Statistical edge with option spreads -none?

[heech]

I am inclined to believe that stat arb does exist, but it's incredibly, incredibly rare... and it's certainly nothing as simple as buying or selling a put/call/calendar/butterfly spread. To have actual statistical "edge" that you can arb, you need a better-than-average continuous model for price and volatility. To put it mildly, it's more or less the holy grail.

For the vast majority of option traders... we're fundamentally no different than stock traders. The shape of the payoff function looks a little different, but the ultimate risk/reward ratio over time will look the same.

 

[vanv0029]

I have been working on a philosophical argument that I think shows
that any option strategy that does not involve predicting market direction
(or at least expiration day ending price) losses money.

My observation is that outcomes from the following thought experimental
will fail any kind of spectral test, i.e. there will be no random variation.
At the current month's expiration buy or sell the nearest strike straddle.
At the next months expiration, underlying price will be manipulated to be
exactly evenly distributed (ignore dividends, interest costs, etc.).
Namely, there will be no 'statistical variation' even over a small number
of samples. In its simplest form, exactly half of the closing prices will
be less and half greater. In a random set of trials there should be much
more variation. I thought of this in trying to explain why I think the
various Cottle type straddle/strangle trades can't make money.

I think this shows that statistics is a nonsense concept like
signs of the Zodiac. One's Zodiac sign has some predictive power providing
society believes in it as a convention.

There are two cases. Statistics involving infinity is nonsense
since no operations can be defined for infinite sets. Statistics is
just conventions and axioms here. The ignored 20th century logician
Paul Finsler showed this. The only algebraic operation that makes
sense for infinite sets is one-to-one correspondence making.

The other case is finite experiments involving human society. I think
the right analogy for seeing option trades is to see a fixed dice game.
The dice are magnetized and everybody has a magnet some small and some
large. Market makers basically have the advantage of seeing retail
magnets but no magnet of their own. The Tarp banks have both better viewing
of magnets and large magnets themselves. This makes statistics and the
various theorems pure fantasy.

This also shows that no retail trader can make money even if trading
costs are ignored from non price change predicting option trades - any
option strategy that does not make a directional prediction will lose money.
It also suggests that market makers can't make money unless their profit
is tolerated by the vertically integrated and now Tarp receiving large
magnet holding money center banks.
~

 

[asiaprop]

this is where your assumption is wrong. The distribution of the underlying prices is NOT normal. That is one of the most fundamental issues in options trading. And this is where some of the models have the edge. Options are priced according to which probabilities the market for up and down moves implies. Those probabilities are pretty much always skewed. If you model disagrees with this skew with a statistical edge then you have a strategy with edge.

Those are just some very basic observations after having traded options for institutional outlets for years.

Quote from optionsgirl:

Well, I was speaking in a relative sense. Here is an example for WFC options:

June 17 put: 0.38
June 20 put: 0.90
June 23 put: 1.90

If I bought a June 20 put, I would assume there would be a 50 percent chance that the stock would go down $3 and a 50 percent chance that it would go up $3. It isn't exactly a 50% loss and 100% percent gain, but it's roughly close. I suppose this isn't very accurate since the chance of a $3 move either way isn't really 50-50, but I still assume the chance of a $3 move up is about the same as a $3 move down.

 

[asiaprop]

I like to add that another factor that may differentiate retail from professional traders is capitalization. I would assert that pure vol trading strategies for a retail guy/gal are impossible without at least putting USD 1 million on the line.

Quote from asiaprop:

that is nonsense. The same set ups that some of the top prop desks have in trading index options, for example, can be easily employed at home. The biggest perceived difference is transaction costs and the time to set up those systems. Sometimes extensive programming and development skills are needed to code a platform that easily extracts the implied vols from traded prices, to calibrate my models and then identify other "mispriced" options. But it can all be done.

 

[asiaprop]

...there is no guaranteed way to make money, no matter what. So even being able to buy options below fair value and selling those above fair value can lose you money. The reasons can be many, among others, liquidity issues, overleverage, commission, and so forth. While the identification of fair value and deviations of that is the right approach many other issues need to be considered.

Quote from Nanook:

Continuing with the dice analogy here is an excellent explanation (courtesy of Maverick74 on 04-19-06):

"Let's try this for a minute. Let's remove the whole idea of the greeks. Let's simply look at options as a bet on fair value. When I interviewed in Chicago for the first time to work for a market making firm on the floor, all the companies asked this question.

