I assume they are doing it consistent with the OTC variance products and market disruption event is the day when S&P fails to publish (e.g. when trading was stopped for more then 20% of S&P 500 companies by weight). Those days are excluded. Not really a concern, unless you expect september 11- like events to occur frequently.
Does anyone have any question on the simple strategy sheet? I was going to go on to talk about convexity and some other features of variance, but do not want to start unless every is on board (whoever wants to be on board).
I honestly don't remember, it's in the vega notional field though - I think it's 1k vega notional every month.
That makes some sense but you only lose 8 vols in oct 08?when I did the analysis a few years ago 2008 wiped out like 10years on pnl. I think bac/merril had an index for it.
8 vols? No, more like 85 vols - 1k of vega notional and losing 85 thousand. Worst draw down is a 110 vols or something around that, also in 08, which is worth 5 or so years of accrued P&L. Filtering trades based on prior realized variance vs current implied (the right panel) improves the drawdowns a lot, but it is true for regular gamma sellers too. Merril index is for quarterly expiration and it does "delete" a bigger amount of P&L accrued, but you also make more in the good times (curve is meaningfully steeper because of the term premium). I will export the variance history for all tenors today and post more results.
Many thanks for posting the spreadsheet and for the whole thread. In nthe spreadsheet, I think you convert the VIX number to a Var strike, I guess b/c VIX is calculated in calendar days. You use the formula: =C40/SQRT((D40*365)/((B40-A40)*252)) whereC40 is VIX, D40 is business days and B40-A40 the difference in calendar days. Not sure if it is the other way round? D40*252 etc
A fair amount but all in OTC - we have not tried the variance swap futures yet. The % varies with the opportunities, obviously, I'd say from 5% to 25%. EDIT: Sorry, misread you - you are asking about gamma swaps. You know, through my entire life as a vol trader, I was asked to price gamma swaps like ten times and I have traded them maybe once. The discount from the regular variance is not significant enough, so funds prefer conditional variance, early term variance or KI fwd var.
You want variance strike to be higher if the business day year fraction is lower then the calendar day year fraction. So, you want to divide variance by the business day year fraction and multiply by the calendar day year fraction: x = [BD/ 252] / [CD/ 365] = [BD * 365] / [CD * 252] var = vix / sqrt(x) makes sense?