Anyone else find it bewildering and inexplicable SPX intraday PnL fluctuation due to wide b/a spreads throwing off vols, skewing marks in the process, rendering PnL like an epileptic on a pogo stick? Using TOS inherent bad IV model to drive greeks, mentally compensating for difference btwn what should be and what is actually happening to PnL flying the plane via the risk curve (screenshots attached) and finger-in-the-wind method of greeks PnL attribution Homegrown vol surface data mining/warehousing/scripting tools, mostly trying to lean/tilt T+0 slightly to where I think the mkt is heading (skating to where the puck is going) while trapping it in a range under the expiry tent, managing delta/theta/vega exposures via ratios and skew/term structure shifts. Being right directionally, having the "right" greeks profile doesn't often translate into corresponding expected PnL, w/o greeks PnL attribution/decomposition can't exactly identify which drivers (spot, time, vol) are affecting the sensitivities. Sound off, questions/comments/snide remarks welcome. P&L Attribution Greeks P&L = [Δ * δS] + [½ Γ * (δS)2]+ [θ * δt] + [σ * δIV] Delta P&L Gamma P&L Theta P&L Vega P&L
My 2 cents: TOS IV developed for the masses, not for us engineering types that have affinity for correctness and precision. When using TOS, base your information only off the BID/ASK and derive everything else yourself and your PnL tracks correctly. -- An easier process is to alter your trading such that TOS is "close enough"! -- While I do my own calcs from BID/ASK, it is not a trivial undertaking.
Look at the range of today's action. Look as SPX plotted against VIX or VIX9D. Had nothing to do with ToS, and everything to do with market using the FED to jimmy past resistance. If we don't get any more Tweets overnight, expect the VIX to *drop* tomorrow -- over a full point, counting today's -0.28[??]
I did not find the timestamp for the data relating to the SPX Skew sheets, so unable to comment on your IV numbers! (the plots imply the IV may not contain the primary TOS flaws) -- I'm assuming you have a process for deriving that column (IV) from the BID/ASK of the complete chain!
Seems like spreads always widen around big events. Maybe set your models to calc with mid prices? Trying to tame Vol/IV jumps is like making a deal with the devil...
this is just regular run of the mill avg day, non-event IV phenomena.. my main point is that w/o greeks PnL attribution it's difficult to isolate PnL drivers by spot, vol & time - there're 4 drivers, not just price alone
I must be blind, you did provide the timestamp! Your computed IV #'s are much lower than mine! -- Not saying yours are wrong, as I focus on longer DTE (7 days to 100 days) typically, and have not spent time verifying the very low DTE derivations! For example, my ATM IV for that timestamp (assuming you are on Pacific time) is:
Can you comment on: I am using a proprietary process which requires TOS to be running, which requires some hand-holding/babysitting of TOS as TOS cannot run for weeks without some TLC. Curious of the API you reference and how it interacts to extract the option prices. -- Am guessing it is NOT thru TDA-API, as that has gone from mediocre to unusable this year! -- A penny for your thoughts.
am working w/ a couple coders/devs who have full-time server pumping in TD API data bypassing Excel altogether, storing it in SQL/PostgresDB and running some Python/Matlab/R analytics on it. www.hanweck.com is an institutional grade product, am not the programmer myself so I work together w/ techs who assist in the below schematic