Sensible Use of Quant Models and Modelers
Thank you very much for helping make sense of the mess.
By innate, I meant that math and quant models now seem to be part-and-parcel of derivatives and structured products, many of which are really valued for risk management and hedging. Now, unless there is some way to turn the clock back and get rid of all the math- and quant modeling based products and services, we will need to learn to live with them and hopefully improve them by learning what doesn’t work.
Folks such as Derman, Taleb, and, Wilmott have been critical of “misuse” and “ignorant” use of models and do not seem to have asked to reject math and models altogether despite their limitations. Their point is that models did not work because most individuals using them did not use them appropriately (for example, Wilmott had been recently organizing a bootcamp to train an army of modelers who will get it right). For whatever it is worth, it seems a plausible argument and perhaps cannot be summarily dismissed without stronger evidence and / or rationale. This observation doesn’t negate your point about the relative poverty of many models (and related methods) to capture the richness of real social phenomena such as finance. The above folks who still seem to have credibility as well as others such as Mandelbrot, all seemed to have acknoweldged this point. For example, if one reads some of Derman’s decade old articles, he talks about similar limitations of models that we are discussing here.
Agreed, some of the most profound systems are based on a few simple principles. It is yet to be determined if it may be feasible to simplify the complexity of financial innovations while improving their utility. Not sure if it is feasible, however, everyone should benefit from it given the ongoing backlash against complexity of such products. This seems to be a challenge that all modelers, old and new, should aspire to handle. That being said, if we acknowledge the complexity of finance, then I am not sure if we can really create simple systems to handle such complexity: it would seem to contradict the law of requisite variety.
OK. At least you agree that being ‘legit’ requires knowledge [and probabaly accreditation] of the math and quant models probably as a ‘right of passage’. That puts this point to rest and hopefully those interested can take it for whatever it is worth. On a related note, the same observation may apply to many other fields and professions including engineering and trading as most 'real experience' is acquired in 'learning by really doing' ('strong empiricism' as Taleb would call it - also ties back to your point about empiricism).
The issue is not about proving absolute truth or falsehood which is the domain of most purely academic pursuits, not suggesting that they are not important. However, mastering the models to recognize their applied and pragmatic limitations and then ‘tinkering’ (isn't that the name of the latest book that Taleb had been planning on writing?) them as needed based on learning-by-doing and rechecking to see if they deliver the expected outcomes is what seems prudent. The focus should probably be less on the neatness of the model and more on the neatness of the outcomes within an ever-adaptive dynamic context [something like Stefen Zota seemed to suggest earlier]. I think that folks such as Derman, Taleb, and Wilmott would probably agree with the method despite its madness.
Thank you very much for helping make sense of the mess.
By innate, I meant that math and quant models now seem to be part-and-parcel of derivatives and structured products, many of which are really valued for risk management and hedging. Now, unless there is some way to turn the clock back and get rid of all the math- and quant modeling based products and services, we will need to learn to live with them and hopefully improve them by learning what doesn’t work.
Folks such as Derman, Taleb, and, Wilmott have been critical of “misuse” and “ignorant” use of models and do not seem to have asked to reject math and models altogether despite their limitations. Their point is that models did not work because most individuals using them did not use them appropriately (for example, Wilmott had been recently organizing a bootcamp to train an army of modelers who will get it right). For whatever it is worth, it seems a plausible argument and perhaps cannot be summarily dismissed without stronger evidence and / or rationale. This observation doesn’t negate your point about the relative poverty of many models (and related methods) to capture the richness of real social phenomena such as finance. The above folks who still seem to have credibility as well as others such as Mandelbrot, all seemed to have acknoweldged this point. For example, if one reads some of Derman’s decade old articles, he talks about similar limitations of models that we are discussing here.
Agreed, some of the most profound systems are based on a few simple principles. It is yet to be determined if it may be feasible to simplify the complexity of financial innovations while improving their utility. Not sure if it is feasible, however, everyone should benefit from it given the ongoing backlash against complexity of such products. This seems to be a challenge that all modelers, old and new, should aspire to handle. That being said, if we acknowledge the complexity of finance, then I am not sure if we can really create simple systems to handle such complexity: it would seem to contradict the law of requisite variety.
OK. At least you agree that being ‘legit’ requires knowledge [and probabaly accreditation] of the math and quant models probably as a ‘right of passage’. That puts this point to rest and hopefully those interested can take it for whatever it is worth. On a related note, the same observation may apply to many other fields and professions including engineering and trading as most 'real experience' is acquired in 'learning by really doing' ('strong empiricism' as Taleb would call it - also ties back to your point about empiricism).
The issue is not about proving absolute truth or falsehood which is the domain of most purely academic pursuits, not suggesting that they are not important. However, mastering the models to recognize their applied and pragmatic limitations and then ‘tinkering’ (isn't that the name of the latest book that Taleb had been planning on writing?) them as needed based on learning-by-doing and rechecking to see if they deliver the expected outcomes is what seems prudent. The focus should probably be less on the neatness of the model and more on the neatness of the outcomes within an ever-adaptive dynamic context [something like Stefen Zota seemed to suggest earlier]. I think that folks such as Derman, Taleb, and Wilmott would probably agree with the method despite its madness.
