes to you and wants to sell a given company's bonds, chances are that you will not be able to immediately find a buyer. You will probably be forced to hold on to the bonds for a bit. Now you're exposed to interest rate and credit risks. What do you do? You can hedge some of the interest rate risk by paying on interest rate swaps, and you can hedge some of the credit risk by buying protection in the credit default swap market. But without some kind of model, it's going to be impossible to determine how to use these tools to hedge your risks. Your only other options are to guess, or to do nothing. If you are a large player in the markets, this is extremely dangerous. Certainly, it's foolish to assume that you are ever perfectly hedged. In this example, several things could go wrong: e.g. credit default swap traders could have a different view of a company's creditworthiness than corporate bond traders, forcing you to pay exorbitant prices for protection on the bonds. But this is a smaller risk than holding the bonds outright. In short: models can be helpful in risk management, as long as you are aware of the residual risks you are exposed to. Just don't allow them to make you complacent or overconfident.

by 1gor 18y ago

ctkrohn 18y ago

I'm not saying that you should try to predict the future of the market. I would agree with you that in many cases, this is a fool's errand. Mathematical finance is about a lot more than prediction, however. Risk management is even more important, since there are so many market participants who find themselves exposed to risks that they must hedge. Even your neighborhood mortgage banker is exposed to interest rate risk -- both the level of rates and the volatility of rates. Any multinational corporation is exposed to foreign exchange risk. An option trading desk will often find itself exposed to volatility and correlation risks. It would be foolish to leave these unhedged, and without some sort of a model you're forced to. The right way to approach this isn't to treat markets as some inscrutable thing, but to propose the best method you can and be aware of the risks you remain exposed to.

eugenejen 18y ago
I don't think Taleb disagrees with this part. What he found annoyed is the modelers frequently forget what we have is a model! While in physics, any elegant theory or model will be "disapproved" by experiments. Financial modelers may just stubbornly refuse to consider the model is wrong! Unfortunately whether string theory is right or wrong may not affect out daily life. But a blow up in millions people's retirement money is in fact a tangible disaster that affect many people's life. What I get after reading NNT's books is "being humble, prepare to be humiliated" I think it also the same like what PG said for startups, "get users feedback asap, and iterate it fast" An arrogant developer or hedge fund manager is just a disaster waiting to happen.
1gor 18y ago
The only true hedging is - you got an asset - you sell it forward. You are hedged. What passes for hedging and 'risk management' today in the financial industry is something different. You measure how one instrument has correlated with another instrument (or a basket) in the past build an offsetting portfolio based on these correlation matrices -- and voila! But your assumptions that those correlations between assets will stay the same in the future (including volatility etc) are baseless. As you see, it is also predicting the future. And it is wrong.
- ctkrohn 18y ago
  It's often impossible to sell an asset forward, especially an illiquid one. Suppose you are a trading desk that makes markets in corporate bonds. If a client comes to you and wants to sell a given company's bonds, chances are that you will not be able to immediately find a buyer. You will probably be forced to hold on to the bonds for a bit. Now you're exposed to interest rate and credit risks. What do you do? You can hedge some of the interest rate risk by paying on interest rate swaps, and you can hedge some of the credit risk by buying protection in the credit default swap market. But without some kind of model, it's going to be impossible to determine how to use these tools to hedge your risks. Your only other options are to guess, or to do nothing. If you are a large player in the markets, this is extremely dangerous. Certainly, it's foolish to assume that you are ever perfectly hedged. In this example, several things could go wrong: e.g. credit default swap traders could have a different view of a company's creditworthiness than corporate bond traders, forcing you to pay exorbitant prices for protection on the bonds. But this is a smaller risk than holding the bonds outright. In short: models can be helpful in risk management, as long as you are aware of the residual risks you are exposed to. Just don't allow them to make you complacent or overconfident.