A quicker move to stability in the micro scale lends to overall stability. From that foundation of stability a better price discovery results.
I think you are saying that having HFT around leads to a stable price, say, a second sooner. So if I wait a second, there shouldn't be a difference in this regard compared to if HFT was not around, or at least the non-HFT market won't be more than one second behind. Is that what you are saying?
I could turn this on its head and say that HFT increases market maker risk since the HFT traders could move against the market maker faster than the market maker can respond to them.
Almost exactly my point. Human market making is dwindling. It is moving into the deeper areas of the book for heavy volume trades, almost a niche market. The market of providing liquidity has a finite amount of room, which is almost directly proportional to market volume. Well over 90% of HFT is MMing, which means there's been a reduction in human MMing. Many humans have been burned in the scenario you suggest, also many HFTs, thus the large reduction in HFT over the past few years. The market is rebalancing due to many HFTs losing BIG.
I may be confused about what you mean by risk. You wrote that HFT decreases MM risk since with HFT you can avoid people moving against you by being faster than them. You also write that many market makers are losing big because they are getting moved against by other people using HFT. That seems like an element of risk to me. The only way I can make these two things make sense to me is to say that SOMEONE is fastest, and then that someone can do MMing without risk since no one else is fast enough to move against them. However, that wouldn't lead to tighter spreads as I see it, since that fastest person could then exploit everyone else trying to do MM by being faster, so no one else could do MM, resulting in no competition which is going to increase spreads. Costs also increase due to the overhead of maintaining the lead in HFT. I think I'm misunderstanding your point.
If the channels that pay to get to you are in the base plan, then the base plan would be more expensive without them. I don't hate those channels so much that I'd pay money to avoid having them delivered to me. I can just not watch them.
But that's not robot labor. You are doing the same thing the store clerk would be doing, so it's really about moving labor away from the store and into the customer, which when you put it like that doesn't sound so great for the customer. Robot labor would be if you just walked through the exit and the store would automatically have scanned your bags and your cell phone and charged you the right amount through that. The receipt would then be on your cell phone and you could take it somewhere to clear up any errors. I don't use self serve, but if they could make real robot labor like that work out well, that would be awesome. Even more awesome would be Amazon.com with Prime for groceries at low price.
I agree with most of your points. I'll say that having many access points is a benefit because then with the right tech you can use a lower signal strength (combating the increase in interference you mentioned) so adding more access points can increase bandwidth if there are many users in an area, even if it also increases interference. Using aimed beams makes interference much less of a problem. I also agree that some parts of the spectrum aren't likely to be useful for two-way data transfer, but we are not at the point yet where all the actually useful data spectrum is used by a uniform protocol to transfer data from any device to any other device, so there is still gains to be made on this front. If my TV is more important than your bittorrent, then QOS in the protocol is the answer for that. I do think we'll eventually be at a point where there will be no practical reason to use a wire for personal data access other than in unusual circumstances, because wireless will be good enough. I'll readily concede that likely there will always be applications where a wire is the right way to go.
Hi!:) well, I'm fine with you responding here, if Slashdot isn't up to continuing the conversation in the other thread. Might get an off-topic mod at some point, though.
As I understand it price discovery is for the market to arrive at a price of a good, while valuation is about determining what a good should actually be worth. I've always liked the quote "the market can stay irrational longer than you can stay solvent", which relates the two concepts. If it takes lots of interaction (and nothing else) between market participants for the market to settle down at some particular price, and HFT allows more interactions faster, then price discovery should occur faster with HFT, I agree. I just don't think that the price HFT arrives at is better than before, it is only faster. If there is a particular point you want me to take from the linked article, then I'll look for that. I did skim the abstract and conclusion. I'm not surprised that there is a link between HFT prices and news, but I'd be surprised if HFT prices aren't worse valuations than human speed prices are, since HFT prices in reaction to news have to be based on computerized reading of the news, which can't be all that accurate or insightful.
There are two separate issues there. One is replacing people with computers, which I think is great when it works well. The other is allowing trades to occur at the speed of computers, which is where I don't see the benefit.