They said you are going to play a game with one six sided die. You roll the die, whatever number comes up, you get that much money. So if you roll a 3, you get $3. If you roll a 6, you get $6. You are going to play the game over and over again. There are two players in this game. The roller and the house. The house is selling the bet, or selling premium if you will. The roller is buying the juice. The question is, as the roller, how much would you pay for the right to roll the die. Then, when you are the banker, how much would you sell the bet for?

For most of you you can figure out this is a very simple probability game that you probably played in your stats class in college. The fair value of the bet is 3.5. You get this number by summing the outcomes and dividing by the total.

(6+5+4+3+2+1)/6=3.5

That means the fair value of this bet over a long series of throws is 3.5. So if you are the roller, you want to pay less then 3.5 for the bet, say 3.4. If you are the banker, you would want to sell this bet for more then 3.5, say 3.6. What you just did is you made a market. You are 3.40 bid at 3.60 offer.

Folks, this in a nutshell is what options are all about. It does not matter if you are the roller or the banker. The person buying the premium or selling it. If you buy the option below fair value, you will make money. If you sell it for more then fair value, you will make money. Market makers have been doing this since 1973 and nothing has changed since then except the technology.

The whole idiocy of buying vs selling premium is as bad as the whole red state/blue state garbage. If you can grasp this very simple concept, trading options will become much more simpler for you.

Yes, over the long run, the buyer of premium will generally outperform the premium seller for one reason and one reason only. Not because he/she is a better trader. But because of something called luck. That's right, luck. Luck, is very much like volatility in that it doesn't have a positive or negative bias. It simply is what it is. You will have both good and bad luck in your life. The difference here is that when you have good luck as the long premium trader, you might retire off of it. Bad luck to the option seller will bankrupt him/her. Good luck will do nothing for the option seller as there is very little upside in what they are doing."

http://www.elitetrader.com/vb/showthread.php?s=&postid=1044299&highlight=dice#post1044299

Everyone here seems to have their own definition of 'statistical edge'.

Placing any option spread is not a "set it and forget it" strategy. Anything can happen at any time and you need the "edge" (i.e., money management ability) to modify/adjust your position when and if needed.

 

[spindr0]

Quote from asiaprop:

Sometimes extensive programming and development skills are needed to code a platform that easily extracts the implied vols from traded prices, to calibrate my models and then identify other "mispriced" options. But it can all be done.

Everything that you say may be totally true but extensive programming and extensive development skills aren't usually within the purview of most of the retail world which has not spent their professional life in the pits, at prop desks or in related industry positions. Ergo, we end up doing things a bit differently.

 

[Bben1006]

You seem to forget that option contracts are merely created by a shake of hands of two willing parties--whereas, the equity market presumably reflects some precieve valuations.
~B

Quote from heech:

I am inclined to believe that stat arb does exist, but it's incredibly, incredibly rare... and it's certainly nothing as simple as buying or selling a put/call/calendar/butterfly spread. To have actual statistical "edge" that you can arb, you need a better-than-average continuous model for price and volatility. To put it mildly, it's more or less the holy grail.

For the vast majority of option traders... we're fundamentally no different than stock traders. The shape of the payoff function looks a little different, but the ultimate risk/reward ratio over time will look the same.

 

[jwcapital]

I think everyone has touched on great points. Selecting strategy appropriate to the environment and managing it is the "edge." Let's suppose you believe the S&P 500 E-mini will go down a hundred points. Now the key is to select and manage your strategy. One, you would like to pick the strategy that gives you optimum profit. Since you know what is going to happen, there can be no loss (or can there be?). It looks like the best trade would be to short the futures. Now, let's suppose we always use options in our trading. So, let's say we use the covered put strategy (short the futures and sell the ATM put). Your profit will be about 60 points. Bear put spreads and bear call spreads will work, but not make as much. A DOTM bull put spread will make even less. Interestingly, if you do a long put, you could actually lose money, for the market takes too long to go down. So, we are back to strategy--picking the right one and managing it.

 

[xflat2186]

Quote from timbo:

Well, you imply it's impossible to gain edge in any strategy -- albeit, favoring dmo. Who cares about hv (or any statics). I'm just saying edge doesn't depend on a number, but favors the expectation. It's forward, not backwards.

Why bring MM's or tbills into the convo and why take an cheap shot?

 

[heech]

Quote from spindr0:

Everything that you say may be totally true but extensive programming and extensive development skills aren't usually within the purview of most of the retail world which has not spent their professional life in the pits, at prop desks or in related industry positions. Ergo, we end up doing things a bit differently.

I personally doubt there's much alpha left in trying to arb the volatility surface. Everyone has pretty much the same option pricing models, and everything tends to move in line.