SPX Credit Spread Trader

[yip1997]

Quote from uglyboy:

Exactly, it would seem to imply you could buy butterflies on either side of the current underlying with the body at 1sigma, and expect.......I want to say positive expectancy, but I'm afraid........an advantage. Similarly, selling at 1 sigma and buying at >2 sigma to create a credit spread might take advantage of this. Again, I don't think this is actually possible.

I agree that the most obvious thing to take home is that unusual events - the ones that make CSs lose money occur more commonly than logarithmic distribution implies - thus the skew.

I had similar idea in the past, and asked Riskarb if there were any combo that will have positive expectancy, and the answer was "No". In fact, I used +dgamma and RA corrected me with the proper term "long wings".

Ugly,

Thanks for your post. It put our focus back to learning better trading strategies, and sharing trading ideas.

 

[El OchoCinco]

If any spread had positive expectancy, wouldn't we all just trade that one?

I said it before and I will say it again, the strategy is meaningless. Success is based on your risk management and trade management skills. So find the strategy that best suits your skill and risk tolerance and manage your risk, whichever strategy that is...

 

[yip1997]

Quote from lloyd111:

Hi all, do any of you fine people have any experience or insight about artificial intelligence trading systems. I have received numerous calls attempting to recruit me into one such system being offered by "my trading futures.com" aka mind engines. Sounds intriguing but I cannot find any reviews either positive or negative about this. Any serious feedback would be much appreciated. Ty in advance.

Lloyd

Lloyd,

I used to work on Artificial Intelligence. More specific, mainly on artifical neural networks, genetic algorithms and fuzzy control.

10 years ago, a graduate came back to our school to recruit and told me that the neural network algorithm we developed worked well for currency trading in some calendar days. I tried to learn the price patterns with neural network. It didn't work. Period.

I believe you can develop some AI software for your decision making process. Nothing replaces human intelligence yet. I have some "in-house" software for my decision making process. They are a combination of control theory, statistics and AI. I don't know if the software helps me to make more profit but it certainly makes me very confidence in making decision.

 

[yip1997]

Quote from optioncoach:

If any spread had positive expectancy, wouldn't we all just trade that one?

I said it before and I will say it again, the strategy is meaningless. Success is based on your risk management and trade management skills. So find the strategy that best suits your skill and risk tolerance and manage your risk, whichever strategy that is...

Phil,

I totally agree with you.

 

[uglyboy]

Quote from optioncoach:

If any spread had positive expectancy, wouldn't we all just trade that one?

I agree with that statement. I also want to point out, I don't think that these butterflies have + expectancy. I'm just trying enjoying looking deeper at the strategy we are discussing in this thread.


How does it feel to finally be the hammer and not the anvil in the "positive expectancy wars"?

 

[yip1997]

Ugly and others:

I like to throw another idea to see if anyone has an answer for me.

BS model is based on risk-neural valuation, and so it uses the risk-free interest as u.

From Brownian process, u is the average rate of return. If you solve the stochastic process, you should use the statistical mean, that is on average 8% higher than risk-free interest rate.

1. Why shouldn't we use the mean return of the stock market?

2. If the stock return is higher than risk-free rate, and the options are priced using risk-free rate, does it have any implication on trading strategies? What trading strategies should we use?

 

[uglyboy]

Quote from yip1997:

Ugly and others:

I like to throw another idea to see if anyone has an answer for me.

BS model is based on risk-neural valuation, and so it uses the risk-free interest as u.

From Brownian process, u is the average rate of return. If you solve the stochastic process, you should use the statistical mean, that is on average 8% higher than risk-free interest rate.

1. Why shouldn't we use the mean return of the stock market?

2. If the stock return is higher than risk-free rate, and the options are priced using risk-free rate, does it have any implication on trading strategies? What trading strategies should we use?

Yip, I'll give this one a try, although I am by no means an expert. Please don't assume that my answer is accurate. Hopefully someone smarter can pitch in too.

The dynamic replication process (i.e B-S equation and others) demonstrates that you can replicate the pay-out of an option by combining cash/bonds and stock. As the underlyings price moves, you can continuously alter the proportion of cash/stock to match the options expected pay-out. This ratio of stock/cash is how the delta is generated. The risk-free rate is used for the cash portion of this dynamic replication. The "average stock market return" is used to generate the expected return on the stock portion of this dynamic replication (hence the need for the logarithmic distribution assumption). So I think that using the "stock market return" as the return on the cash/bond portion of the replication process would give an inaccurate result of the expected reward.

In theory this process is stochastic and continuous (i.e. calculus) without transaction costs. It also assumes that the Brownian motion of stock prices is actually Brownian - no jumps, no discontinuity. Some people think that the skew is due to fear of sudden jumps (eg a crash). The B-S equation doesn't account for jumps, which is why it can't explain the skew. There have been several jump-diffusion models over the last decade - I'm not really capable of commenting on them intelligently. There are lots of other problems with the B-S model - it is after all a model, which is a simplified approximation of reality.

I think that this post was probably a bit confusing. Partly because I'm not at the best of times that quick, partly because I spent the day skiing+beer ( 200cms at Whistler!). I will try again if you want me to.

 

[yip1997]

Quote from uglyboy:

Yip, I'll give this one a try, although I am by no means an expert. Please don't assume that my answer is accurate. Hopefully someone smarter can pitch in too.

The dynamic replication process (i.e B-S equation and others) demonstrates that you can replicate the pay-out of an option by combining cash/bonds and stock. As the underlyings price moves, you can continuously alter the proportion of cash/stock to match the options expected pay-out. This ratio of stock/cash is how the delta is generated. The risk-free rate is used for the cash portion of this dynamic replication. The "average stock market return" is used to generate the expected return on the stock portion of this dynamic replication (hence the need for the logarithmic distribution assumption). So I think that using the "stock market return" as the return on the cash/bond portion of the replication process would give an inaccurate result of the expected reward.

In theory this process is stochastic and continuous (i.e. calculus) without transaction costs. It also assumes that the Brownian motion of stock prices is actually Brownian - no jumps, no discontinuity. Some people think that the skew is due to fear of sudden jumps (eg a crash). The B-S equation doesn't account for jumps, which is why it can't explain the skew. There have been several jump-diffusion models over the last decade - I'm not really capable of commenting on them intelligently. There are lots of other problems with the B-S model - it is after all a model, which is a simplified approximation of reality.

I understand the dynamic replication, and so risk-neural valuation process, and it should use risk-free interest if deriving from this perspective.

However if you assume a pure brownian process, you should use the mean return if solved from this perspective. In reality, we use the standard deviation of stock return for sigma. Why don't we use the statistical mean of stock return for u.

Why can't we have free lunch statistically?

 

[MechTrade]

Quote from uglyboy:

I agree with that statement. I also want to point out, I don't think that these butterflies have + expectancy. I'm just trying enjoying looking deeper at the strategy we are discussing in this thread.


How does it feel to finally be the hammer and not the anvil in the "positive expectancy wars"?

Ugly... You're a real stinker.

All that statistical sigma & + expectancy dialog, only to finally reveal what a level-headed and devious mind you have. LOL.

Mech

 

[lloyd111]

dagnyt, about the Artificial Intellegence system,we share the same sentiment...noone sells a producer for no chump change. TY for the feedback.

yip1997, your reply is exactly what I was looking for. I am skeptical at the best of times and would not use any purchasable systems to begin with. But the AI part was intriguing to me cause it is not a common topic. Great you replied with the background you acquired in this field to share your view. thx a bunch.......to all.

Lloyd

*note* I will continue to trade my way as I too believe that human intelligence for the majority of our population is irreplacable..