Thread by @bennpeifert

ok this is a good one and self contained enough to do on a recovery pton ride.

the tl;dr is that large sophisticated prop firms and hedge funds running volatility arbitrage strategies in listed options effectively look like market makers in how they execute and manage positions.

most investors (individual or retail) think of options as something you trade by pulling one up on a screen and entering a limit DMA order for.

SPX Dec22 3800 call, ok that's 40 @ 41, where do you want to bid? 40.2?

market got crushed, new price 39.2 @ 40.2, filled offerside.

sophisticated arb execution does not work like that.

sticking with the very simple example of a single option line (which generally would not be the case), order instructions to an execution system would look more like:

- buy up to 250k vega of this line
- post only (no take)
- show a randomized small bid size between 10k and 15k vega at a time
- pay no more than 28 vol
- pay no more than -0.02 vega below the real time volatility surface at any time
- hedge delta on each fill in real time

plus a bunch of other customized risk limits and safeguards. under the hood, the robots send DMA child orders to the executing broker (security, size, dollar price) and constantly cancel and replace those orders in low latency as spot and vol move

that order shows up in the market like a tightening on that line relative to the current market makers, for small size. the arb manager becomes the market maker on the bid side, better bid than C*tadel because the arb actually wants the position, not just the spread.

then over time as other market participants want to sell that option, they sell it to the arb manager, at a slightly better price than they would have gotten from the pure market makers. the arb accumulates the position slowly and passively

in general, a vol arb manager rarely has one line to execute. they might be working into a large SPX calendar position, long June23-Dec23 short Nov22-Dec22, across a wide range of strikes. the whole basket would be sized to desired long and short vega...

and working as described above, floating to buy below the surface and sell above the surface, replacing/improving the pure market makers on this whole part of the surface, in small randomized sizes.

that kind of execution would be running with tight risk limits on net delta, net tenor-adjusted vega, etc to tightly manage any legging risk. if you get lifted on a chunk of shorts, the sell orders shut down until the buys catch up, etc.

that way they work into a large complex position that they want, sourcing liquidity from wherever it comes patiently over time, and working in a balanced way that is never directionally exposed to sudden market moves

same thing with dispersion for example- might be working ten thousand line items simultaneously, buying every option on the list that someone in the market sells, and vice versa, staying tightly constrained to the index weights and vega balanced between longs ans shorts

so when ppl ask me Benn, why don't you trade options in your PA, this is why I just laugh and shrug

its like asking a Toyota exec why don't she build your own cars at home, she knows about cars, doesnt sound too hard, right

very large firms and very specialized prop firms might build every piece of this themselves. many others code their specific logic on top of third party platform like spiderrock which provides the low latency market access part as a service