The FX market is all algos. Just try going to indeed.com and looking up FX dealer positions. Almost non-existant. In fact, the FX market was the FIRST market to go all algo, way before equities. Sure there are still guys in the exotic derivatives space but those are structured products. Those old youtube videos you see with dealers screaming in the phone because Soros is selling Pounds in 1992 are long gone. Nobody is getting run over looking for liquidity desperately trying to hedge. It's sad in many ways, but the market is different today. I think this is also what makes daytrading very challenging and why I said I chose to fade moves back when I focused more on daytrading. Granted if we go way back to my stock trading days (2000-2002) yes I was a momentum equities trader. We were picking off people. But those days are long gone. You know I told this story on here a while back about how TOS put a lot of option MMS out of business. Back in the "old" days when options traded in fractions, you know people in general are bad at math and even more horrible when it came to adding and subtracting fractions. My first interview that's all they did was had us add and subtract fractions in our head as fast as possible because that was such an important skill. Anyway, back in the late 90's when stocks were moving very fast, retail traders always wanted to buy and sell at the mid but the problem was they had to figure out where the mid was and do it very fast. Say the AOL Dec 90 calls were 3 1/16 bid at 3/5/8 offer. Where is the mid? Most retail traders would always get it wrong and MMs loved that. Then TOS came along and said, we'll give you the mid price. So now when you put in an option order, even a complex 4 legged spread, TOS will tell you to the penny what the mid is. Well this was no fun for MMs because now all these people who were horrible at math knew the exact mid price of a 4 legged option spread. LOL. My point here is, technology has changed everything. You no longer needed a 26 year old kid who lived in Lincoln Park on the north side of Chicago quoting bids and offers for Susquehanna, you could now have a computer do it as well. So now algos are making markets and retail traders have very sophisticated tools to improve their pricing. Grandma can now use very sophisticated VWAP algos at IB to work her buy order in AAPL all day while she picks her grand kids up at school. People are just not getting run over anymore and that is why vol in general has really come down. Sure, you will spikes like what we had in Feb, but look at how short lasting they are. What use to take 6 months to resolve now gets resolved in days. The markets are faster, more efficient and market players have more info then ever to make decisions very quickly. Good traders will find ways to adapt to this. Bad traders will trade their capital for lessons that will be very expensive. You HAVE to adapt.
Mav, I have been meaning to ask about this for a while. With large players being able to hide so efficiently, have you noticed them getting better at hiding from the number lines?
I wasn't talking about Soros and that sterling collapse. But I admit I was talking about dealers handling the orders from their clients as they come. My understanding is that not much has changed. Yes, many market participants now have direct access to the market so they can run their own positions on the electronic platform, either manually or through the algo. But it is the real money flows, clients that don't want or don't need own currency desks (so they either call in to execute right away or put their orders into dealer's book), the big transactions that require MM to take the order from the clients and put it through the system to lay it off. Remember, MM doesn't know whether client wants to buy or sell, but he still needs to quote the price for the big ticket order. I hope I won't sound like someone that disagrees, but since you know so much, how are banks handling big fx orders from their clients? If the client that has big M&A to do and requires tons of foreign currency, calls the bank for the price, what are they telling them? That they will run the algo and will see what the execution will be when it finally gets done? Or if the client want's to know the price upfront that they can put the order at that price but it may not fill (if the market moves away)? I'm not asking out of curiosity, this is the fundamental knowledge that I would like to have in order to better understand the particular market that I'm trading.
Nobody needs to do anything quickly. The biggest players in the FX market are actually companies like GOOGL and GE and KO. They have to hedge their cash flows in 55 different countries. But they do it gradually over time. They are not "calling in to move it fast". Again, those old videos on youtube where during a time and place where currency "speculation" was very high. Banks use to have 400 guys on the FX desk. Now they have 2. It's just a different world. If I wanted to move 100 million Euro right now I could do it one swoop at probably 2 ticks over ask. I trade a lot of FX and use to trade a lot of FX and I've had to adapt to these changes and use ACD in a way that fits well for FX which is different than how I use ACD for trading energies. Again, the key term here is to adapt.
It's a good question. I think with vol in general coming down it has changed the NLs. The good thing is that NLs are adaptive to these changes. With vol dropping, the OR gets smaller and the A levels get smaller so on a "relative" basis the score is still accurate. But the difference is in the old days if you showed me a +24 NL for Crude Oil, that baby would be smoking. You could see 10 to 15 pts in a straight line. Today if CL has a +24 you might get a move from 61 to 64 over the course of a month with a lot of back and forth. Again, the key takeaway here is to make adjustments.
