If I recall correctly, the financial insitutions were on the verge of forcing an economy shutdown by killing all liquidity. If the government would not have stepped in, within weeks the first healthy corporations would have missed payroll because they could not get any form of credit. With healthy corporations falling, he economy would soon collapse. These companies need credit because their assets were tied up in more profitable venues, exactly as the financial industy dictates. The financial industry, scared by their own failure, refused to do their job. The government decided to give free liquidity to the financial industry, begging them to do their job. Even forcing them to do their job. I think that was a mistake. The government should simply have saved the economy by stepping in, and giving out loans to corporations (forming a national bank), and let the financial industry rot until they wised up again.
Of course. The Fukushima scale is billions. However, trillions don't even start to describe the damage that will be done when sea levels rise with a single meter. Who will insure coal plants for that? Oh, they'll not be held accountable because unlike Fukushima, it will be impossible to attribute the damage? Doesn't this mean that coal and oil is also implicitly heavily subsidized?
1). Global cooling was never much of an issue in the 70s. We were shit-scared of nuclear war, not of global cooling. Ozone was picked up and mostly fixed in the 90s. Imagine that: a man made climate problem fixed with regulation. Global warming in the 2000s morphed into climate change when we realized that the models only predict massive change, not exactly the direction. So, all in all, not much in the way of scaremongering here -- it seems pretty well backed by science.
2) Why is climate change bad? Because if sea levels rise, millions if not billions of people will be displaced from where they currently live. If billions of people are forced to move into the territory of others, invasions, wars, poverty, hunger and death follow.
3) Cost of trying to counter global warming? How much did it cost to stop using gasses that killed the ozone layer? How much did it cost to stop industry polluting the air and rivers with sulfur and heavy metals? How much will it cost to install CO2 reduction technology? It doesn't sound like a very difficult thing to do. The goal is to let the free market price technology that reduces CO2 emissions highly. This can be done in at least two ways: one is to cap emissions, forcing the use of the technology and making it thereby more effective and cheaper in the long run. If the technology becomes more viable, make the cap more strict. Second is to tax it. Exactly same effect. Start with a small tax, and when the technology becomes better heighten the tax so that it becomes economically necessary to stop polluting. None of these methods need to hurt the economy, as it is not necessary to either set the cap, or the tax, initially at such levels that it makes production impossible. It just needs to get the right incentive in place. In a free market, it is impossible to introduce this technology, as the polluters will force you out of business. You need government to create a fair playing field.
Example: I am Dutch. We, the Dutch, don't like to be displaced. We happen to like our swamp (and both the Belgians and the Germans are not likely to take all 17 million of us in. Especially not the Belgians). So currently we are planning to invest billions of dollars to keep us safe. We made the calculations, we can do it, the only unsolved problem at this moment is how to pump the output of the Rhine into the North Sea when the North Sea rises a meter or so. We'll solve that as well -- maybe we elevate the Rhine or something.
My question to you is: when we do this, and it turned out to be necessary, and it was shown that it was indeed human industry that was the culprit, where do we send the bill?
Sure, and the fact that the US has been determining politics in their part of the world for the last half century or so has no bearing on the matter at all. US is pure good, and therefore can do no evil (limitless incarceration, illegal rendition, torture, death lists, blowing up antagonists). Al Qaida is pure evil, so everything is allowed.
What if I were to say that Al Qaida's methods are flawed and make them evil? What if I were to say that the US's methods are flawed and are making them evil? Where's the flaw in that logic? I prefer the US a million times over Al Qaida, though in matters of the middle East I have a hard time to call the US's actions objectively better than that of their foes. And it is getting worse.
You're at war? With whom? A couple of individuals? No really: at war with individuals? Not countries? Individuals? And you're using your army? Against individuals? That's pathetic. Al-Qaida is tiny. The mafia is much bigger, and we don't put the military on them (and the mafia does way more damage). You are giving these guys waaay too much credit, and they revel in it. If you would simply beef up local police and Interpol, you would be in a much better shape. To fight nations, you need an army. To fight individuals you need police. This is stupid.
