Why predict?
- Thread starter EmRock
- Start date
nitro
OK, so this drew no comment from you. If it's too basic, that's fine - your question could be interpreted in many ways after all.
What if I offer you the range 48-52 in 100 throws? Would you take the bet? (Turns out to be close to the 60:40 you talk about.)
Against the Gods is essential reading, imo.
That is a bad bet as the number of throws approaches infinity, and it is at best a phantom edge. The answer I was looking for is: set the odds for the payout - as stated in a previous post.
The entire idea is that no matter how you wiggle and squirm, you can't wring and an edge out of something random. You have to look at the structure of the game itself.
The one thing that is not the same when comparing coin tossing with markets is that, each coin is unconditionally independent on any other coin toss before or after. That is not true of markets, so there is something to exploit there.
nitro
Yes, it is a bad bet: it turns out that you are offering 36% chance and getting paid as if it were 50% (if anyone would take this bet).
Isn't this what you were getting at - mispricing? Knowing something about the structure that for some reason your counterparty does not know or has mispriced. (Again this raises questions as to why you are right and they are wrong.)
You say that you cannot get anywhere without knowing the expectancy. Then you say that market prices are not unconditionally independent. So expectancy can never be known. Isn't this a bit of a problem for you?
Hi Nitro,
Let me try to summarizes what I've got from your reply - The "edge" of an option MM boils down to the following three things:
1) Better volatility estimation model
2) Better executions (ie. better managing legs and hedging risks)
3) Having a better vol fitter who's basically the Paul Tudor Jones in volatility betting
Am I right to say that?
I fully appreciate your proposed framework of trading odds instead of trading the future, and I understand your point made about focusing on payouts as it is a built-in element for all my trading. What I'm confused is the research direction in getting that favorable odds. All my "edges" are derived from tedious manual back-testing of market ideas (empirical research if you will) instead mathematical modelling. I have very little knowledge in quantitative finance and I'm wondering if this is something that I should try to learn more in the future for the continuity of my trading career.
The one big question I have is whether it is possible to really accurately estimate volatility or return distribution. I have little training in quantitative finance and my limited knowledge comes from the William Eckhart interview in Market Wizard where he claimed the market has infinite variance. If what he say is true then I assume the standard error is impossible to accurately estimate and eventually it all comes down to having a good vol fitter with his assessment of volatility which may or may not be right all the time (ie. selling insurance at below cost like being on the wrong side of 87'?). Is this the case?
Do you think all this would ever come to a point where competition is so fierce that eventually profits are reduced down to how good the vol fitter is? Without a good vol fitter wouldn't the whole thing simply be reduced to earning the tick spread (in the case where everyone has similar volatility estimation models)? In the latter case of earning tick spreads it would simply be a matter of who has the best information technology infrastructures and the fastest (and lowest cost) executions?
The reason I ask is that if eventually all the math doesn't matter (ie. complex math and execution edges being arbitraged out by other complex math edges from competitors) and profit eventually depends on the art of the vol fitter then I would spend more time on developing the "art" side of things like "feeling the market". Of course the math is always important in getting a more complete perspective. What do you think?
Thanks for your posts. You have definitely expanded my thoughts greatly.
Hi,
1) Yes. But it is not so much a better estimate from the model. It is that the model is very flexible and is adjusted by a human being to fit what you see. The software has to support that.
2) It is not just better executions. No one is coming to laputa to make a two sided market, 1000 up, on the 1400 September SPX put. In other words, you have to be standing in the pit to do that. You could make markets in electronically traded options like IWM or ES, but it is not the same thing because at least as important as anything else, the order flow you see coming into the pit dictates how you fit your vol. So yes execution, but equally, you have to be at a place where institutions know they can come to you to take the other side of their trades, and not 50 lots.
3) You may be able to get away from having a great trader be your vol fitter. I am not sure. But he has to know what he is doing for sure.
It's not your fault really. You come to a site like ET where there are supposed to be knowledgeable traders. You see that 99% of the threads are dominated by things like this MACD, or that pivot, blah blah blah. You go to the local bookstore and you see technical analysis of stocks and commodities magazine, or active trader, and all of them push these sorts of trading styles. Further, you come to ET and you see people bad mouthing academics, saying that they live in ivory towers and that the markets are not efficient abnd they don't know what they are talking about blah blah and more crap. Take the book, "A Mathematician plays the stock market" by Paulos. It is a book that should be read carefully, maybe several times. But people will not like the conclusions that he makes, and will dismiss it altogether. The point is Paulos understand the math all too well, but they draw the wrong conclusions from their analysis!
First, what you are doing now (searching for edges by backtesting etc) is not a bad skill to have, it's just that it is number 10 and there are 9 others that are far more important. I cannot stress enough that taking tradestation etc and trying to come up with indicators/systems that then get back tested is a dead end if that is where you begin and end. You may find statistical anomalies that way, and they may make money, but you will also lose money, and probably lots of it. That is the reality you have set for yourself when you attempt to trade this way. Remember what I said above, trading, when done right is nearly as certain as getting a paycheck over the course of two weeks to a month.