Low latency allows for quicker reaction time to adjust prices, minimizing risk the market will move against you. If MMs can get out of a bad position, as a liquidity provider, much more quickly risk is reduced and parties are more willing to offer MM services for less profit, i.e. tighter spread.
I could turn this on its head and say that HFT increases market maker risk since the HFT traders could move against the market maker faster than the market maker can respond to them. Your argument seems to require an assumption that market makers are automatically able to perform HFT faster than the people moving against market makers. Without that assumption, it seems to me that risk might as well go up due to HFT as down. Am I misunderstanding your argument? As I understand it, this is your main argument and main point that shows that HFT is a boon, so I'd like to understand it.
I agree that banning something because of its opportunity cost is a slippery slope at the end of which we all end up in a prison where we are told exactly what to do each day. Also, it's hard for a regulater to know in the first place what the optimal thing to do is. So there is a balance. However, trading in markets is different from, say, growing marijuana in your garden for fun. The marijuana is only going to directly affect the person using it, and the resources spent are contained to the person growing it. So there is a strong freedom argument that banning marijuana is wrong because it infringes on our freedoms. Trading in markets is specifically about the economy which affects all people, and it's not just a place for people to have fun, it is all about everyone's money, jobs and savings. When markets go wrong, it affects everyone. So I'm sympathetic to the freedom argument that we can't go around banning every single harmful thing, but I'm not sure that that argument is so strong for trading in the market as it is for most other things like growing marijuana (which is never the less illegal).
According to Google, the world's gross domestic product in 2009 was $58.26 trillion=$58260 billion, so $141 billion is about 0.24% of the world's total yearly output. I think that's a tremendous amount of resources and more than I would have thought. I'm also not sure that only hedge funds do HFT, but perhaps "hedge fund" is defined broadly enough that that would be true.
Do what everyone else does with preprint papers in your field. I don't know what people do with preprints in your field, but a MS thesis isn't peer reviewed in the way papers are, so it fits nicely in the preprint category assuming you've actually done at least a tiny amount of work that qualifies as original research. Ask what researchers do with preprints in your field and then do that. Even better, ask your adviser what you should do with your MS thesis. If you are going to to publish papers with your master's thesis work in it, then you may want to downplay your thesis in favor of those papers.
The prediction is "will continue for at least 10 years", and it is made before density enters the picture. Consider the difference between law and mechanism for a while longer.
Well, net neutrality supporters want the government regulating everything, including the private networks of private companies that you as a customer merely pay for access to, so the inevitable result is going to be abuses like restricting torrent traffic to prevent "economic terrorism."
Please private update your outdated private Newspeak private dictionary. The approved private term is private now "money rape" private. It has always private been private "money rape" private. Private private private.
15 more years of Moore's law sounds good. Let's take it and reevaluate in 15 years:) Though, if the doubling is every 18 months, and we continue for 15 years, that is only 10 doublings, which is 1024 layers. I don't know how much time it takes us to print the surface of a single layer chip today (?), but if it takes an hour then that's only 42 days with no improvements. Also, if we can find a way to make the layers separately and stack them at a later stage, they can be made in parallel. I suspect that heat dissipation will be a limiting issue long before the time to produce becomes the main issue.
Thanks for the reference to the original article, which I had not read and now I have. The article is in two parts. The first part is an observation: as you increase the number of transistors on a single component, the price per transistor becomes a u-shape. If you graph the sweet spots of these u-shapes, you get an exponential decrease in price per transistor. Moore speculates that he sees no reason that the (exponential) trend shown on the graph on the second page won't continue for at least 10 years. That is an economic observation. The rest of the article is a technical discussion about the mechanism behind the exponential observation he has already made at that point in the article, and it is only there, after the law, that he gets into density.
To be perfectly accurate, the exponential observation Moore makes in that article is that the minimum price per transistor will be found at increasingly large numbers of transistors, although that doesn't actually imply that the price of those transistors will decrease exponentially. However, the graph on the second page makes that argument clearly enough, so I think it's fair to summarize Moore's law as saying that the cheapest cost-per-transistor obtainable will drop exponentially in the way described on the graph on page 2. The discussion about density is just about mechanism, it is just the way that Moore explains how his observation is holding true and can continue to hold true. The fundamental observation that we call Moore's law can also hold with other mechanisms.
