That is a better question than mine, and one of my two premises did not survive it.
You confirmed the first half: the worst window does begin part-way into a bad run, 9.05% below its recent high. The second half, that conditioning therefore produces a worse tail, is not supported by your data, and it fails across every lookback you tested. I asked for the conditional figure because I assumed it would be worse. It is not, and the unconditional four-session number is the one that matters and makes sense then.
The reframing from magnitude to concurrency is useful. What keeps less than 5% of the book from crossing the threshold on the same day needs no NAV series to ask, and a single daily mark is exactly the mechanism that makes simultaneity the default rather than the tail.
One challenge back, since you asked for them. Your 19.78% bracket applies HYG’s concentration ratio of 1.08 to the disclosed worst month, and you flag that HYG carries none of the 30% BB CLO sleeve. I would expect that sleeve to push the ratio up rather than down – mezzanine CLO in March 2020 moved later and further than cash high yield – so 1.08 reads to me as a floor for the blend rather than a point estimate. Worth testing if the CLOIE Post-BB series is obtainable.
One addition to your list of what would make this an actual HINC result, from the other proposal in this pair.
These two fail in opposite places. Here there is no downside floor, so a move on the scale of your bracket prints in full and the whole constraint sits in the liquidation machinery you have just modelled. On mWIN it is the reverse: minExpectedAnswer sits 0.8% below the $130,000 Authorised Denomination, so a move of that size cannot be reported at all. That band can only express its own worst disclosed drawdown once NAV has accrued roughly 7.6% above issue value, and only until it reaches the 150,000 ceiling.
That bears directly on your concurrency test. On mWIN the common trigger cannot fire, because the print that would move the book through its thresholds is outside the band before it gets there. Depending on which layer catches it – Horizon’s documentation describes DON-level rejection with the last valid price retained, while the Midas data feed contract reverts – the book either runs against a stale mark or the market stops. Neither state liquidates. So a seventh item, at least for the mWIN side: which layer governs a bounds breach, because it decides whether the liquidation economics are ever reached.
Thanks for committing the manifest. Appreciate the ongoing discussion here.