https://investors.atlassian.com/fin...Quarter-Fiscal-Year-2020-Results/default.aspx gross. That one is going to sting. I was originally positioned long the Oct11/18 switched out to above position after seeing how this comany clusters with other stocks. It might not be too bad actually. I should be able to sneak out with a .25 loss at 9.30
This question is coming from a place of curiosity rather than judgement - what made you take such a large position in TEAM cals? Were you betting on earnings being announced on the week of 25th? In which case, the trade would have served you very well. Also, I see that you have liquidated your call cals - are you planning on holding onto the put cals? PS - agreed, you were totally the only volume yesterday on this strike. But now that the earnings date has been announced, expect a lot more volume, so hopefully liquidity and spreads will become better.
abs 4 week EPS change vs abs jump. Big indicator here guys. I am currently taking a sneak peak on 3 equities with 60 earnings dates. The relationship is clear as mud. In theory, the more efficient the market is, the smaller the move should be. If we see a large absolute change in the estimate, it means the market is incorporating new information into the estimates. @oldmonk any chance you can post data on this? I have to manually do this on BBG.
Here is a link to Francis Diebold lecture on Connectedness. Stay tuned to the end to see quite an amazing graph. He uses realized intraday volatility to determine connectedness. Has anyone done something like this? Creating a dynamic graph (instead of his static full sample one) would be so helpful in vol trading. It would make relative value trades much easier to find! https://www.youtube.com/watch?v=TMOrEJrSpno&t=0s
Yes. You have. You used a variation of Margrabe's formula for vol to measure directional info flow (directional connectedness) between very highly correlated assets. IIRC, you concluded that connectedness went generally from south to north. Measuring directional information flows (what Diebold calls to/from connectedness) is pretty widely dealt with in the literature over the past 25 years. There are many ways of measuring it. Hasbrouck uses an eigen-decompostion method. Diebold prefers VAR or more specifically an MA representation of a VAR (requires the assumption that the underlying process be covar-stationary, which might be problematic as calling the vol process such is a bit of a stretch). And for assets with active options, you're already familiar with the Margrabe decomposition. I've also ween a communality-based (principal axes) method. And probably other methods also, which I can't recall at the moment. The variance decomp matrix shown in the video first appeared, AFAIK, in Belsley et al. "Regression Diagnostics" (1980) and was specifically referenced in the stats.stackexhange link on the ridge penalty I posted in this thread a few days ago. And in fact was one of the main reasons I posted it. In the whuber answer to that question. A good rule of thumb is that if you come across a whuber post, read it all the way through. At least twice. In this instance he suggests a variance decomposition grouping of columns allowing you to, e.g. use an L2 penalty for within group coeffs, and perhaps an L1 penalty for between group coeffs (sort of like group lasso or elastic net but with more control and tuned to your X matrix). The concordance between the rolling connectedness peak and the Lehman blowup is meaningless. The rolling window should have been centered (and symmetrically weighted), putting the real peak 3+ months earlier. Edit: this type of short-term estimation of directional info-flow is commonly used in higher frequency trading to estimate instantaneous (and constantly shifting) lead-lag relationships among traded assets.
Hi TBS Wow yes it looks like a big indicator. So I guess you're looking for the big EPS changes which leads to smaller jumps. I'm guessing the market sees the profitability of the company (+ve or -ve Current EPS), so when the official Earnings time arrives there's little surprise in the market (... is my thinking correct as I know little about EPS?). But what is the "abs 4 week EPS change"? Is it the 4 week absolute change in "Current EPS" estimates leading up the official Earnings announcement? ... and where does one get data for this ... the Quandl Zacks data looks a little pricy? I'm catching up as I've been distracted getting a business off the ground.
Some food for thought for you guys as I study VAR/VECM and time varying volatility. Check to see if the past earnings move could have been a structural shift in the earnings distribution. I'll use STMP as an example, now that I understand what a structural shift is we have actually seen this everywhere (KHC, FL, BBBY etc..)! Here are the historical Close to Close move for STMP. Prior to the extremely large move, the probability of a move that size was extremely small, however, once that move occurred it clearly demonstrated a shift in the earnings distribution. The next 2 earnings moves were also massive relative to historical moves. This shows how the earnings distribution for STMP changed. What I am curious to know is how do we identify a structural change in the opposite direction? Ie. If a company usually moves 20% on average with the smallest move being 6% and this earnings they only moved 2% after earnings, has there been a structural shift? STMP OM sent me a data set with a bunch of data for the 4 week eps change. I have it cleaned but have not looked at the predictability yet. I am currently working on my modelling/forcasting skills so I have not looked into it yet. Yes, it is the abs 4 week eps change. Zacks data is pricey, yahoo finance offers that indicator for free. However, it is probably useless without historical data. A 5% eps change on NFLX is much different than a 5% eps change on AAPL. Edit* Here is another example with FL, sorry I dont have 10 years of data to show but that first large move of 18% for Footlocker was actually a huge move relative to it's past 12 quarters (almost 2x the size of the second largest). Look how ever since that day, the whole distribution of FL changed. This was during the retail meltdown. Scanning for stocks (ULTA) that just had a huge earnings move could be a great long gamma trade for the next earnings. The implied Move is always slow to adapt! *Double Edit. Here is ULTA, this totally could be a structural shift here. What do you guys think?
One of those nights, little to no sleep and lots of code. Here is a plot where I bin the historical implied moves into equally weighted bins. I then show the distribution of the 30d straddle pnl for each bin. We can see a pattern that as the implied move increases the Long ATM straddle Pnl deteriorates. I have come to the conclusion that selling a straddle where the Implied Move is 18% and the stock does not move (MongoDB) can almost be considered tail risk. You lose almost 100% of a very expensive investment. In shrinkage terms, the largest implied moves are probably too large and the smallest implied moves are probrably too small. Each bin consists of 1400 observations.
I don’t understand the tail risk statement. Doesn’t your data support the idea that the higher the implied move, the less likely you will earn on a long straddle? This is contrary to what I would have expected.