Moore's law is not about density, it is about the price per transistor. Doesn't matter how you bring the price down, the law will still hold if it does indeed go down as the law predicts. Density has been the most important component of that so far, that you are right on, but it doesn't have to continue to be true. If you can stack slices of transistors on top of each other, and keep making that process cheaper, that will help with Moore's law too.
No magic necessary, we're just not there yet. First of all, use all the spectrum all the time instead of carving it up like we do today. Just need QOS to make sure emergency communication can get through. Requires a cheap way for small devices to send and receive at any frequency, which I suspect we haven't got yet, but I don't know of a reason that it shouldn't be possible eventually. Second, add access points everywhere to get a better signal at lower strength and shared with only the people in your immediate vicinity. For example we could install an access point in every light bulb. After that, you could try directed channels where the wireless device knows where the access points are and sends a focused beam only in the direction of the access point. That way the channel is not shared.
If it's really necessary, you could set up several directed wireless channels from the same device, though that would require some precision aiming. If ALL the spectrum multiplied by however many channels can be kept separate is not enough for your application, then then yes that would be hitting a fundamental barrier, though this barrier depends on the volume of your device since that determines how many separate channels can be sent to an access point. Wires have a similar limitation since you can only attach so many individual strands to a device of a given surface area. In any case, the bandwidth of such a system would be tremendous - far beyond anything you can do on wires today. So it's not a physical limitation, it's a technology and price limitation. No magic necessary, just R&D. I agree that wireless will probably be behind perpetually, but it's possible that at some point we'll have so much bandwidth and so low latency that it doesn't matter.
OP doesn't assume guidance in his post. If the energy requirement is such that a present trait is not worth it in the sense that people without that trait survive better because they don't have to spend the energy, all else being equal evolution will tend to remove that trait from the gene pool precisely because "it is not worth the trade off", as the OP said. The determination of whether it is worth the trade off is made through practice instead of by some deliberate intelligence, but that determination is still being made none the less.
So speeding up trading, which by natures means reducing latency, stops being beneficial at some point? Lower latency by it's very nature no matter how low the latency allows for better price discovery, see below.
Not better, faster.
Faster trading, by very nature includes lower latency, encourages competition over smaller amounts of profit. If there is a spread of 10 points on a security; somebody that can be faster means they'll cut it down to 8 or 9 points. Somebody who gets a little faster will then undercut them yet again. When the spread is one, getting there faster means you get filled before the other guy. With a spread of 1 or 2 you've reached a point of significantly diminishing returns.
Say Alice offers to sell at 2$ and buy at 8$. Bob wants to undercut Alice. So Bob offers to sell at 3$ and buy at 7$. How is speed of transactions relevant to this situation, or are you thinking of something else?
Any time it settles down faster is an increase in efficiency and by very nature lends to better price discovery. Market makers on the floor only act on the activity already in the market. HFT only does it faster. So the activity of human MMs doesn't add any real information?
I'll give you that if we all trade 1000x faster using computers, then prices might settle down less than a second faster. That has never been in dispute. The question is how that is useful. It seems to me that there is no difference if you are willing to wait a second to trade. the point is exactly "HFT only does it faster". Faster isn't better for the world, unless that speed is somehow useful outside of the competition of getting there first.
I wasn't posing it as a strong argument. It's *something* beneficial, in which you claimed HFT only harms. You've moved your argument away from the original point.
Crashing your car has the benefit that then you'll be driving in a brand new car. So that is a benefit of crashing your car. Yet crashing your car is still harmful, and pointing out the benefit of having a new card to replace it doesn't negate that. So if someone tells you that crashing your car is just harmful, it's not a good point to say that he is forgetting about the benefit of having a new car, except perhaps to make him feel better.
