Tracking The Tasks

I've introduced The Quant Innovation Spreadsheet as a kind of one view tool assessing an innovation. The posted UnRisk CM example is made concise allowing one view in one page.

But I split the Changes column into 3 sub columns: Change Values | Polarity | Turning Point

If we take the VaR Calculations task of UnRisk CM

Change Values: VaR numbers quantify the border between the known and uncertain
Polarity: VaRs alone don't say much abut the breaks
Turning Point: Back and stress tests validate VaRs (across instruments and risk factors)

Regulators often just want VaRs, but for fund management decisions it's vital to know if the number of VaR break days, their distribution…are plausible…because the methodologies may be too "optimistic/pessimistic" in certain regimes.

In general tasks in quant innovations use methodologies, models, solvers…and interpreters that need to be validated. Especially of you need to deal with ill-posed problems.

Remember, in mathematical problem solving we 1. transform a textual problem description (an instrument term sheet) into a model (a general stochastic model), 2. transform it into a model that is of good nature for reality near calculations (an interest rate PDE with calibrated parameters), 3. solve the PDE (by Finite Element techniques…), 4. interpret the results (by dynamic visualization, cross model validation…)

In this process you need to avoid quite a lot of traps and only if you've organized your objects, models, methods...orthogonally, you'll be able to detect (Polarity) and avoid (Turning Point) them. Otherwise the Change Value of the tasks is in question.

I've outlined the difficulties in Model Validation - First defense Front of Risk Management (2012)...

And tasks are the most important units of a Quant Innovation. Finding problems in them is the (my) first evaluation job.

But tracking the tasks is not enough…there are the big principle traps…like The Problems With Computing "Expected" Returns in Finance.

We talk about principles, methodologies, models…technologies, but applying them is like "writing a story". And tracking it's continuation…Point of View | Timing | Actors is indispensable.

I'll come to this in one of the next posts.