good lord...did you really confuse probability of strike touch and a no touch option? https://optionalpha.com/lessons/probability-of-profit-vs-probability-of-touch#:~:text=Probability of touch is the,not stay at that level.
Thanks for the explanation, but I didn't get this part. Why does retail software use yesterday's close in calculating IV and greeks? It only needs to know the underlying spot price and time-to-expiration.
There are others: rates, if the stock is ex-div, etc. but any modern production system should be able to handle this.
I guess by "modern production system" you didn't mean anything used by retail. I think the rates, dividend and hard-to-borrow fees could be inferred from the options market prices and shouldn't need to be specified?
Let's make an example with 30 day realized vol. You take the log return of yesterdays close vs todays close of the last 30 days. You square them, sum them up, divide them by 30 and take the square root. That gives you the annualized realized vol for the 30 day bucket...an average of the log returns of the last 30 days. Now do that for a 1day realized vol. You only have one single data point. There is no averaging. It only measures the log return from yesterdays close vs the intraday price. It doesn't account for intraday volatility which doesn't make sense at all for a one day option. So to go back to the example. Would you trade intraday when the only datapoint you have is yesterdays close? No 5min charts. How do you know if the day is gonna be slow or trending when all you have is yesterdays close? With 1 day options, how do you know if the volatility is too high or too low when you don't know how realized vol from the open was compared to lunchbreak vol? Or if vol is mispriced because mid day realized vol was low but macro data is coming at 2pm? I'm using 525600/minutes till expiration for these to capture the intraday vol squings, others are using seconds but for that you really need acurate data and loads of computing power. On top of that, how do you calculate weighted vega for one day options? Because if you trade them, you probably want to hedge with longer dated ones. How do you hedge delta when gamma figures are so insanely high? You will get issues because your delta hedger is not fast enough which causes something called oscillation risk. People are spending a lot of time on this problem. http://www.topquants.nl/wordpress/w...5/01/Van-Gulik-Risk-management-at-Optiver.pdf
That makes sense. Thanks! Intraday data is the prerequisite for calculating intraday realized vol. Although I had a different understanding about the 30 day realized vol you mentioned. I thought it should be based on the return of today's close and the close of 30 days ago (averaged over multiple samples). The calculation you described would actually be 1 day as the returns are based on dates that are 1 day apart.
Perhaps we're talking about the same: You want to look at the log return of close(-30->30 days ago) and close(-29), then the close(-29d) and close(-28), then close (-28) and close(-27), and so on and so forth until you arrive at log(close(-1)) - log(close(0->today)). You square them up, divide them by 30 (because you have 30 datapoints) and take the square root. That's the 30 day volatility. https://www.wallstreetmojo.com/realized-volatility/ Think about it this way: If you sell a call that has 30days to maturity, strike 100 and IV of 44 with the stock at 100$, you will have a delta of -52$ and a gamma of 3$ while receiving a 500$ premium. Meaning if the stock trades down 1$, you will have a delta of -49$. If you hedge at the initial sale, you will need to buy 52 shares. Stock down 1$, you sell another 3 shares. Next day, stock trades up 1$, you will have to buy 3 shares to be delta neutral...and you have a 3$ loss. If you hedge your delta once per day, you will have a series of losses by buying high and selling low. On the other hand you received 500$ premium. So now you are interested in how much you can lose on average per day through your delta hedging until the 500$ option premium is eaten up. That's why you calculate the average daily log return and NOT the log return from today minus 30 days ago.