You are falling back on old arguments in which you have not provided any evidence. How does HFT harm? Nothing backing it up but what you believe. Please backup what you are saying, I'm very seriously interested in discovering the truth. The transatlantic cable is a large investment in which banks are lining up to pay. It's not all HFT, in fact much will be MFT that needs to react FAST with a few trades. You pose an interesting theory but again have nothing to back it up. Also the current pipe is clogged up something fierce. I wonder what the costs of laying a new pipe just for bandwidth would cost? Subtract that from the actual cost of the pipe and then you have your HFT difference, in which you have to subtract out the MFT stuff and other low latency applications.
I think you might not understand what I wrote. If HFT doesn't provide significant benefit to the world, yet significant money gets allocated to pursue HFT, then the harm to the world is the opportunity cost of those investments. That is an a-priori logical argument, it doesn't need anything to back it up, it's just logic. To put it more simply, suppose there were only two investments in the world, investment A and investment B. Investment A gives a return of 5% but it only benefits you. Investment B gives a return of 4%, but it also benefits the world (positive externality) by 20%. The smart money goes to A, and the harm of that is 19% of the value. So 19% is the opportunity cost for the world of allowing A. In this case A could be HFT or insider trading or something that pollutes. Obviously I just made these numbers up to illustrate the idea of opportunity
I do imagine that if the market was limited to trading to once a day, then increasing speed to once an hour could provide some benefit. That is not what we are discussing though - we are discussing going from a minute to a second to a millisecond to a nanosecond and so on. The difference is that a day and an hour makes a real difference to human being trying to do something even if that human doesn't care about being first. The difference of a minute to second doesn't mean as much, and a second to a millisecond doesn't matter much at all other than for being first.
Trickle down technology happens because money is flowing to HFT and HFT requires technology. If the money did not go to HFT they would go somewhere else and I don't see why that could not equally result in technology development. This is the broken window fallacy. In any case I doubt that a very large amount of the money flowing into HFT results in open technology that gives benefits beyond the HFT firms where the technology is developed, since HFT firms are in direct competition and revealing their technology puts them at a disadvantage. To put it pointedly, I doubt that putting money into HFT is a cost-effective way to improve Linux. If that were the goal, it would be much more efficient to impose a tax on trading and use the money to fund Linux development. So I don't think that trickle down technology is a strong argument in favor of HFT.
How does faster trading lead to a lower spread? I understand that computerized trading leads to lower costs which would reduce spread in a competitive market and I understand that increased competition would lower spread. How does faster trading do that, if I'm willing to wait a little bit to buy or sell?
It seems to me that HFT can't really help with price discovery. At most, it seems to me, it could make the price settle down less than a second sooner than it otherwise would, which is pointless. In any case, for the price to reflect something real, which is how price discovery leads to a benefit, the price has to reflect something real about the world. HFT acts only on the activity already in the market, so it doesn't add any new information, and if it does add information, say by reading the news very quickly, all that happens is that this information gets into the market less than a second sooner than it otherwise would. Is that wrong?
The harm that HFT does is that it wastes resources that could otherwise be allocated in an efficient way. Consider the recent Slashdot story about a transatlantic cable laid down to transfer information a few milliseconds faster between exchanges. I believe that a few milliseconds faster trades between exchanges doesn't benefit the world. If I'm right here, one effect of HFT is to extract money from the stock market without providing a benefit and then pouring it into projects like this cable that also doesn't do anything useful. Another way of saying it is that the harm of HFT is the opportunity cost of the resources that HFT extracts from the market.
It doesn't, the advantage comes in incremental movements where a few get ahead of everybody else. It's the nature of competition. You pose the question in such a way as to bias the outcome. It does not reflect real world mechanics.
The whole point of trading faster is to be further in front. Similar to racing cars, those that come in 1st win more than those that come in 2nd, 3rd, 4th... If you made all of them equally faster there's no point, but if the one in 4th makes some improvements and then is able to take 1st, he wins.
