How market makers price earnings

[TheBigShort]

Fair point. But how do you know it’s supply and demand vs a real fundamental issue?

If you trade it systematically then the supply/demand return will be low as you will be forced to trade a lot of “pushes.”

I first want to mention I have read some of your previous posts and find you very intelligent so thanks for continuing to answer my threads/questions. If I assume that the earnings moves are random then it does not matter about the fundamentals. If earnings are not random, then I will eventually find a model that predicts earnings moves and I will become a lot richer ALOT quicker.
I have also matched implied move with actual move, and many times over it fails to model the distribution appropriately. The reason I beleive it has not been arbed away is that for the less liquid companies there is a capacity constraint aswell the drawdowns could get intensive. Like you could put on 5 earnings trades in one day each at 3% of total fund value and lose on all 5.

Lastly I want to mention that humans have an auto correlation pattern(base there decisions on recent past). Look at WMT they have very small earnings moves , last 2 quarters had huge moves so this quarter the expected move was 4.5, the sd for WMT is just over a buck. Thats an easy sell and it was profitable. And if we were wrong and WMT moved a large amount say $8 the losses would have been small.

 

[Kevin Schmit]

For earnings estimates and surprises I access to a Bloomberg terminal. Estimates/surprises up to 2006. What stock/stocks would you like? I'll send the data on monday

Thanks for the offer! Unfortunately I am looking for a market-wide dataset rather than a few individual names. If you can easily (via loop in API or Excel AddIn VBA) create a file with all current Russell 2000 names, please do so and post it. However that also might impact your monthly symbol limit.

I've got Zacks estimate and surprise data going back ostensibly to 2006, but the data are absolute shit in the early years. I way overpaid for it ($1800) considering that the daily updates for the exact same dataset are available from Quantopian for $25/month.

 

[Kevin Schmit]

In image 1 those are the distributions of the last 48 earnings for multiple companies that recently had their earnings. (histogram with 20 buckets)
Image 2: same companies. Y axis = price change close to close after event. x axis = surprise
[

Interesting plots.

In the top panel, histograms aren't that good for assessing normality, but it doesn't look like 70% of those are normal. For a better visuall assessment post qqplot/qqline.

In the bottom panel (the scattargrams) some of the plots do show clearly the expected positive relation, in a few x-wise variation is obscured by a few outliers on the far right, with all the other points clustered around zero on the x axis. It looks like you "suprise" scaling is degenerate on these two scatterplots. Maybe try a different scaling with a more stable denominator like (reported - mean-estimate) / revenues or (reported - mean-estimate) / close-price.

 

[EricJRupe]

Like I would have bee happy selling that straddle for 6. So i am a bit confused

 

[srinir]

Can you explain what you mean by group returns?

Just grouping the stock that exhibits certain characteristics. For ex. small cap growth looks like traders are always optimistic and probably puts/straddles are undervalued. Then look at entire group of those straddles 2 days prior and 2 days later, instead of looking at individual names.

 

[TheBigShort]

Just grouping the stock that exhibits certain characteristics. For ex. small cap growth looks like traders are always optimistic and probably puts/straddles are undervalued. Then look at entire group of those straddles 2 days prior and 2 days later, instead of looking at individual names.

Thanks srinir, I will do a study this week on it and document what i find

 

[TheBigShort]

Thanks for the offer! Unfortunately I am looking for a market-wide dataset rather than a few individual names. If you can easily (via loop in API or Excel AddIn VBA) create a file with all current Russell 2000 names, please do so and post it. However that also might impact your monthly symbol limit.

I've got Zacks estimate and surprise data going back ostensibly to 2006, but the data are absolute shit in the early years. I way overpaid for it ($1800) considering that the daily updates for the exact same dataset are available from Quantopian for $25/month.

Interesting plots.

In the top panel, histograms aren't that good for assessing normality, but it doesn't look like 70% of those are normal. For a better visuall assessment post qqplot/qqline.

In the bottom panel (the scattargrams) some of the plots do show clearly the expected positive relation, in a few x-wise variation is obscured by a few outliers on the far right, with all the other points clustered around zero on the x axis. It looks like you "suprise" scaling is degenerate on these two scatterplots. Maybe try a different scaling with a more stable denominator like (reported - mean-estimate) / revenues or (reported - mean-estimate) / close-price.

Thanks Kevin, I need to do some research on this topic and will get back to everyone when I find something out

 

[TheBigShort]

Hey people, CRM straddle looks like an unreal buy. In fact it is currently pricing in an implied move of 3.5%....30 minutes to close this should be even lower. If you take away it's last 4 quarters, the stock should move ~$13 68% of the time. It also has only moved less than 2% twice so the risk reward is there.

 

[TheBigShort]

The horizontal line is the 3.5 mark. oldest on the left, newest on the right.

 

[truetype]

If you can easily (via loop in API or Excel AddIn VBA) create a file with all current Russell 2000 names, please do so and post it. However that also might impact your monthly symbol limit.

I strongly advise against violating BBG's TOS in such a fashion. They take it very seriously, with good reason.