You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Trading Volatility
- Thread starter sle
- Start date
If you really want to trade volatility you need to look into IV p&l. See how much of an option's price change is because of volatile. I have started collecting intraday greeks and IV for SPY. Every option for every month. If you are interested in getting some of this data then PM me. I should have some demo screenshot of the software and data soon.
Given that the majority of people here will be trading shorter dated options, the value of vega p&l (especially intraday) is going to be very small. An IV database that has been made consistent time-wise and strike-wise is very useful indeed, especially for strategies where you are mixing up vega and gamma, as in calendar spreads.Quote from mike007:
If you really want to trade volatility you need to look into IV p&l. See how much of an option's price change is because of volatile. I have started collecting intraday greeks and IV for SPY. Every option for every month. If you are interested in getting some of this data then PM me. I should have some demo screenshot of the software and data soon.
I found that backtesting has very little value in volatility trading, you better of concentrating on forecasting volatility and response of the implied volatility to the changes in realized volatility.
Just my 2c on the topic of software tools.
Forecasting is better but how can you forecast without data? I will have a few features where you can download the raw data in csv format for loading it into excel for your own forecasting.
Also, you can get this data live throughout the day.
Also, you can get this data live throughout the day.
Quote from sle:
Given that the majority of people here will be trading shorter dated options, the value of vega p&l (especially intraday) is going to be very small. An IV database that has been made consistent time-wise and strike-wise is very useful indeed, especially for strategies where you are mixing up vega and gamma, as in calendar spreads.
I found that backtesting has very little value in volatility trading, you better of concentrating on forecasting volatility and response of the implied volatility to the changes in realized volatility.
Just my 2c on the topic of software tools.
In general, to forecast realized vol you would use daily and high/medium frequency underlying data, in your case S&P. This said, I've yet too see good volatility forecasting tools that take into account multiple aspects of the world, except of the ones that were developed by traders themselfs. For I'm implied vol data I like to use constant maturity vols (e.g. 1w, 2w,1m etc) and some sort of parametric fit for the skew (normal SABR in rate space for IR options, a tan parametric for S&P and other equity indices).Quote from mike007:
Forecasting is better but how can you forecast without data? I will have a few features where you can download the raw data in csv format for loading it into excel for your own forecasting.
Ps. Ability to see the fitted surface and the source for the fitting ( individual option vols) would be a useful feature, I think.
PPS. Live data through the day is useful, yes.
sle- Below are some questions I have on a strat I've been working on...just say the word if you'd like me to have this moved to another thread, but i think it may be topical and of interest to others setting up a modelling/testing framework.
-------------------------------------------
Basically, I have a model that forecasts the underlying's return distribution for the next week. I'm fairly confident in the stationarity of the predictions, but to this point haven't been able to test whether or not the market is already pricing in iv,skew, and kurt that reflects what I think "I know" from my forecasts. Certainly, my model's theoretical values will tie out with the market's much of the time, but when there are significant/tradeable differences, I need to determine if my model is providing an edge or as sle mentions above, if it's just missing some important exogenous data.
Before I can quantify the quality of the forecasts relative to the market, I need to determine how to trade them. Weekly options are obviously a good fit here, but I wanted to see if there are any opinions on the best way to implement my perceived edge. Theoretically, I can use my distro forecast to price all options in a weekly chain and trade the most over/under priced, but having said that, I'm somewhat partial to trading something that looks like a long fly for a few reasons:
1. defined risk
2. my other models tend to underperform in declining vol environments (they all trade the underlying directly, this would be my first options strat)
3. i believe that there's some positive expectancy from excess risk premium which would put the wind at my back in case my model sucks.
But I'm very open to suggestions here...if anyone has recommendations on how they would trade this type of forecast, it would be great to hear.
I think this is where it gets interesting (and probably more broadly applicable to others)...once I settle on how to trade it, I still need to test it. I wouldn't be doing any delta hedging, so that simplifies things a bit, but the real problem is that access to historical weekly ATM IVs (and even worse, vol skew) is rather limited. Is it worth bothering to obtain that data, or does it make more sense to just trade it small going forward?
If the answer is to trade it small going forward, then I need a way to compare the results of the trades my model selects vs. ones taken randomly. I have an idea of how I would do this objectively and account for the uncertainty of mid-market fills, but I won't bloat the thread anymore until it's clear that's the best way to go. If it is, I'll put it in here and people can rip it apart(should look like a distant cousin of bootstrap).
Appreciate any thoughts on the above.
-------------------------------------------
Basically, I have a model that forecasts the underlying's return distribution for the next week. I'm fairly confident in the stationarity of the predictions, but to this point haven't been able to test whether or not the market is already pricing in iv,skew, and kurt that reflects what I think "I know" from my forecasts. Certainly, my model's theoretical values will tie out with the market's much of the time, but when there are significant/tradeable differences, I need to determine if my model is providing an edge or as sle mentions above, if it's just missing some important exogenous data.
Before I can quantify the quality of the forecasts relative to the market, I need to determine how to trade them. Weekly options are obviously a good fit here, but I wanted to see if there are any opinions on the best way to implement my perceived edge. Theoretically, I can use my distro forecast to price all options in a weekly chain and trade the most over/under priced, but having said that, I'm somewhat partial to trading something that looks like a long fly for a few reasons:
1. defined risk
2. my other models tend to underperform in declining vol environments (they all trade the underlying directly, this would be my first options strat)
3. i believe that there's some positive expectancy from excess risk premium which would put the wind at my back in case my model sucks.
But I'm very open to suggestions here...if anyone has recommendations on how they would trade this type of forecast, it would be great to hear.
I think this is where it gets interesting (and probably more broadly applicable to others)...once I settle on how to trade it, I still need to test it. I wouldn't be doing any delta hedging, so that simplifies things a bit, but the real problem is that access to historical weekly ATM IVs (and even worse, vol skew) is rather limited. Is it worth bothering to obtain that data, or does it make more sense to just trade it small going forward?
If the answer is to trade it small going forward, then I need a way to compare the results of the trades my model selects vs. ones taken randomly. I have an idea of how I would do this objectively and account for the uncertainty of mid-market fills, but I won't bloat the thread anymore until it's clear that's the best way to go. If it is, I'll put it in here and people can rip it apart(should look like a distant cousin of bootstrap).
Appreciate any thoughts on the above.
Are there any strategies that can be used effectively during the IV run-up before earnings calls?
I'm relatively inexperienced but I've seem promising results doing backtesting in Matlab where I "buy" an ATM straddle on an equity 2-4 weeks before the earning call. I've then done a monte carlo approach by letting the price randomly walk while applying a linear increase in IV (based on historical IV before earnings) as the call approaches. I would then liquidate the position just before the announcement.
My idea is that I am delta neutral if the price doesn't move and then it's purely theta vs. vega. If the expected IV buildup is enough there is a small profit in the trade. If the price does move in anticipation of earnings then there is a potential for the straddle delta to pay a substantial amount.
I'm relatively inexperienced but I've seem promising results doing backtesting in Matlab where I "buy" an ATM straddle on an equity 2-4 weeks before the earning call. I've then done a monte carlo approach by letting the price randomly walk while applying a linear increase in IV (based on historical IV before earnings) as the call approaches. I would then liquidate the position just before the announcement.
My idea is that I am delta neutral if the price doesn't move and then it's purely theta vs. vega. If the expected IV buildup is enough there is a small profit in the trade. If the price does move in anticipation of earnings then there is a potential for the straddle delta to pay a substantial amount.