How does the HFT being further in front benefit the world, though? I think it only benefits the HFT trader. I agree that trading faster is a good idea for the people who are able to be faster than everyone else by doing so. I agree on this in the same way that I agree that insider trading grants a benefit to the person doing it. I also think that it is reasonable to think that investment in HFT has the side effect of developing technology that other people can benefit from such as Linux kernel patches. I also agree that care would be needed in developing a system to stop HFT, and just saying that trading can only occur at for example 1 second intervals may not be the best way to go. I also agree that if computers can trade as well as humans, then replacing some of the humans with computers is a good thing.
Where I don't agree is that having the opportunity to beat a competitor is a benefit to the world. Beating a competitor is simply allocating funds inefficiently if the basis on which the competitor is beat does not produce value. If an efficient farmer puts an inefficient farmer out of business, value is produced by giving the opportunity to profit from production to the more efficient producer. If a fast trader beats a slow trader, I don't see that kind of benefit. In fact I see harm, since, as you acknowledge, fast trading does not in itself produce value over slow trading, and perhaps the slow trader is actually doing smarter things, so he should be beating the fast trader.
In short, if I were a HFT I would want HFT to continue for my own sake. At the same time, if I were a market regulater, I would want HFT to stop because it only benefits HFT traders by extracting resources from the market and allocating them inefficiently. That is the exact same reason that I would want insider trading to stop. The function of the market is to allocate resources efficiently, and both HFT and insider trading are harmful because they disrupt that.
Your NDA discussion must have been with someone else. I have said nothing about NDAs. As for cars, exactly, the whole point is that making cars faster is useful, just as growing crops faster is useful. If we could grow crops at twice the speed, we'd get twice the food for the same effort. When the inefficient farmers go out of business because they can't compete with the efficient ones, that means that food is being produced more efficiently. Great. The problem is that trading faster doesn't result in something of value for the world that trading slower would not also accomplish. Rewarding fast trading only has the result of wasting resources in trading ever faster. Again, I ask you, if everyone who now trades in less than a second could suddenly trade in less than a nanosecond, at no additional cost, how does that help anyone? I say it doesn't help anyone because the only point of fast trading is to be faster than the other people, it doesn't matter how fast you are going - only that the other people are slower. Is that wrong? If trading is 1000 times faster for everyone, that helps no one to get ahead, so the outcome is the same. So making sub second trading faster doesn't do anything useful. Do you also think that moving from millisecond to nanosecond trading is pointless? (except for the people who ahead, obviously)
So building cars faster than humans with computerized machinery has no intrinsic benefits? That's the logic you are utilizing.
That's the opposite of my logic. I'm saying that sub-second trading is not intrinsically useful. Faster cars are intrinsically useful. In any case, you are talking about computerized trading, which is necessary for HFT but you can have computerized trading that is not HFT. In a market where you can only trade at a slow speed, you can still have computers involved or even to do the trading itself. It just won't matter as much how fast your computers and connection are.
The problem with HFT isn't just that it is too fast for a human to pull the brakes in case of something going wrong. The basic problem with HFT is that it is profitable without creating anything. So the HFT profits are just extracted from everyone else while contributing nothing of value. The best I've ever heard HFT apologists come up with is that HFT provides liquidity to the rest of the market. But a HFT trader won't buy something to hold on to for more than a fraction of a second (or by definition it wouldn't be HFT), if things go according to plan, so if I'm trading at a reasonable human speed of a few actions per minute, that does nothing to help me sell or buy. All that occurred was that some amount of money is now in the pocket of the HFT, and no benefit was provided to anyone else trading at human speed that I can see.
The problem here is that trading faster than humanly possible is pointless in itself - there is no intrinsic benefit. The only reason HFT traders want to trade at superhuman speed is that they themselves gain a benefit from trading faster than everyone else. Suppose tomorrow that we could somehow make everyone able to trade 1000x faster. Does that add anything of value to the world? I don't see it - but feel free to enlighten me (yeah, if it goes from a week to minute, then yes, that could be a benefit, but we are talking about sub-second trades here). Suppose tomorrow we make trades faster than 1 second impossible. I don't see anything has been lost to the world there either. Contrast to something actually useful like agriculture. If we could make agriculture 1000x faster, that would have a tremendous impact on the world, because agriculture is useful. The only effect of HFT I can see is to distribute wealth and effort in an inefficient way.
