Nothing in the output marked that number as inherited. It sat in the data looking exactly like a decision.
A $5 spread risks about the same money whatever the share price is. That is the whole problem in one sentence.
| What we had | What it should have been | |
|---|---|---|
| The width | $5, inherited from the code | a stated number in the specification |
| Trades affected | more than 100,000 | — |
| What it looked like | a deliberate $5-wide strategy | a default nobody had weighed |
That count is from our own capture data. It is not a strategy result — no test on this site has published one.
We are publishing it anyway, because the version of this site where the default was never found is the version that publishes a confident number built on it.
There were no earlier numbers. Nothing has been published. What it means is that the trades captured before 6 August described a strategy nobody had specified.
It sets the maximum loss. Delta and tenor decide where the trade sits; width and credit decide what it is worth and what it risks.
Every test names four numbers before it runs, and the width is one of them. A run that cannot state all four does not get a verdict.
No. Often the value is perfectly sensible. The bug is that nobody weighed it, and that nothing in the output said so.
The width is now part of the specification printed at the top of every test, alongside the delta, the tenor and the minimum credit.
A run that cannot name all four does not get published. That rule exists because of this.