As to your point 3. theshowmecanuck is probably right. The US is them main source of climate change denial, simply because it is well trained in scientific denial. The country is set up to make people think how you want them to think. It seems that climate change denial is even bigger in the US than evolution denial. It's funny, how the country that's doing most of the fundamental research is so willfully ignorant. The American paradox.
The paradox is I think caused because at this day and age, there are two types of people in the US. The ones just moving in, searching for freedom to do what they're good at, and the ones that are there, feeling entitled, and actually, not good for anything. The US is losing its pizzazz, probably because they are becoming just another corrupt country. Maybe a solution for the US would be the following rule: fourth generation Americans are considered to be native Americans. As tradition demands, native Americans are stripped of their possessions and put into a reservation. That defines America.
I'm probably stupid, but what I get from this is that the guy called Charles has claimed that 3 dead bears in 11% of the area roughly equates to 27 dead bears in 100% of the area, while the guy called Eric vehemently tries to claim that 7 bears in 11% of the area roughly equates to 63 bears, not 27. Eric is misunderstanding the math, Charles fails to see immediately where Eric is missing the boat. And that's about it.
Of course, when real money is on the line, people do tend to look at the data. Of course, volatility smiles are there. And yet. Nobel prizes are awarded for techniques that fail any type of empirical scrutiny. Black-Scholes and Markowitz. People talk about sigmas, betas, correlations, and greeks as if they are measurable quantities, while (some) professionals know that they're not. Students are still taught that this is a good first approximation (all models are wrong, some are useful). All this to teach the hidden assumption that you can use past price to tell anything about an instrument. When people become professionals, they might learn the ugly truth (after they've been bitten a couple of times). If they become individual investors, they're screwed, as they were taught a lie.
The situation is much like a society that is teaching astronomy by first starting with a flat earth assumption, and explain bodily movements by gods in chariots. Only when you are going to work in practice, or become a post-grad, we will tell the nasty details of a round earth, orbiting the sun. This society never got to figure out relativity.
Do you realize how many people have been taught the basics of pricing through normal assumptions, and were thereby taught something that is going to bankrupt you if you apply it? Is that sound? And are you really sure that the professional market is aware of how little is in those models? I'm sure a few know exactly what they are doing, but I've met lots of representatives of the bulk: they truly believe that fundamentally, second-order statistics are sound, but just need a bit of tuning to be used.
The point I'm trying to make here is that a sector that has so little fundamentals as the financial sector, should be very careful about how they quantify, teach, and communicate risks and events. Even to themselves. This is a sector that fools itself very easily. So of course, nobody blindly trusts Black-Scholes models, but they still run them to get a feel. And when the market does something wild, they lose a lot of money. Option traders still use them for computing the delta, to compute their risk-neutral exposure. They still go broke because of this. Nobody blindly trusts the beta of a company, but it still gets communicated in reports. In general, nobody really trusts all these multvariate statistics, but they still form a basis that people use as a guideline. It is just a pity that this guideline is thoroughly underestimating the risk people actually run.
An anecdote. Actual bank, circa 2005. They were doing a risk computation, so they applied VaR. They had to, Basel-II requirements. (my numbers are going to be off, it's too late here, I'm not going to compute them) They knew their computations led to unacceptable risk if you plug in the true numbers (as VaR invariably does), so instead of computing the value at risk at the 99.9% probability range, as was required, they computed it at the 99.9999999% range. Technically, they were trying to manage risk from a 12 sigma risk level. They were fully aware that they were running a risk much higher than this 12 sigma level, yet their instruments didn't allow them to put in the right numbers. When I watched them I had the feeling they were steering an airplane with a random altimeter, which they were constantly compensating for. Of course, they never looked out of the window.
