Curriculum·R405 Strategy Development and Validation·about 33 min
The pipeline
By the end of this lesson you can
- →Name the six stages a strategy passes through and what each one is allowed to conclude
- →Explain why the party being measured must not control the measurement
- →Compute what a model change is worth to the person whose limit it sets
- →Identify the stage at which most strategies should be killed and usually are not
Senior · enrolled learners
This lesson opens with JPMorgan's Chief Investment Office, 2012.
- What happened
- JPMorgan's Chief Investment Office held a large synthetic credit portfolio managed in London. In January 2012 a new Value at Risk model was introduced for the office, replacing the previous one and roughly halving the reported risk figure, which allowed the position to grow without breaching limits. The model was implemented as a series of Excel spreadsheets completed manually, with data copied and pasted between them. It contained an error in which, after subtracting the old rate from the new rate, the spreadsheet divided by their sum rather than by their average, an error with the effect of understating volatility by a factor of about two and therefore lowering the reported Value at Risk. The bank stopped using the model in 2012 once the errors were found. The loss on the portfolio reached about $6.2B.
- The decision point
- The model that set the limit on the position was revised, implemented and operated by people whose activity that limit constrained, and the revision moved the number in the direction that permitted more of what they were already doing. Whether the error was deliberate is not the useful question. The pipeline had no stage at which an independent party had to reproduce the number before it was relied upon, and without that stage an error and a preference are indistinguishable from the outside.
- Recorded loss
- $6,200,000,000
What you will be able to answer
- →What are the six stages?
- →What is the independence rule?
- →What is a model change worth?
- →Where should most strategies die?
Orientation and Year One are open: anyone can read them without an account. From Year Two onward the lessons are for enrolled learners, because progress through the later years only means anything if it is tracked against a record.
It is free. We do not sell the list and there is nothing to buy at the end of it.
Sources and review
- https://blog.revolutionanalytics.com/2013/02/did-an-excel-error-bring-down-the-london-whale.html
- https://www.accountingweb.co.uk/tech/excel/excel-errors-jp-morgans-whale-sized-classic
- https://www.thekeycuts.com/dear-analyst-38-breaking-down-an-excel-error-that-led-to-six-billion-loss-at-jpmorgan-chase/
- https://web.mit.edu/Alo/www/Papers/august07.pdf
Confidence medium·Volatility low·Reviewed 2026-08-07·Owner unassigned
Contested
The loss is reported at about $6.2B and the spreadsheet error is described in the bank's own task force report and in subsequent official reviews. Whether the model revision was motivated by the limit it relaxed was examined at length and this lesson does not assert intent, per P6. The structural point holds under either reading.
Value at Risk is one risk measure among several and its limitations are well documented independently of this incident. This lesson uses it as an example of a measurement controlled by the measured party rather than as an endorsement or a criticism of the measure.
R411-01 owns turning rules into code and R411-03 owns staged deployment for automated systems. This lesson owns the stages and the independence requirement. Keep the split.