It's much like insider trading - as an insider you can do something that gives you an advantage in the market, yet the thing you can do isn't actually useful to the world. All it does is redistribute wealth to you even though you didn't do anything useful. That is not a good thing.
I don't know the exact people in question so I can't know what they did. I can only imagine an "ideal bean counter" who would indeed take something like that into account. I doubt that the difference in expected value is enough to justify a 50 year copyright, in fact I'm pretty sure it isn't.
Slashdot Mirror
A quicker move to stability in the micro scale lends to overall stability. From that foundation of stability a better price discovery results.
I think you are saying that having HFT around leads to a stable price, say, a second sooner. So if I wait a second, there shouldn't be a difference in this regard compared to if HFT was not around, or at least the non-HFT market won't be more than one second behind. Is that what you are saying?
I could turn this on its head and say that HFT increases market maker risk since the HFT traders could move against the market maker faster than the market maker can respond to them.
Almost exactly my point. Human market making is dwindling. It is moving into the deeper areas of the book for heavy volume trades, almost a niche market. The market of providing liquidity has a finite amount of room, which is almost directly proportional to market volume. Well over 90% of HFT is MMing, which means there's been a reduction in human MMing. Many humans have been burned in the scenario you suggest, also many HFTs, thus the large reduction in HFT over the past few years. The market is rebalancing due to many HFTs losing BIG.
Ref: MMs http://en.wikipedia.org/wiki/Market_maker.
I may be confused about what you mean by risk. You wrote that HFT decreases MM risk since with HFT you can avoid people moving against you by being faster than them. You also write that many market makers are losing big because they are getting moved against by other people using HFT. That seems like an element of risk to me. The only way I can make these two things make sense to me is to say that SOMEONE is fastest, and then that someone can do MMing without risk since no one else is fast enough to move against them. However, that wouldn't lead to tighter spreads as I see it, since that fastest person could then exploit everyone else trying to do MM by being faster, so no one else could do MM, resulting in no competition which is going to increase spreads. Costs also increase due to the overhead of maintaining the lead in HFT. I think I'm misunderstanding your point.
If the channels that pay to get to you are in the base plan, then the base plan would be more expensive without them. I don't hate those channels so much that I'd pay money to avoid having them delivered to me. I can just not watch them.
You are right that some of the labor is mechanized, which is also true when a store clerk does things. I feel for you if 20 minute waits is the norm.
But that's not robot labor. You are doing the same thing the store clerk would be doing, so it's really about moving labor away from the store and into the customer, which when you put it like that doesn't sound so great for the customer. Robot labor would be if you just walked through the exit and the store would automatically have scanned your bags and your cell phone and charged you the right amount through that. The receipt would then be on your cell phone and you could take it somewhere to clear up any errors. I don't use self serve, but if they could make real robot labor like that work out well, that would be awesome. Even more awesome would be Amazon.com with Prime for groceries at low price.
I agree with most of your points. I'll say that having many access points is a benefit because then with the right tech you can use a lower signal strength (combating the increase in interference you mentioned) so adding more access points can increase bandwidth if there are many users in an area, even if it also increases interference. Using aimed beams makes interference much less of a problem. I also agree that some parts of the spectrum aren't likely to be useful for two-way data transfer, but we are not at the point yet where all the actually useful data spectrum is used by a uniform protocol to transfer data from any device to any other device, so there is still gains to be made on this front. If my TV is more important than your bittorrent, then QOS in the protocol is the answer for that. I do think we'll eventually be at a point where there will be no practical reason to use a wire for personal data access other than in unusual circumstances, because wireless will be good enough. I'll readily concede that likely there will always be applications where a wire is the right way to go.
Free karma! :)
Hi! :) well, I'm fine with you responding here, if Slashdot isn't up to continuing the conversation in the other thread. Might get an off-topic mod at some point, though.