Here is a picture of the before and after of one of the largest trading floors in the US which was UBS in Stamford, CT.
Reading this thread I came across you post and it touches on a subject that that has been on a back of my mind for a long time. There is that big discussion on whether markets are random (whether due to efficiency or whatever) and sometimes I think it really depends who you are or how you look at the markets, similarly like Newtonian vs Quant physics. Let me just state clearly: I don't believe in random markets. With that being said though, I totally disagree with the way Fisher( I think I might have seen similar explanation at other venues) presented his case in the book based on those 5min or whatever it was time frames . Basically he claimed that markets are not random because ORs are statistically more significant (using time-basis) than any other time of the day. It makes no sense to me to prove it that way, it is meaningless to me. On intraday basis time is irrelevant, the market either moves up or down (think in terms or renko charts). Yes time is relevant in bigger sense because markets open and close, but anything in between is business that needs to be done that day. Now, anybody subscribing to the way Fisher proved it in the book, I urge you to repeat as many times as you want an experiment: take a coin, decide which side is uptick/downtick, play that coin and plot on piece of paper the price. You know it is random PA because of the coin flips are random, then see if what you plot is any different from the charts you see everyday on every timeframe. Feel free to perform statistical analysis of ORs for that random distribution and I'm sure you will get to the very same conclusion that they are significant. How could that be if the distributions are random per your coin flips? The ORs (even despite random coin flips) are still more significant than any other range later in the day (remember to disregard intra day time frames), but they are important for a very different yet fundamentally simple reason: The Opening Range (which is opening print and its vicinity) is statistically significant because... this is where the price MUST trade and will trade before it can get to any other price range at any later time. And how far it goes, whether it goes, which way it goes, it is unknowable until after the fact. So in my opinion, proving that ORs are statistically significant in fact doesn't prove anything about randomness or not of the markets. So does that mean the markets are random? No. For one thing fundamental, markets don't trade by themselves, they are based on human decisions and emotions. That is not random unless someone proves that human brain is nothing short of flipping coin or dice. Maybe it is for some part of the population, but not the majority. And therefore Mav's comment: is in my opinion spot on correct. The market will trade through OR because it must to and will go one way or another or both (one after another). And as a trader you just need to capitalize on these directional moves. And because there are people and their emotions behind the moves (even if its algo it had to be programmed) there is only so much capital that can be made or lost in any given day, only so much pain can be inflicted any day. And therefore the inner beauty of ACD, it forces you to start not at some imaginary arbitrary level but at the simplest level of all, the opening price range, a level that must and will trade, and as the price moves from there maybe it works for you on the way up, or maybe on the way down, in worst case you get seesawed and lose only a little bit (hopefully).
A few comments on randomness. From an economic perspective, the issue is not whether markets are random, it's a matter of how efficient they are. The most commonly held view (and the one I agree with) is that markets are "mostly"random. This should make intuitive sense if you think about it because none of us trade every product, every minute of every day. We would if there was an economic benefit. Clearly by the fact that we are selective shows that we believe that markets are mostly noise and we are waiting patiently for signals. So this lines up with the theory. With regards to how efficient the market is, it's not a question of whether one makes money. If there are risk premiums built into markets (and most have them) then any idiot should be able to capture it since it's being offered. The question is once accounting for risk, does one beat the market. Again, in economics we typically use something called a naive model to test this hypothesis. A naive model in many cases is either simply capturing the risk premium offered or simply choosing the mode, i.e the most common outcome. For example, the long term return in the S&P 500 including dividends is about 10% a year. The long term sharpe ratio is about .50. The sharpe is really what we are after here as it normalizes for leverage and risk. Most investors (and traders) do not beat the long term sharpe of the S&P 500 over long periods of time. They may beat the market in units of return but that has to be normalized. If the markets are "relatively or mostly efficient" it would stand to reason that once accounting for risk, most market players will not earn a superior risk adjusted return (and they don't). What about option sellers? Same thing. The naive model that we will test here is to always sell premium (since it produces the most common outcome). So we run a model that simply sells say ATM S&P 500 options every month and record the "risk adjusted returns". If we then look at Trader A who has a "so called proprietary" method of selling premium the question is not whether he/she makes money, but do they outperform the naive model on a risk adjusted basis. Again, usually they don't. In commodity markets (where I mostly trade), they are generally more random meaning it doesn't make sense to use a long only model or short only model as your naive comparison. Rather we would want to run a random monte carlo simulation where trades would be placed randomly. We could do 500 to 1000 runs, take the avg return and the average variance and generate a sharpe. The expected value of this would be around zero. Therefore any strategy that generated any positive return would be noteworthy and if accounting for opportunity costs as well, then it could be considered economical. So to sum up, one should expect the market to be mostly random otherwise you should be trading every product available 24/7 to extract the most amount of profits. Clearly one of the reasons we use ACD and more specifically the number lines is to attempt to extract the "rare" signal from the "common" noise. THAT is our goal.