I would venture to say safely that the only thread worth reading on ET is the Don Bright thread. Why? Because it is the one example of a system that has worked for decades, and if you dissect it carefully, you begin to understand how to uncover other edges. Unfortunately, it is spread out over many years on many threads. The other thread I would look at are the pairs trading threads. Again, don't stop there. Expand into the academic research. Read the Vidyamurthy book, read it twice or three times. Look at the references. Read those. Read about how to deal with situations like we have now where the cheap companies are getting bought out by the expensive ones, creating greater risk for pairs traders. See how the theory is put into practice. If the math is too hard, don't throw your hands up. Sit in on a community college and get the training. There is no other way.
It is a hundred times easier to estimate vol than it is to estimate the direction of the underlying. That is well known.
It is well known that power laws in volatility is best estimated by modeling the tails correctly. Eckhart just gave up. Also, these people assume a static view of vol. The modern trend is to do some sort of dynamic programming. Hence my second concept above, modeling the volatility of vol correctly.
It is possible for a MM firm to lose money, if the event takes place overnight where delta/gamma/etc hedging cannot take place. Also, the event would have to be huge.
I doubt it. The gains may be less spectacular, but they willl always be there because MMs don't make markets unless they have a built in edge on execution. Otherwise, why the heck do it? Institutions are more than happy to have liquidity providers take on risk to have their insurance needs met. But I do see a tightening of spreads, especially as more options markets go electronic.
You underestimate what picking up a dollar here and a dollar there when the size you are quoting is 1000 up or more, in the SPX, where the size of the contract is 5 times the size of ES. Traders work for their money. They are constantly looking for the next deal. MMs do this hundreds if not thousands of times a month. Some trades win some lose. But nearly as certain as the sun rises in the east and sets in the west, MMs show profits month after month, year after year, decade after decade. The vol fitter is important, but you overestimate his importance. As important is the fact that you stand there and are willing to make markets. Just showing up is 80% of it. Thing is, to show up costs millions of dollars just to play.
These people are already rich. They could close shop and retire and golf all day long, drive Ferrari, go to parties with a model on each arm, and retire to their yatch or mansion. Yet they prefer to get up at 6:00 am and go and hang out with a bunch of guys and play the game.
The math is the heart and soul of options trading. In fact, the math is at the heart and soul of every profitable system I have ever seen. Whether the practitioner knows that or not is another matter, but if you look carefully, underlying it is some sort of edge that is often wrapped in mystical baggage. A good exercise is to think of why a trend line break has a better than a 50% chance of working (forget for now that it may be self-fulfilling. Assume there is a mathematical reason for the edge). You don't have to know engineering to drive a car, but you do have to know engineering if you want to build cars. You are trying to build cars, not just drive them. Catch my drift?
People resist the truth. I have said countless number of times that the math is what makes it work. In fact the vol fitters intuition is of a mathematical nature, the least of which because he understands the limitations of the model. The art of the vol fitter is the icing on the cake. We don't underestimate his role, but he is but one cog of the wheel.
nitro
You assume mispricing occurs because you know more than the other party. Like I said on a previous post, why does a single share of stock trade? Does the value of GE change every ten seconds. Ten minutes? An hour? Have you never wondered about this?
The number one reason the price of an asset moves, assuming no news, is because of liquidity demands by institutions. The institution on the other side of the trade may have many reasons to sell or buy the asset, driving the price way out of whack. They are not worried about right or wrong. They want the stock and they wanted it now, or they need to sell the stock and when the decision is made, it is gone. Stand in front of an institution and sell short to them when it wants INTC and it wants 100M shares. See what happens to you. It is this distinction, and probably only this distinction, that you need for a profitable system: figure out when the move of an asset is based on information and is the start of a trend, or when it is liquidity driven and is temporary. That is the holy grail.
I don't understand the question.
nitro
In the absence of the individual(s) who adjust volatility, based on their perceptions of what needs to be considered to permit an efficacious adjustment, would your model/system still be profitable? Is the adjustment provided by the vol fitter a stopgap measure, a temporary fix, an iteration, or none of the aforementioned?
lj
I doubt it. The experiment has never been done AFAIK. The current theory is that it may be possible to automate the vol fitter (vf). I have my doubts. The vf is constantly adjusting for what he thinks is important, and just like any other trader can get "whipsawed" [the model]. The key to all options trading is you make markets where you can hedge. So even if the vol is off by a little bit, there is usually some room for error.
BTW, the vf does more than adjust vol, as these models have more than one calibrateable parameter. But you get the idea.
nitro
Read some of the SFI papers; they hit all these topics.