The first step [for the pros only]

[babutime]

Quote from sle:

Mispricing of the options IS mispricing of the expected moves, since value of the option is dicatated by the implied volatility (which is the mean expectation of the moves).

But IV is dictated by the option price itself- unless all those codes I wrote a cpl years back for this undergrad class was in vain... What am I missing?

The mean expectation of the moves .. is the vol of the vol also important to consider then?

Wish I could get my hands on some of that kinda data...

Quote from sle:

either vs history or vs your subjective interpretation

My tools to do so has been limited so far to only ThinkOrSwim's thinkback feature. It's quite a brudal way to do so. LiveVol has a handy feature that lets you look for past performance of ATM straddles on earnings. A bit too expensive at this time. But backtesting has its own pitfalls- especially for earnings, dont you think? Since you cant really do any out of sample testing. So your point on subjective interpretation makes the most sense. I guess I'll have to sharpen that with practice and age and study the dynamics of each earnings vol individually.

Thanks sle !

 

[babutime]

Quote from newwurldmn:

Earnings as a basket aren't any trickier than other vol. The one advantage is that the pnl volatility allows you to trade fewer options meaning less tranaction costs relative to potential gain/loss.

Could you elaborate on how I would trade fewer options? Don't quite grasp that.

Gawd I need sleep... pulling an all nighter here...

 

[amazingIndustry]

Sorry, I was referring to you Babutime ;-)

Quote from babutime:

Thanks a lot amazing. Kind words.

And I'm not Sonoma... .. got the wrong handle there..

 

[sonoma]

Quote from babutime:

...But backtesting has its own pitfalls- especially for earnings, dont you think? Since you cant really do any out of sample testing.

Why can't you use your data source for out of sample testing?

 

[babutime]

Quote from sonoma:

Why can't you use your data source for out of sample testing?

I was referring to back-testing for earnings related moves. So even if you analysed 3 years worth of earnings activity, you have 12 data points to work with.

Now, to be fair, that is a lot of data points in the context of earnings. But they all could have happened under so many different conditions that even if one got a 95% chance that volatility, prices and IV have acted in a certain way in the past (during earnings) doesn't mean it will again.

So doing an out of sample testing still won't help much.. But I can see it being a viable solution if one combines it with current conditions.

 

[newwurldmn]

Quote from babutime:

I was referring to back-testing for earnings related moves. So even if you analysed 3 years worth of earnings activity, you have 12 data points to work with.

Now, to be fair, that is a lot of data points in the context of earnings. But they all could have happened under so many different conditions that even if one got a 95% chance that volatility, prices and IV have acted in a certain way in the past (during earnings) doesn't mean it will again.

So doing an out of sample testing still won't help much.. But I can see it being a viable solution if one combines it with current conditions.

Yeah. it's tough. If you look at earnings moves for August-September of 2011, they will look extremely high, but that's because the world was much more volatile, so you have to account for that. All of a sudden history is harder to look at.

 

[sonoma]

Quote from babutime:

...So doing an out of sample testing still won't help much.

Using historical data is fine for understanding the mean and variability associated with earnings, but I don't consider that out of sample testing. I save "out of sample" to describe testing a hypothesis you've generated using your data source and you want to explore the robustness of the hypothesis.

 

[babutime]

Quote from sonoma:

Using historical data is fine for understanding the mean and variability associated with earnings, but I don't consider that out of sample testing.

Me neither. I don't think anyone would since that's not the definition at all. Given the limited number of earnings, the variables to be taken into account- everything from macro factors to how the underlying was trading, are enormous.

Not worth all that effort. Treating each event individually and trading small and assigning past events lower weights is probably the best way.

 

[sonoma]

Quote from babutime:

Me neither. I don't think anyone would since that's not the definition at all. Given the limited number of earnings, the variables to be taken into account- everything from macro factors to how the underlying was trading, are enormous.

Not worth all that effort. Treating each event individually and trading small and assigning past events lower weights is probably the best way.

Just wanted to be sure we were on the same page. Because the data source you have is rich with info. It's not the monster some have access to, but it's a good start.

 

[babutime]

Quote from sonoma:

Just wanted to be sure we were on the same page. Because the data source you have is rich with info. It's not the monster some have access to, but it's a good start.

You mean IB or ThinkOrSwim? ThinkOrSwim is just a paper trade account

IB on the other hand I'll try some java with it soonish. Should be fun to fool around with.