But efficient at what? There are only two ways for the market to go (irrespective of time) and that's up or down and you are either right or wrong with your predictions. People base their decisions on many things, mostly emotions, including fear and greed. It doesn't matter (to me) whether they trade one product or all, once in a while or 24/7. So as long as people are trading the markets the markets are not random. Just because there are millions of participants and we can not say what will happen next or we can not (in time) comprehend who is the weaker side in order to capitalize on the move, doesn't make the markets random. That's why I say let's leave economic perspectives for economists... last time they though they knew it all about the markets LTCM happened... Unless someone proves that human decisions are random (they might look like from time to time), markets are not random. Taking millions of independent not-random individuals and bundle them together into one (market), might sometimes look like randomness, but in my opinion it is a convenient excuse for the obvious inability to measure complex systems. Now finding a way to capitalize on it that is totally different aspect. But lets not blame on "randomness" our deficiencies to solve this complex puzzle.
It's not a deficiency argument, it's a mathematical one and one that can be proved. Efficient markets are generally hard to forecast because markets are a discounting mechanism. The best explanation for this was John Maynard Keynes. In the 1930's he explained what speculation is and its challenges. He said it was like predicting the winner of a beauty pageant. But the catch was, it wasn't about who you thought the winner should be but rather who you thought everyone else thought the winner should be. In other words, the answer might be obvious to you but your objective is to figure out who everyone else will choose. So imagine now whoever can predict the winner will get a lot of money and imagine that this becomes an iterative repeating process. Meaning, the game gets played, everyone sees the results and we play again. Each time we are learning about the others choices. So now at some point we have to predict who we think everyone else will predict on who they think will win based on what they think based on what the next group will think and so on over and over. This is known as a discounting mechanism. In other words, we have to immediately remove the obvious and we have to discount 5, 10 or even 20 steps ahead. It becomes infinitely complex. So when we talk about efficiency, what we are referring to is the markets ability to discount the future. A perfectly efficient market would 100% discount all future events to the point they would have no impact on the market. Very few people believe this obviously. But we do believe there is some middle ground where the market is discounting much, but not all of the future. How much we don't know. Can we measure this? Why of course. Just look at trader performance. If most market participants could accurately forecast the future they would. LOL. We can tell by looking at their performance, again on a risk adjusted basis, that they can't. And since most of us all have "most" of the available information available to us in the form of real time news feeds, charts, prices, economic reports, etc, we all see it, yet most of us make the wrong choices. That must mean that all the things we are looking at are for the most part already priced in, especially once adjusted for variance of our errors. Again to sum up, this is not to say the markets can't be forecasted. The question is to what degree and in what magnitude in terms of units of risk. Reminds me of a guy on ET (I won't name him) he started a journal. It was during the credit crisis. He purchased or I should say "made a call" that he was long the ES at some level, say 1300. Over the next 6 months the market traded against him down to say 1000, but eventually came all the way back and he was back in the black and called the exit with a slight profit and declared victory. Technically speaking, he predicted the future! But the trade which was suppose to last a few days or a week ended up lasting over a year and he took 300 pts of heat to make about 10 handles. His forecast was accurate for sure. But again, the question is, after accounting for the risk he took and how long it took to make back his losses, any reasonable person would not have declared victory. And that is where the subtly is. Yes, you are right, the answer is binary in the most simplistic form, up or down. But the real answer is far more complicated. How much risk are you taking, how much time are you wasting and what opportunities are you giving up? That makes the question far more complicated then simply saying it's either up or down. The word "efficiency" speaks to this process. How "efficient" are you at extracting value from the market. I can take a tree down in the forest with a hammer, but it's not terribly efficient to do so.