I honestly don't know what the industry is exactly doing these days, as I said goodbye to hedge-funds around 2006. I however honestly doubt that a fundamental shift has occurred, and that the intellectual fraud of model-driven risk management has been abandoned. I do know that gross exposures are (finally) getting some attention as an indicator of risk, but I also see that the sigma myth is still alive and kicking.
The intellectual fraud lies in the empirics: the statement that second moment statistics can be reliably estimated based on data. Yes, today your models say that the event that is about to occur is so unlikely that you don't even have to consider it. It is 23 sigmas! Tomorrow, using exactly the same method, your model will tell you that your 23 sigma event is a 5 sigma event. The mathematics is just the mathematics -- what is wrong is the statement that the Gaussian (or even the t-distribution) is a good model of reality. That the Central Limit Theorem holds, and that second order moments are 'finite' (i.e., stable).
To really see where the fraud is, do the following experiment. For fun: download Nasdaq data from 1971 to now. You can get it from Yahoo finance. Load it in a spreadsheet (I'll take excel as an example). Next to the adjusted close column, add a column of log returns [log(x_t) - log(x_t-1)] of these adjusted close prices. Now we are going to look at the (common) assumption that the log returns are normally distributed. So let's try to estimate the standard deviation of the log returns. Go to the end of the spreadsheet, and in the one but last row next to the column of log returns type something like this: =STDEV(H10247:$H$10248). Now apply this to the entire column. This will give you a running standard deviation of the log returns of the entire nasdaq from beginning to end. You get 10247 standard deviations, each a different estimate of the 'historical risk' in the Nasdaq. If you plot this series, you will see jumps, you will see 1987, you see 2002, you see 2008, you see the current crisis. Zoom in to the last few months. You still see jumps. Think about this. You are seeing jumps in a standard deviation calculation where 99.99% of the data is shared between the two computations!
From this data, some people argue that, of course, events change the risk profile, but that is nonsense: the risk profile is assumed to be predicted by the standard deviation, not the other way around. The standard deviation is the risk. By definition. How can an event, if described correctly by this standard deviation, then change the standard deviation? This is magical thinking, not scientific.
What you see in a graph like this are a long series of falsifications of the normality assumption -- the assumptions that second order moments are stable in the market. We're talking about 10,000 observations, and a single data point changes the risk profile. Constantly. With 10,000 measurements, the t-distribution is equal to the normal distribution. You will see this for any series. Long series of falsifications of the normality assumption. This is well known, but ignored. That's the intellectual fraud.
Yes, you get it. There's no way to quantify this 1 time event, nor is there a way to quantify all these other 1 time events that happen all the time. There is no meaningful difference between a so-called 23 sigma event, and a 5 sigma event. The math fails in the tails, and assuming that you're safe against one time events is what is causing most of the mayem in the financial world.
So please tell me, what is the financial sector using these days, if not risk models that are ultimately dependent on a (log-) normal assumption? What fundamental shift has happened that threw away 50 years of economic theory forming? What has replaced it?
And as for smart people missing the bleeding obvious: Mandelbrot showed empirically that they were completely wrong more than 40 years ago. This was highly publicized, accepted as correct, and the smart people ignored this because Gaussian fairy land is a very convenient mathematical framework to publish in. I would be very surprised if the quants these days work on fundamentally different assumptions from the academics.
I'm arguing against people equating risk with sigma, and in particular those who say that there's a problem predicting 23 sigma moves, as if that's saying something about the likelihood of the event occuring.
I'm also arguing against the financial sector who is still using sigma (and beta, and the entire apparatus of mutlivariate statistics) as something that can be used in practice. You are arguing that the financial sector has moved on, but have they? Are they truly beyond the Gaussian assumption, have they really moved away from Value at Risk models, is Black-Scholes truly abandoned, are standard risk-of-ruin calculations abandoned, or are they still trying to fix Gaussian models with martingale and jump dynamics as they were doing last time I looked into it. What has replaced the entire apparatus that wreaked havoc 5 years ago? I still see bailouts on the horizon, banks very highly leveraged, and a sector that has reasoned away risk using the same flawed arguments that have been en vogue for the past 40-50 years.
