So I pulled bars for the last 1239 whole trading days at a 1-minute resolution. For each of the 390 minutes in the regular session, I looked at the price change (close minus open) and calculated the standard deviation of that minute for all of the days which essentially gives me a "volatility" value for each minute of the day: Now, everyone already could guess that the open and close are more volatile than mid-day. At 5, 10, and 15 minutes before close there are spikes from EOD position management and things like Reg-T cutoff being applied at 10 minutes. There was something unexpected however: spikes at every 30 minute interval on the hour and the half-hour. The first few are the easiest to see. The REALLY big one is at 11:00 AM PST / 2:00 PM EST and its magnitude is similar to the first couple minutes the market is open. What is causing it though? I suspect automated processes. I ran this study because I considered selling M/W/F weekly options on SPY on the final trading day (as I have watched 'live theta decay' throughout the day), and I wanted to know what time periods during the day would be optimal to do it. I am also going to examine option decay on the final day as a function of time of day, and risk of further upward movement (calls) based on the price movement since the open (theory is that the second half of the day tends to have reversals from the first half, so if it's gone up $2, its less likely than before to move up another $2, and the intraday time correlation tends to be negative).
Nice. Only thing about 2pm spikes though.... it's probably FOMC announcement causing the volatility for the most part.
Finally a worthwhile post. thanks for sharing. Your 11 am bar worked like a charm today. very volatile. what was the median and the mode? also wondering what kind of outliers you could throw out of the data like the guy said 2 pm is probably fomc. 11 am could be because of the Dept of energy report which causes more volatility yes i know it doesnt come out at that time. good data for sure. thanks
Here is some more I ran last night. First the mean of the upside (that minute to EOD price change) by time of day, last 60 days analysis (it needs to be divided by 60 to get the actual but the chart shows the relative values): Standard deviation of the same (essentially realized intraday volatility to the close). Much more significant than the average which is only slightly positive say 0.03-0.10 because of the general trend direction. It drops around the same time from around the 0.7 range to the 0.6 range so this may be a good entry point for option plays. It does not quite fit the "idealized" option theta decay during the last day, as I have seen the actual option theta ATM drop during the day at a fairly constant rate for a good part of the day, but this reveals that the realized volatility doesn't drop through much of the day (about 8:15 PST to 11:30 PST has the same upside risk for selling a call - but earlier it'll have more premium).
%% 1st part of theory maybe more useful/long lasting than $2 move theory; ExCEPT on TREND DAYS. Especially since an ETF split + rising price makes a fixed $2 amount not long lasting/practical. Thanks.............................................................................................
That's why I tested various "if it move up X, what is the further upward risk?" distribution. Long tail risk can blow it up. I don't have much skill for market timing, so I usually calculate mechanical strategies in order to better define the risk profile. I am pretty skeptical of "income producing options strategies" that I read about. All I can say is I have worse chance of predicting the future than rolling dice.