Statistical edge with option spreads -none?

[asiaprop]

sorry but your answer shows clearly that you do not seem to possess a whole lot of technical options expertise.

I dont question the way you trade but the issue arose whether option pricing models are important to exploit an edge. I claim they are essential. You did not answer in the slightest my questions about those options that empirically proved to be the most mistpriced. There is no way around a very refined model in order to be capable of accurately pricing the wings and far dated options.

Quote from dmo:

Since you talk about your "hedging style," I assume that you are talking about relative mispricings rather than absolute mispricings, and that you are spreading one option against another.

If that is correct and the "mispricing" is relative, it really only has meaning in the context of whatever strategy you are using to exploit the "mispricing."

Far be it from me to suggest you change something that's working. But how could you do it using standard pricing models? You can assign an IV to every strike. That IV is the "price" of the option. You can plot those IV's and massage them any number of ways to identify a "mispricing." You can utilize curve-fitting packages and functions to evaluate each point (IV) in the curve.

But there is no "mispricing" without a strategy to exploit it. Unless you're a directional trader, success trading options is a two-step process, identifying a mispricing and exploiting it.

If you have a model that identifies mispricings and a strategy to exploit them, then bravo. But your model is useless without the strategy. And that's the point I was trying to make in my post to optiongirl - that it's not one model or another that makes you successful, it's what you do with the information it generates.

 

[atticus]

Quote from asiaprop:

please take no offense but let me ask you then, how does your off-the-shelf program price far out of the money options as well as options couple months out on, lets say, Nikkei or Kospi index options that are only actively traded between the 90-110 strikes as well as first 2 front months? The further out ones and farther away from the money ones are the ones that are most often mispriced and even large sell-side desks often misprice them. I myself could not find a single off-the-shelf program that could remotely do the job. In the end it came down to our desk taking a very successful model that was traded by a guy in NY after a lot of tweaking and adjusting to make it fit to our hedging style as well as peculiarities of NKxx and KMxx options

Please enlighten me, maybe I re-invented the wheel ;-)

WTF does "hedging style" have to do with model output? I would love to know which banks are mispricing Kospi vanilla options.

 

[asiaprop]

GS, JPM, DB, even SocGen and BNP among others, ask for some calendar spreads and ratios and you will sometimes be quite surprised. Or try 3-6 months out varswaps right after the open after turbulent US markets the night before...enough said...

I am not gonna volunteer any more information as I still continue to work in the region and from time to time picked off those shops for nice arb profits and plan to continue in doing so..., my point was that models matter if you want to identify mispriced options, I am not gonna change my opinion and at least hoped I made couple others think about it...and yes...I believe mispriced options can also be identified on the retail side with some efforts put into your systems.

Quote from atticus:

WTF does "hedging style" have to do with model output? I would love to know which banks are mispricing Kospi vanilla options.

 

[TraderSU]

Quote from asiaprop:

GS, JPM, DB, even SocGen and BNP among others, ask for some calendar spreads and ratios and you will sometimes be quite surprised. Or try 3-6 months out varswaps right after the open after turbulent US markets the night before...enough said...

Sorry for newbie question. Aren't they mis-priced for a reason? Guess no model can accommodate every aspect of pricing. For example BS model doesn't know anything about Washington.

Also curious if this mis-priced options always shift toward their so called "fair value". Why can't they move further away from their "model value" making "cheap options" cheaper?

- SU

 

[TallTrader]

Statistical edge is only applicable if one is assuming an underlying pricing model, like Black Scholes. But even that is recently being questioned as being valid.

 

[sonoma]

Quote from optionsgirl:

We read about the implied statistical edge concerning options all the time, but it seems that there is no statistical edge with any options strategy that I can think of. Now my mathematical aptitude isn't that great, but this is why I am posting it here for critique. I think most option strategies act more or less as a stop loss. Supposedly, the most conservative strategy would be a butterfly, but is it that much more conservative than a regular vertical spread? If there isn't a statistical edge with a butterfly over a simple vertical spread, then people are losing money on commission.