You are seriously arguing that this has changed? I've worked at hedge funds, I am working in risk management, the assumption of normality is the only game in town. And yes, I am arguing that all those extremely smart mathematicians and physicists are missing the obvious, simply because they are doing what they were hired for: to create models. Stating that we don't have the mathematics to do the job will not land you a position as a quant. They'll hire someone that will try the best they can: steer a car by rearview mirror, driving in the ravine at the next haircut. All we do is drive a little less fast.
I guess that on the trade floor there might be a few that get it and are using algorithms that do assume that they can lose everything at any time and for instance only trade in options, but at the macro level, capital requirements are still stated in risk weighted assets. Options are still being sold. Individuals and banks are still going naked (leveraged) long. Risk is still assumed to be log-normal. Textbooks are still printed stating this as a fact (not as an assumption). Have you seen a shift in teaching economics, away from multivariate stats? No, as nothing has changed.
You do realize that a 23 sigma event implies the belief in a risk model where an event like this will happen once every 10^112 years?!! You also do realize that the entire financial sector is using these risk models, and are therefore still assuming that 23 sigma events will only happen once every googol lifetimes of the universe? A hedge like this, that will only fail once every googol universes as predicted by economics (and is thus a safe bet on the surface), tends to fail every few years.
Sigma measurements of risk is intellectual fraud, on a scale that is costing us billions.
That was because in the pre-9/11 time, hijackings were for ransoms. If you sit down and shut up during a ransom, chances are pretty good that everyone survives. If you start shooting, chances are that people, including innocents die. When you mistake a suicide mission for a ransom, everyone dies. Doesn't make the original tactic wrong, just outdated. By a single day.
You can use common sense as evidence against the free market, because a 100% unregulated free market is going to be a nightmare. If you believe otherwise, you're a moron, and the Earth being 6,000 years old must sound like a plausible alternative to you. That's the level of idiocy we're talking about here. Markets need to be regulated. We tried unregulated markets. That was not a good idea. People die because of it.
Ah okay. Yes... maintaining other people's C++ code. Especially that of someone who just figured out how static variables could work together with operator overloading and templates. Shudder. You're right, Java is a godsend for that. I've seen some pretty brutal Java code too, but if you can't toss it, you can usually wrap it and hope it would not make too much of a mess later on. Uncontrolled C++ is a whole different beast. Don't let amateurs code, but.... no choice.
Are you truly claiming that figuring out what a button called "one click shopping" does is equivalent in complexity with reverse engineering windows 7 from a title bar? Are you serious?
Most of the stuff you mention on how to figure out if how these things work are either plainly illegal, borderline illegal, or only illegal if you use it to copy other peoples work for commercial gain. Do you really think you can chemically analyze Coca-Cola and bring an equivalent competitor on the market claiming you broke the (unpatented) Coca-cola recipe? Try it and see how far you will get.
Ah well, after using java for a decade or so, I still go back to C++ now and then to feel a breadth of fresh air. Code where you actually don't pay for what you don't use; no per object overhead when you don't need it; no garbage collector that needs to be tuned independently from the program; no runtime environment + jit that optimizes away the errors of newbies, not the needs of the proficient. No magic, just a program that you can understand.
Hmm, so if I say that this little pill here helps against erectile dysfunction, everyone will immediately invent Viagra? To figure out what's in the pill, you need to chemically analyze it. To figure out what's makes a machine produce 3d objects, you will have to take it apart. To figure out what's in JPEG, you have to reverse engineer it. To figure out what goes on behind Amazon's button, you look at the title on the button.
Slashdot Mirror
Few year, and they'll have robots. They're practicing in the Middle East now. We're next.
