wilmott.com

QuoteOriginally posted by: CuchulainnQuoteOriginally posted by: Traden4AlphaThe nonlinearities in finance are worse than those in engineering because finance is nonlinear at BOTH ends of the range. Whilst huge transactions suffer market impact nonlinearities, tiny transactions suffer high relative administrative costs.But - in contrast to Hooke's law - do we know the operating ranges within which these models are correct? (i.e. give expected results).The linear range of Hooke's law isn't that simple because the linearity range is a very strong function of:1. spring geometry: springs with wide loops of thin wire have a much much larger operating range than narrow loops of thick wire2. material properties: IIRC, the linear range is a near linear function of the yield strength of the material divided by the Young's modulus of the materialBetween the two factors, there's probably a greater than 100:1 ratio in the linear range of Hooke's law across the range of common spring designs and materials.As for finance, isn't the linearity range a strong function of liquidity with both ends of the range being worse as liquidity drops to zero and the cost of finding a counterparty climbs (making small-transaction end much worse) and the market impact of transacting in size also much worse. Worse, isn't the market liquidity a function of people's propensities to buy and sell and those propensities are driven by a number of factors shared by all participants (e.g., perceptions that the price of X can only up, deleveraging dynamics, financial crises, etc.) Thus, the financial markets go nonlinear when they are most needed. Imagine if we had springs that obeyed Hooke's Law as long as we really didn't the spring. Under such conditions, a prudent engineer would never use Hooke's law.

QuoteThus, the financial markets go nonlinear when they are most needed.IMHO Forget black swans and the Taleb thread - this is the truth of modern finance in a nutshell.

QuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: CuchulainnQuoteOriginally posted by: Traden4AlphaThe nonlinearities in finance are worse than those in engineering because finance is nonlinear at BOTH ends of the range. Whilst huge transactions suffer market impact nonlinearities, tiny transactions suffer high relative administrative costs.But - in contrast to Hooke's law - do we know the operating ranges within which these models are correct? (i.e. give expected results).The linear range of Hooke's law isn't that simple because the linearity range is a very strong function of:1. spring geometry: springs with wide loops of thin wire have a much much larger operating range than narrow loops of thick wire2. material properties: IIRC, the linear range is a near linear function of the yield strength of the material divided by the Young's modulus of the materialBetween the two factors, there's probably a greater than 100:1 ratio in the linear range of Hooke's law across the range of common spring designs and materials.As for finance, isn't the linearity range a strong function of liquidity with both ends of the range being worse as liquidity drops to zero and the cost of finding a counterparty climbs (making small-transaction end much worse) and the market impact of transacting in size also much worse. Worse, isn't the market liquidity a function of people's propensities to buy and sell and those propensities are driven by a number of factors shared by all participants (e.g., perceptions that the price of X can only up, deleveraging dynamics, financial crises, etc.) Thus, the financial markets go nonlinear when they are most needed. Imagine if we had springs that obeyed Hooke's Law as long as we really didn't the spring. Under such conditions, a prudent engineer would never use Hooke's law.You can simulate the properties of 1 and 2 before manufacturing it, yes? Because of known material properties, FEM etc.

QuoteOriginally posted by: CuchulainnQuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: CuchulainnQuoteOriginally posted by: Traden4AlphaThe nonlinearities in finance are worse than those in engineering because finance is nonlinear at BOTH ends of the range. Whilst huge transactions suffer market impact nonlinearities, tiny transactions suffer high relative administrative costs.But - in contrast to Hooke's law - do we know the operating ranges within which these models are correct? (i.e. give expected results).The linear range of Hooke's law isn't that simple because the linearity range is a very strong function of:1. spring geometry: springs with wide loops of thin wire have a much much larger operating range than narrow loops of thick wire2. material properties: IIRC, the linear range is a near linear function of the yield strength of the material divided by the Young's modulus of the materialBetween the two factors, there's probably a greater than 100:1 ratio in the linear range of Hooke's law across the range of common spring designs and materials.As for finance, isn't the linearity range a strong function of liquidity with both ends of the range being worse as liquidity drops to zero and the cost of finding a counterparty climbs (making small-transaction end much worse) and the market impact of transacting in size also much worse. Worse, isn't the market liquidity a function of people's propensities to buy and sell and those propensities are driven by a number of factors shared by all participants (e.g., perceptions that the price of X can only up, deleveraging dynamics, financial crises, etc.) Thus, the financial markets go nonlinear when they are most needed. Imagine if we had springs that obeyed Hooke's Law as long as we really didn't the spring. Under such conditions, a prudent engineer would never use Hooke's law.You can simulate the properties of 1 and 2 before manufacturing it, yes? Because of known material properties, FEM etc.Quite so! Now whether the actual material meets the spec in the sim and whether the actual manufacturing technique delivers the specified dimensions is another issue. And in that regard, it's not unlike the problem of underwriting mortgages -- woe unto the financial system that does not inspect for quality!Yet here we see another ubernasty nonlinearity that bedevils money manufacturers but not spring manufacturers. For the most part, springs operate independently. No failure of any N springs will increase the probability of failure or the N+1 spring. But financial instruments are coupled to each other by price in the market, regulatory issues, and by counterparty relationships. If N instruments fail, the probability of the N+1 instrument failing will be higher.