Here is a pattern that I see with a 50/50 chance with a random option. Get out when I double my money. Get out when I lose half. Double your money. Lose half. Double your money. Lose half. If I started with $100, I would end up with about $100 (not counting commission) no matter what the strategy is. Obviously, this isn't a realistic scenario, but it's just an idealized illustration of profits and losses with option hedging.

I thought there is a statistical edge with a far out-of-the-money short spread. This would be a steady trickle of money, but I think once in a while all those pennies will be lost by a major move and you end up about breaking even again.

I realize this is quite an inaccurate way to describe statistical probabilities, but this is just my guestimation that there is no statistical edge with options --or at least anything worthwhile. I think the only way to have an edge is to predict one or all of these things: implied volatility, historical volatility, and price movement of the stock.

I doubt any non-professional trader who lacks extensive experience in financial engineering could both recognize and then successfully trade any type of static statistical edge. Nor is it necessary unless you're running millions. Better to think of "statistical edge" in terms of understanding that profitability zones occur during the life of a trade, even if the final payout behaves in a statistically perfect manner. You have to act when those profitable zones appear. Of course, looking at trading in this fashion implies intermittent hedging but isn't that where the fun is, anyway?

 

[atticus]

Quote from asiaprop:

GS, JPM, DB, even SocGen and BNP among others, ask for some calendar spreads and ratios and you will sometimes be quite surprised. Or try 3-6 months out varswaps right after the open after turbulent US markets the night before...enough said...

I am not gonna volunteer any more information as I still continue to work in the region and from time to time picked off those shops for nice arb profits and plan to continue in doing so..., my point was that models matter if you want to identify mispriced options, I am not gonna change my opinion and at least hoped I made couple others think about it...and yes...I believe mispriced options can also be identified on the retail side with some efforts put into your systems.

I was using DB Prime until 2008. I can state from experience that SocGen, RBS and UBS routinely misprice Euro-convention forward-start exotics. I've never seen any shop price vanilla equity index strikes out of line with, for example, Superderivates or Fenics on the FX-side, much less handing anyone a ready-arb.

I asked you a question on page 9 of this thread -- what persistent LISTED vanilla vol-skew, in any market, have you been able to isolate at nominal gamma?

 

[drcha]

I think of it as swing trading--getting in and out on the right day/time makes up for the lack of a true statistical edge. That seems to me what a lot of people are saying here.

 

[rallymode]

Quote from asiaprop:



However, option "mispricing" can be consistently identified with a well developed and updated vol surface and underlying model. Just imagine an insurance firm buying a lot of the at the money 2nd front month puts. Huge demand will make those overvalued relative to the front month or further out expiries. Selling those and offsetting the remaining exposure in a smart way will give you an edge. Point is to identify those "mispricings" before others step in to do the same.

I would agree only if the premise of your argument is an illiquid spot. Such edge is neither scalable nor consistent, though still an edge for argument sake.

 

[asiaprop]

good questions...sure they are mispriced for a reason, be it as simple as supply/demand imbalances. And yes, this can further widen out. But given you hedged your remaining exposure with other options around those particular mispriced ones your pay day will at latest be expiration. Sure, there are a number of other risk considerations but to me it proved to be well worth the endeavor. I make it sound very simple, I admit it is not as simple as I described. For instance if you offset some of your mispriced May expiry risk with the June expiries you may want to shift back into May expiries as soon as you capture some or most of the mispricing.

Quote from TraderSU:

Sorry for newbie question. Aren't they mis-priced for a reason? Guess no model can accommodate every aspect of pricing. For example BS model doesn't know anything about Washington.

Also curious if this mis-priced options always shift toward their so called "fair value". Why can't they move further away from their "model value" making "cheap options" cheaper?

- SU