If I recall correctly, the financial insitutions were on the verge of forcing an economy shutdown by killing all liquidity. If the government would not have stepped in, within weeks the first healthy corporations would have missed payroll because they could not get any form of credit. With healthy corporations falling, he economy would soon collapse. These companies need credit because their assets were tied up in more profitable venues, exactly as the financial industy dictates. The financial industry, scared by their own failure, refused to do their job. The government decided to give free liquidity to the financial industry, begging them to do their job. Even forcing them to do their job. I think that was a mistake. The government should simply have saved the economy by stepping in, and giving out loans to corporations (forming a national bank), and let the financial industry rot until they wised up again.
A corporation is a non-physical entity, and is therefore not taxed. Its owners are. They have representation. That's about all that matters.
Of course. The Fukushima scale is billions. However, trillions don't even start to describe the damage that will be done when sea levels rise with a single meter. Who will insure coal plants for that? Oh, they'll not be held accountable because unlike Fukushima, it will be impossible to attribute the damage? Doesn't this mean that coal and oil is also implicitly heavily subsidized?
1). Global cooling was never much of an issue in the 70s. We were shit-scared of nuclear war, not of global cooling. Ozone was picked up and mostly fixed in the 90s. Imagine that: a man made climate problem fixed with regulation. Global warming in the 2000s morphed into climate change when we realized that the models only predict massive change, not exactly the direction. So, all in all, not much in the way of scaremongering here -- it seems pretty well backed by science.
2) Why is climate change bad? Because if sea levels rise, millions if not billions of people will be displaced from where they currently live. If billions of people are forced to move into the territory of others, invasions, wars, poverty, hunger and death follow.
3) Cost of trying to counter global warming? How much did it cost to stop using gasses that killed the ozone layer? How much did it cost to stop industry polluting the air and rivers with sulfur and heavy metals? How much will it cost to install CO2 reduction technology? It doesn't sound like a very difficult thing to do. The goal is to let the free market price technology that reduces CO2 emissions highly. This can be done in at least two ways: one is to cap emissions, forcing the use of the technology and making it thereby more effective and cheaper in the long run. If the technology becomes more viable, make the cap more strict. Second is to tax it. Exactly same effect. Start with a small tax, and when the technology becomes better heighten the tax so that it becomes economically necessary to stop polluting. None of these methods need to hurt the economy, as it is not necessary to either set the cap, or the tax, initially at such levels that it makes production impossible. It just needs to get the right incentive in place. In a free market, it is impossible to introduce this technology, as the polluters will force you out of business. You need government to create a fair playing field.
Example: I am Dutch. We, the Dutch, don't like to be displaced. We happen to like our swamp (and both the Belgians and the Germans are not likely to take all 17 million of us in. Especially not the Belgians). So currently we are planning to invest billions of dollars to keep us safe. We made the calculations, we can do it, the only unsolved problem at this moment is how to pump the output of the Rhine into the North Sea when the North Sea rises a meter or so. We'll solve that as well -- maybe we elevate the Rhine or something. My question to you is: when we do this, and it turned out to be necessary, and it was shown that it was indeed human industry that was the culprit, where do we send the bill?
Sure, and the fact that the US has been determining politics in their part of the world for the last half century or so has no bearing on the matter at all. US is pure good, and therefore can do no evil (limitless incarceration, illegal rendition, torture, death lists, blowing up antagonists). Al Qaida is pure evil, so everything is allowed.
What if I were to say that Al Qaida's methods are flawed and make them evil? What if I were to say that the US's methods are flawed and are making them evil? Where's the flaw in that logic? I prefer the US a million times over Al Qaida, though in matters of the middle East I have a hard time to call the US's actions objectively better than that of their foes. And it is getting worse.
You're at war? With whom? A couple of individuals? No really: at war with individuals? Not countries? Individuals? And you're using your army? Against individuals? That's pathetic. Al-Qaida is tiny. The mafia is much bigger, and we don't put the military on them (and the mafia does way more damage). You are giving these guys waaay too much credit, and they revel in it. If you would simply beef up local police and Interpol, you would be in a much better shape. To fight nations, you need an army. To fight individuals you need police. This is stupid.
