letting go of the past (maybe?)

[Baozi]

ha.. no idea if this is the right place to post.. I guess it's something in between options and trading psychology..

anyways..

it's already been more than one year that I religiously download every day the option chain for the assets I'm following, and for each asset and each tenor calibrate my distribution parameters on the IVs (in this case I use tempered stable distribution) which in turn allows me to calculate the probability density function. I spent months automating this process, and I'm kinda proud of what I have done so far, but..

..the sad truth is that so far I haven't been able to make a dime out of it! the calibration itself takes around 1 hour to complete, so the tool can be used only for end-of-the-day calculations. And I honestly have no idea how to use it. The sharp movements on the day opening automatically invalidate the scenario of the night before, and the model doesn't seem to have any (meaningful) predictive value on a longer timeframe.

By now, I found my edge elsewhere, and my study/research is going in a very different direction. So rationally I know I should stop wasting time every day just going through the motions and recording the data into an ever growing spreadsheet.. but so far I sunk so many hours in this project that it HAS to be good for something... right? RIGHT ??

 

[smallfil]

ha.. no idea if this is the right place to post.. I guess it's something in between options and trading psychology..

anyways..

it's already been more than one year that I religiously download every day the option chain for the assets I'm following, and for each asset and each tenor calibrate my distribution parameters on the IVs (in this case I use tempered stable distribution) which in turn allows me to calculate the probability density function. I spent months automating this process, and I'm kinda proud of what I have done so far, but..

..the sad truth is that so far I haven't been able to make a dime out of it! the calibration itself takes around 1 hour to complete, so the tool can be used only for end-of-the-day calculations. And I honestly have no idea how to use it. The sharp movements on the day opening automatically invalidate the scenario of the night before, and the model doesn't seem to have any (meaningful) predictive value on a longer timeframe.

By now, I found my edge elsewhere, and my study/research is going in a very different direction. So rationally I know I should stop wasting time every day just going through the motions and recording the data into an ever growing spreadsheet.. but so far I sunk so many hours in this project that it HAS to be good for something... right? RIGHT ??

If it has no edge, it has no value. A negative expectation game, like in Las Vegas casinos, only favors the casinos. In this case, it only favors those traders who have an edge. They will take your monies because they worked hard to design a trading system that works, has an edge and has been backtested.

 

[Baozi]

If it has no edge, it has no value. A negative expectation game, like in Las Vegas casinos, only favors the casinos. In this case, it only favors those traders who have an edge. They will take your monies because they worked hard to design a trading system that works, has an edge and has been backtested.

Correct. This is also what the rational part of my brain is telling me. And trading should be a rational endeavor.. and yet so many times the psychological part messes things up..

 

[BlueWaterSailor]

By now, I found my edge elsewhere, and my study/research is going in a very different direction. So rationally I know I should stop wasting time every day just going through the motions and recording the data into an ever growing spreadsheet.. but so far I sunk so many hours in this project that it HAS to be good for something... right? RIGHT ??

What you learned in the process of setting it up and making it work is incredibly valuable - and that part is not going to be lost regardless of what you do about this project. In fact, if I had to guess, I'd bet that at least a part of what you learned contributed to finding your edge... that's how this stuff works, little improvements from here, there, and everywhere that build up to a unique, optimized-to-you strategy.

I have a code repo, with some stuff that I wrote so long ago that it has cave dust and mammoth coproliths in it; truly cringy crap if looked at objectively. AND YET: all of that, and the process of learning how to do it better and not make those mistakes went into building the foundation of my programming skills today. The only attitude that makes sense is to be humbly grateful to the "me" who had the guts to dive in and make those stupid mistakes back then - because I wouldn't be here without them.

 

[M.W.]

What you are doing only benefits you when you make markets or trade low latencies and when you trade illiquid options. You calibrating the IV grid (that is what you are in effect doing through the estimation of distributional properties) only makes sense if you intend to trade illiquid options for which there is no recently traded contracts available and hence you approximate IV for the illiquid option by using liquid ones and interpolation techniques. What you are doing does not help you in spotting relative value unless you trade mentioned illiquid contracts. From what you described you probably want to consider different approaches to gauge a view of where IVs should be relative to traded levels.

ha.. no idea if this is the right place to post.. I guess it's something in between options and trading psychology..

anyways..

it's already been more than one year that I religiously download every day the option chain for the assets I'm following, and for each asset and each tenor calibrate my distribution parameters on the IVs (in this case I use tempered stable distribution) which in turn allows me to calculate the probability density function. I spent months automating this process, and I'm kinda proud of what I have done so far, but..