QuoteOriginally posted by: CuchulainnQuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: CuchulainnQuoteOriginally posted by: Traden4AlphaThe nonlinearities in finance are worse than those in engineering because finance is nonlinear at BOTH ends of the range. Whilst huge transactions suffer market impact nonlinearities, tiny transactions suffer high relative administrative costs.But - in contrast to Hooke's law - do we know the operating ranges within which these models are correct? (i.e. give expected results).The linear range of Hooke's law isn't that simple because the linearity range is a very strong function of:1. spring geometry: springs with wide loops of thin wire have a much much larger operating range than narrow loops of thick wire2. material properties: IIRC, the linear range is a near linear function of the yield strength of the material divided by the Young's modulus of the materialBetween the two factors, there's probably a greater than 100:1 ratio in the linear range of Hooke's law across the range of common spring designs and materials.As for finance, isn't the linearity range a strong function of liquidity with both ends of the range being worse as liquidity drops to zero and the cost of finding a counterparty climbs (making small-transaction end much worse) and the market impact of transacting in size also much worse. Worse, isn't the market liquidity a function of people's propensities to buy and sell and those propensities are driven by a number of factors shared by all participants (e.g., perceptions that the price of X can only up, deleveraging dynamics, financial crises, etc.) Thus, the financial markets go nonlinear when they are most needed. Imagine if we had springs that obeyed Hooke's Law as long as we really didn't the spring. Under such conditions, a prudent engineer would never use Hooke's law.You can simulate the properties of 1 and 2 before manufacturing it, yes? Because of known material properties, FEM etc.You never know the real materail properties in advance ... if you can't derive them from process behaviour (even the process changes material behavioer), FEM is "useless" ... the BlaNk Swan recalibration approach is correct for most of the material dependent predictions.Matereal specs are in the range of up to 15% ... to overcome this uncertainty you need the "Blank Swan of metal treatment" (we make real money from in-process recalibration approaches) ...T4A is correct in general ... Numbersix in quant finance.

QuoteOriginally posted by: CuchulainnLast night was a program pension funds. Paul, the journnalist has a question regarding section 12.7 (see @23.00 minutes on clip). The manager did not know the answer!Zwarte zwanen (black swans)The graph at 21.00 shows how reality and prediction differ. These guys took 24K measurements based on the period 1916-2004. Instead of once in 300K years predicted event, 64 occurrences actualised in that period.So NNT was correct, it seems. // I blame popular TV science programmes for promulgating the "once in a million years" myth applied to everything.

That's not really the biggest problem with pension funds. The biggest problem is that the poor sods who hope to get a pension out of them have no clue about what is going on with their money.

QuoteOriginally posted by: CuchulainnQuoteOriginally posted by: CuchulainnLast night was a program pension funds. Paul, the journnalist has a question regarding section 12.7 (see @23.00 minutes on clip). The manager did not know the answer!Zwarte zwanen (black swans)The graph at 21.00 shows how reality and prediction differ. These guys took 24K measurements based on the period 1916-2004. Instead of once in 300K years predicted event, 64 occurrences actualised in that period.So NNT was correct, it seems. // I blame popular TV science programmes for promulgating the "once in a million years" myth applied to everything.What's the Dutch for 'Mind projection fallacy'?