Who solved the problem that the parameters for fat-tailed models cannot be estimated from data?
As to your point 3. theshowmecanuck is probably right. The US is them main source of climate change denial, simply because it is well trained in scientific denial. The country is set up to make people think how you want them to think. It seems that climate change denial is even bigger in the US than evolution denial. It's funny, how the country that's doing most of the fundamental research is so willfully ignorant. The American paradox.
The paradox is I think caused because at this day and age, there are two types of people in the US. The ones just moving in, searching for freedom to do what they're good at, and the ones that are there, feeling entitled, and actually, not good for anything. The US is losing its pizzazz, probably because they are becoming just another corrupt country. Maybe a solution for the US would be the following rule: fourth generation Americans are considered to be native Americans. As tradition demands, native Americans are stripped of their possessions and put into a reservation. That defines America.
I'm probably stupid, but what I get from this is that the guy called Charles has claimed that 3 dead bears in 11% of the area roughly equates to 27 dead bears in 100% of the area, while the guy called Eric vehemently tries to claim that 7 bears in 11% of the area roughly equates to 63 bears, not 27. Eric is misunderstanding the math, Charles fails to see immediately where Eric is missing the boat. And that's about it.
Of course, when real money is on the line, people do tend to look at the data. Of course, volatility smiles are there. And yet. Nobel prizes are awarded for techniques that fail any type of empirical scrutiny. Black-Scholes and Markowitz. People talk about sigmas, betas, correlations, and greeks as if they are measurable quantities, while (some) professionals know that they're not. Students are still taught that this is a good first approximation (all models are wrong, some are useful). All this to teach the hidden assumption that you can use past price to tell anything about an instrument. When people become professionals, they might learn the ugly truth (after they've been bitten a couple of times). If they become individual investors, they're screwed, as they were taught a lie.
The situation is much like a society that is teaching astronomy by first starting with a flat earth assumption, and explain bodily movements by gods in chariots. Only when you are going to work in practice, or become a post-grad, we will tell the nasty details of a round earth, orbiting the sun. This society never got to figure out relativity.
Do you realize how many people have been taught the basics of pricing through normal assumptions, and were thereby taught something that is going to bankrupt you if you apply it? Is that sound? And are you really sure that the professional market is aware of how little is in those models? I'm sure a few know exactly what they are doing, but I've met lots of representatives of the bulk: they truly believe that fundamentally, second-order statistics are sound, but just need a bit of tuning to be used.
The point I'm trying to make here is that a sector that has so little fundamentals as the financial sector, should be very careful about how they quantify, teach, and communicate risks and events. Even to themselves. This is a sector that fools itself very easily. So of course, nobody blindly trusts Black-Scholes models, but they still run them to get a feel. And when the market does something wild, they lose a lot of money. Option traders still use them for computing the delta, to compute their risk-neutral exposure. They still go broke because of this. Nobody blindly trusts the beta of a company, but it still gets communicated in reports. In general, nobody really trusts all these multvariate statistics, but they still form a basis that people use as a guideline. It is just a pity that this guideline is thoroughly underestimating the risk people actually run.
An anecdote. Actual bank, circa 2005. They were doing a risk computation, so they applied VaR. They had to, Basel-II requirements. (my numbers are going to be off, it's too late here, I'm not going to compute them) They knew their computations led to unacceptable risk if you plug in the true numbers (as VaR invariably does), so instead of computing the value at risk at the 99.9% probability range, as was required, they computed it at the 99.9999999% range. Technically, they were trying to manage risk from a 12 sigma risk level. They were fully aware that they were running a risk much higher than this 12 sigma level, yet their instruments didn't allow them to put in the right numbers. When I watched them I had the feeling they were steering an airplane with a random altimeter, which they were constantly compensating for. Of course, they never looked out of the window.
