CBOE Variance Futures

quite truthfully..i watch the oi, vol etc in the other new cfe curves , somehow i never looked at VA...promising oi it seems.

of course i cannot get ib to quote the curve yet. if someone knows the correct symbol i'm interested.

sle, guys like you mostly trade these against another product? btw, i did note no feb oi.

my interest in these will be to see if they trade differently than vix futures.
 
Quote from sle:

Lets start at what is a variance swap and how is it "GOOD".

Variance is a square of volatility. Historical variance can be calculated from daily closes using the following formula:

252 * Sum( 1 to N) { log(SPX[t]/SPX[t-1])/(N-1)

So, it's just an average of daily logarithmic returns that is annualized. To be anal and detail-oriented, variance for variance swap purposes is calculated from the close of the strike date (in case of futures, it is the listing date) to the SOQ print on the expiration date.

So far so good? Well, a variance swap pays the difference between the realized variance (as calculated above) and the variance strike:

payoff = variance units * (realized variance - variance strike)

To make things a bit easier in terms of quotation, usually variance strike is quoted as a volatility number, so it's a square root of variance strike:

K = sqrt(variance strike)

and variance units are calculated from "vega notional" and the variance strike as:

variance units = 1/(2*K)

So, if you have sold a 100k vega notional of a 1m variance swap struck at 15.57 (that's today VIX close, we will give there in my next post) and S&P realized 12% volatility, your P&L will be

-100000 * (.12^2 - .1557^2)/(2 * .1557) = 3160.723

Bingo. Not as simle as buying and selling stocks, but simple enough. More about variance and VIX in the next installment.

I did not read carefully, but it strikes me that maybe something is wrong in the formulas (I looked only at the first one).

In variance formula, shouldn't you square the returns, and center them if using N-1, or divide by N instead of N-1 if no centering?
 
i havent looked on IBTWS yet, but i see it on bbg. any idea why it shows the jan settlement at 764.16 (27.64 vol strike) but shows quotes at 16.15/16.45 25k x 5k?

when would it even have traded with a 27.64 strike if it was listed dec 10?
 
Quote from Jgills:

i havent looked on IBTWS yet, but i see it on bbg. any idea why it shows the jan settlement at 764.16 (27.64 vol strike) but shows quotes at 16.15/16.45 25k x 5k?

when would it even have traded with a 27.64 strike if it was listed dec 10?

Lean on the bid!
 
Quote from tradingjournals:

I did not read carefully, but it strikes me that maybe something is wrong in the formulas (I looked only at the first one).

In variance formula, shouldn't you square the returns, and center them if using N-1, or divide by N instead of N-1 if no centering?
You are right, for some reason (formating?) it deleted my square rooting (lets see - ^2). Thank you for pointing out! Can a moderator correct this please, it's been too long?

PS. I also see that a multiplier is missing from my variance P&L calculation, but that I think is my error.
 
Quote from Jgills:

i havent looked on IBTWS yet, but i see it on bbg. any idea why it shows the jan settlement at 764.16 (27.64 vol strike) but shows quotes at 16.15/16.45 25k x 5k?

when would it even have traded with a 27.64 strike if it was listed dec 10?
No, initial strike is specified by the exchange using a variance calculation, I think (have not looked) it would be with 19 handle

edit - yes, initial strike is printed as 396.41, so square root of that is 19.9. i'll get to how these things are quoted in the next post. For now, think that, for practical purpose, they are quoted as fresh variance swap and settled (EOD) as seasoned variance swap vs the original strike.
 
so what this value just floats on speculation .... its striked at a certain date as you specified then as variance starts to be realized it moves around.. wow i need to read more.. seems like your sort of trading against a snap shot of the implieds.. "strike"
 
First of all, thank you all for participating. Also, if you see any errors, please let me know right away (thank you, tj!). Obviously, I try to be as correct as I can, but I am in bed with fever and my attention span is rather low.

Anyway, I promised that I'd post a simple short variance strategy (the "homework answer") but I figured I'd go through how the futures are quoted and settled first, given that it is super-confusing at first.

The way this works:

* day one, when these things are listed, the exchange specifies an initial strike, K0 using a variance swap calculation (as I described above). So, for example, Jan13 futures were listed on Aug 20th, 2012 (they back-dated the calc based on the options listing) and the initial variance strike is 396.41, which is 19.9^2.

* now, next day comes and the market makes make the fair variance at 16.75 @ 17. This is spot starting variance. Now, lets assume that you hit 16.75 bid in $100,000 vega notional, so your Vt = 16.75. You settlement will be determined as follows. First, some important inputs:

DF = a discount factor provided by the exchange, given the level of OIS rates, it's equal to 1 more or less

N = also known as N-effective, a number of business days between the contract listing and contract expiration.

n = number of business days elapsed on which variance was observed

K0 = initial variance strike as specified by the exchange (e.g. in our case of Jan13 futures, 396.41)

X = interest accrued on margin. again, given the interest rates, you can safely assume it is going to be somewhere around zero

First, we convert the price into "effective variance strike", call it Kt

Kt = [ Vt^2 * (N-1-n)/252 + 10000 * Sum(1 to n) log(Si/Si-1)^2 ] * 252/(N-1)

What did we do here? We combined current implied variance from now to expiry with already realized variance and annualized the result. So, now, we can calculate the "price" of the variance futures as

Futures "Price" = DF * (Kt - K0) - X + 1000

I will provide a numerical example in a sec.
 
I will take a theoretical numerical example, but something close to reality:

Lets assume that
-- you hit the Jan 13 futures today for 16.75 in 100k vega
-- since Aug 20th, S&P500 volatility has been about 0.65% per day

If you do the legwork, you will see that
-- N = 106
-- n = 70
For simplicity, we assume discount and accrued value as zero

realized portion = 70*10000*(0.65%^2) = 29.575

implied portion = (16.75^2)*(106-1-70)/252 = 38.9670

Kt = (realized portion + implied portion)*252/(106-1) = 165.50

Now, all thats left to do is to calculate the futures value:

futures settlement = (165.50 - 396.41) + 1000 = 768.1

CBOE, in their never-ceasing kindness, provides daily calculation of discount factor, realized variance, etc, so you don't have to deal with it and just plug in the numbers.
 
Back
Top