..the sad truth is that so far I haven't been able to make a dime out of it! the calibration itself takes around 1 hour to complete, so the tool can be used only for end-of-the-day calculations. And I honestly have no idea how to use it. The sharp movements on the day opening automatically invalidate the scenario of the night before, and the model doesn't seem to have any (meaningful) predictive value on a longer timeframe.

By now, I found my edge elsewhere, and my study/research is going in a very different direction. So rationally I know I should stop wasting time every day just going through the motions and recording the data into an ever growing spreadsheet.. but so far I sunk so many hours in this project that it HAS to be good for something... right? RIGHT ??

 

[murray t turtle]

ha.. no idea if this is the right place to post.. I guess it's something in between options and trading psychology
it's already been more than one year that I religiously download every day the option chain ...

By now, I found my edge elsewhere, and my study/research is going in a very different direction. So rationally I know I should stop wasting time every day just going through the motions and recording the data into an ever growing spreadsheet.. but so far I sunk so many hours in this project that it HAS to be good for something... right? RIGHT ??

%%
I think so, because i still stop + pick up a penny
THAT's on main street[ not Wall Street]+ sure not my main goal.
BUT all the above has more value than WSJ crossword puzzle/LOL.
NO dis-repect to the market maker that make a fortune off penny bid/ ask.
My mom still like crossword puzzles; not me . find something better.
I see SPY levels @$1, not wasting time,chained to that /LOL

 

[TheDawn]

ha.. no idea if this is the right place to post.. I guess it's something in between options and trading psychology..

anyways..

it's already been more than one year that I religiously download every day the option chain for the assets I'm following, and for each asset and each tenor calibrate my distribution parameters on the IVs (in this case I use tempered stable distribution) which in turn allows me to calculate the probability density function. I spent months automating this process, and I'm kinda proud of what I have done so far, but..

..the sad truth is that so far I haven't been able to make a dime out of it! the calibration itself takes around 1 hour to complete, so the tool can be used only for end-of-the-day calculations. And I honestly have no idea how to use it. The sharp movements on the day opening automatically invalidate the scenario of the night before, and the model doesn't seem to have any (meaningful) predictive value on a longer timeframe.

By now, I found my edge elsewhere, and my study/research is going in a very different direction. So rationally I know I should stop wasting time every day just going through the motions and recording the data into an ever growing spreadsheet.. but so far I sunk so many hours in this project that it HAS to be good for something... right? RIGHT ??

If you have no idea what to do with the data, either 1) find out how to use the data to devise a trading plan or 2) ditch the data if you can't use it. You have no idea how much research that I have thrown away before finding something that I could finally use to form a trading plan. It's normal. It's all part of trading. We are traders. If we can't trade something then it's of no use to us.

What I would suggest is don't erase the data. Just leave it there; store it somewhere. You never know when you might find it useful again and need to go back to it.

 

[Baozi]

I have a code repo, with some stuff that I wrote so long ago that it has cave dust and mammoth coproliths in it; truly cringy crap if looked at objectively. AND YET: all of that, and the process of learning how to do it better and not make those mistakes went into building the foundation of my programming skills today. The only attitude that makes sense is to be humbly grateful to the "me" who had the guts to dive in and make those stupid mistakes back then - because I wouldn't be here without them.

Had to look up 'coprolith', but besides this, thanks A lot man, this is giving me the closure I was looking for..

 

[Baozi]

What you are doing only benefits you when you make markets or trade low latencies and when you trade illiquid options. You calibrating the IV grid (that is what you are in effect doing through the estimation of distributional properties) only makes sense if you intend to trade illiquid options for which there is no recently traded contracts available and hence you approximate IV for the illiquid option by using liquid ones and interpolation techniques. What you are doing does not help you in spotting relative value unless you trade mentioned illiquid contracts. From what you described you probably want to consider different approaches to gauge a view of where IVs should be relative to traded levels.

its not really an interpolation because the curve is parametric, the inputs are spot, dte and six other parameters related to the curve shape. but yes of course once you have it you can estimate the values at otc strikes.

what about the related probability density function ? Also that quite meaningless?

 

[M.W.]

You are right, in your case you were likely estimating the parameterizations but it is akin to an estimation of expected implied volatilities. The density can be useful but in my opinion for your use case not in the way you derived it. Look at the larger picture, you are essentially using market IVs to derive the density of an underlying continuous random variable that is also market driven. What edge are you expecting to derive from that? That's why I mentioned that such approach is often times used in market making of otc options, hence likely of little value to you. What you want is the density that is derived from non-traded metrics. That can then be used as foundation for a model that outputs the deviation from traded levels and at which you then decide to buy or sell volatility.

its not really an interpolation because the curve is parametric, the inputs are spot, dte and six other parameters related to the curve shape. but yes of course once you have it you can estimate the values at otc strikes.

what about the related probability density function ? Also that quite meaningless?