I honestly don't know what the industry is exactly doing these days, as I said goodbye to hedge-funds around 2006. I however honestly doubt that a fundamental shift has occurred, and that the intellectual fraud of model-driven risk management has been abandoned. I do know that gross exposures are (finally) getting some attention as an indicator of risk, but I also see that the sigma myth is still alive and kicking.
The intellectual fraud lies in the empirics: the statement that second moment statistics can be reliably estimated based on data. Yes, today your models say that the event that is about to occur is so unlikely that you don't even have to consider it. It is 23 sigmas! Tomorrow, using exactly the same method, your model will tell you that your 23 sigma event is a 5 sigma event. The mathematics is just the mathematics -- what is wrong is the statement that the Gaussian (or even the t-distribution) is a good model of reality. That the Central Limit Theorem holds, and that second order moments are 'finite' (i.e., stable).
To really see where the fraud is, do the following experiment. For fun: download Nasdaq data from 1971 to now. You can get it from Yahoo finance. Load it in a spreadsheet (I'll take excel as an example). Next to the adjusted close column, add a column of log returns [log(x_t) - log(x_t-1)] of these adjusted close prices. Now we are going to look at the (common) assumption that the log returns are normally distributed. So let's try to estimate the standard deviation of the log returns. Go to the end of the spreadsheet, and in the one but last row next to the column of log returns type something like this: =STDEV(H10247:$H$10248). Now apply this to the entire column. This will give you a running standard deviation of the log returns of the entire nasdaq from beginning to end. You get 10247 standard deviations, each a different estimate of the 'historical risk' in the Nasdaq. If you plot this series, you will see jumps, you will see 1987, you see 2002, you see 2008, you see the current crisis. Zoom in to the last few months. You still see jumps. Think about this. You are seeing jumps in a standard deviation calculation where 99.99% of the data is shared between the two computations!
From this data, some people argue that, of course, events change the risk profile, but that is nonsense: the risk profile is assumed to be predicted by the standard deviation, not the other way around. The standard deviation is the risk. By definition. How can an event, if described correctly by this standard deviation, then change the standard deviation? This is magical thinking, not scientific. What you see in a graph like this are a long series of falsifications of the normality assumption -- the assumptions that second order moments are stable in the market. We're talking about 10,000 observations, and a single data point changes the risk profile. Constantly. With 10,000 measurements, the t-distribution is equal to the normal distribution. You will see this for any series. Long series of falsifications of the normality assumption. This is well known, but ignored. That's the intellectual fraud.
Yes, you get it. There's no way to quantify this 1 time event, nor is there a way to quantify all these other 1 time events that happen all the time. There is no meaningful difference between a so-called 23 sigma event, and a 5 sigma event. The math fails in the tails, and assuming that you're safe against one time events is what is causing most of the mayem in the financial world.
So please tell me, what is the financial sector using these days, if not risk models that are ultimately dependent on a (log-) normal assumption? What fundamental shift has happened that threw away 50 years of economic theory forming? What has replaced it? And as for smart people missing the bleeding obvious: Mandelbrot showed empirically that they were completely wrong more than 40 years ago. This was highly publicized, accepted as correct, and the smart people ignored this because Gaussian fairy land is a very convenient mathematical framework to publish in. I would be very surprised if the quants these days work on fundamentally different assumptions from the academics.
I'm arguing against people equating risk with sigma, and in particular those who say that there's a problem predicting 23 sigma moves, as if that's saying something about the likelihood of the event occuring. I'm also arguing against the financial sector who is still using sigma (and beta, and the entire apparatus of mutlivariate statistics) as something that can be used in practice. You are arguing that the financial sector has moved on, but have they? Are they truly beyond the Gaussian assumption, have they really moved away from Value at Risk models, is Black-Scholes truly abandoned, are standard risk-of-ruin calculations abandoned, or are they still trying to fix Gaussian models with martingale and jump dynamics as they were doing last time I looked into it. What has replaced the entire apparatus that wreaked havoc 5 years ago? I still see bailouts on the horizon, banks very highly leveraged, and a sector that has reasoned away risk using the same flawed arguments that have been en vogue for the past 40-50 years.
You are seriously arguing that this has changed? I've worked at hedge funds, I am working in risk management, the assumption of normality is the only game in town. And yes, I am arguing that all those extremely smart mathematicians and physicists are missing the obvious, simply because they are doing what they were hired for: to create models. Stating that we don't have the mathematics to do the job will not land you a position as a quant. They'll hire someone that will try the best they can: steer a car by rearview mirror, driving in the ravine at the next haircut. All we do is drive a little less fast.
I guess that on the trade floor there might be a few that get it and are using algorithms that do assume that they can lose everything at any time and for instance only trade in options, but at the macro level, capital requirements are still stated in risk weighted assets. Options are still being sold. Individuals and banks are still going naked (leveraged) long. Risk is still assumed to be log-normal. Textbooks are still printed stating this as a fact (not as an assumption). Have you seen a shift in teaching economics, away from multivariate stats? No, as nothing has changed.
You do realize that a 23 sigma event implies the belief in a risk model where an event like this will happen once every 10^112 years?!! You also do realize that the entire financial sector is using these risk models, and are therefore still assuming that 23 sigma events will only happen once every googol lifetimes of the universe? A hedge like this, that will only fail once every googol universes as predicted by economics (and is thus a safe bet on the surface), tends to fail every few years.
Sigma measurements of risk is intellectual fraud, on a scale that is costing us billions.
That was because in the pre-9/11 time, hijackings were for ransoms. If you sit down and shut up during a ransom, chances are pretty good that everyone survives. If you start shooting, chances are that people, including innocents die. When you mistake a suicide mission for a ransom, everyone dies. Doesn't make the original tactic wrong, just outdated. By a single day.
You can use common sense as evidence against the free market, because a 100% unregulated free market is going to be a nightmare. If you believe otherwise, you're a moron, and the Earth being 6,000 years old must sound like a plausible alternative to you. That's the level of idiocy we're talking about here. Markets need to be regulated. We tried unregulated markets. That was not a good idea. People die because of it.
Oracle and Sun are both internationals.
Better is: Facebook users should always keep in mind that they are the consumer. Not the customer.
Yes. In short: Free speech should be free.
Ah okay. Yes ... maintaining other people's C++ code. Especially that of someone who just figured out how static variables could work together with operator overloading and templates. Shudder. You're right, Java is a godsend for that. I've seen some pretty brutal Java code too, but if you can't toss it, you can usually wrap it and hope it would not make too much of a mess later on. Uncontrolled C++ is a whole different beast. Don't let amateurs code, but.... no choice.
Are you truly claiming that figuring out what a button called "one click shopping" does is equivalent in complexity with reverse engineering windows 7 from a title bar? Are you serious? Most of the stuff you mention on how to figure out if how these things work are either plainly illegal, borderline illegal, or only illegal if you use it to copy other peoples work for commercial gain. Do you really think you can chemically analyze Coca-Cola and bring an equivalent competitor on the market claiming you broke the (unpatented) Coca-cola recipe? Try it and see how far you will get.
Ah well, after using java for a decade or so, I still go back to C++ now and then to feel a breadth of fresh air. Code where you actually don't pay for what you don't use; no per object overhead when you don't need it; no garbage collector that needs to be tuned independently from the program; no runtime environment + jit that optimizes away the errors of newbies, not the needs of the proficient. No magic, just a program that you can understand.
Hmm, so if I say that this little pill here helps against erectile dysfunction, everyone will immediately invent Viagra? To figure out what's in the pill, you need to chemically analyze it. To figure out what's makes a machine produce 3d objects, you will have to take it apart. To figure out what's in JPEG, you have to reverse engineer it. To figure out what goes on behind Amazon's button, you look at the title on the button.