There are two separate issues there. One is replacing people with computers, which I think is great when it works well. The other is allowing trades to occur at the speed of computers, which is where I don't see the benefit.
Low latency allows for quicker reaction time to adjust prices, minimizing risk the market will move against you. If MMs can get out of a bad position, as a liquidity provider, much more quickly risk is reduced and parties are more willing to offer MM services for less profit, i.e. tighter spread.
I could turn this on its head and say that HFT increases market maker risk since the HFT traders could move against the market maker faster than the market maker can respond to them. Your argument seems to require an assumption that market makers are automatically able to perform HFT faster than the people moving against market makers. Without that assumption, it seems to me that risk might as well go up due to HFT as down. Am I misunderstanding your argument? As I understand it, this is your main argument and main point that shows that HFT is a boon, so I'd like to understand it.
I agree that banning something because of its opportunity cost is a slippery slope at the end of which we all end up in a prison where we are told exactly what to do each day. Also, it's hard for a regulater to know in the first place what the optimal thing to do is. So there is a balance. However, trading in markets is different from, say, growing marijuana in your garden for fun. The marijuana is only going to directly affect the person using it, and the resources spent are contained to the person growing it. So there is a strong freedom argument that banning marijuana is wrong because it infringes on our freedoms. Trading in markets is specifically about the economy which affects all people, and it's not just a place for people to have fun, it is all about everyone's money, jobs and savings. When markets go wrong, it affects everyone. So I'm sympathetic to the freedom argument that we can't go around banning every single harmful thing, but I'm not sure that that argument is so strong for trading in the market as it is for most other things like growing marijuana (which is never the less illegal).
According to Google, the world's gross domestic product in 2009 was $58.26 trillion=$58260 billion, so $141 billion is about 0.24% of the world's total yearly output. I think that's a tremendous amount of resources and more than I would have thought. I'm also not sure that only hedge funds do HFT, but perhaps "hedge fund" is defined broadly enough that that would be true.
Do what everyone else does with preprint papers in your field. I don't know what people do with preprints in your field, but a MS thesis isn't peer reviewed in the way papers are, so it fits nicely in the preprint category assuming you've actually done at least a tiny amount of work that qualifies as original research. Ask what researchers do with preprints in your field and then do that. Even better, ask your adviser what you should do with your MS thesis. If you are going to to publish papers with your master's thesis work in it, then you may want to downplay your thesis in favor of those papers.
The prediction is "will continue for at least 10 years", and it is made before density enters the picture. Consider the difference between law and mechanism for a while longer.
Well, net neutrality supporters want the government regulating everything, including the private networks of private companies that you as a customer merely pay for access to, so the inevitable result is going to be abuses like restricting torrent traffic to prevent "economic terrorism."
Please private update your outdated private Newspeak private dictionary. The approved private term is private now "money rape" private. It has always private been private "money rape" private. Private private private.
15 more years of Moore's law sounds good. Let's take it and reevaluate in 15 years :) Though, if the doubling is every 18 months, and we continue for 15 years, that is only 10 doublings, which is 1024 layers. I don't know how much time it takes us to print the surface of a single layer chip today (?), but if it takes an hour then that's only 42 days with no improvements. Also, if we can find a way to make the layers separately and stack them at a later stage, they can be made in parallel. I suspect that heat dissipation will be a limiting issue long before the time to produce becomes the main issue.
Thanks for the reference to the original article, which I had not read and now I have. The article is in two parts. The first part is an observation: as you increase the number of transistors on a single component, the price per transistor becomes a u-shape. If you graph the sweet spots of these u-shapes, you get an exponential decrease in price per transistor. Moore speculates that he sees no reason that the (exponential) trend shown on the graph on the second page won't continue for at least 10 years. That is an economic observation. The rest of the article is a technical discussion about the mechanism behind the exponential observation he has already made at that point in the article, and it is only there, after the law, that he gets into density.
To be perfectly accurate, the exponential observation Moore makes in that article is that the minimum price per transistor will be found at increasingly large numbers of transistors, although that doesn't actually imply that the price of those transistors will decrease exponentially. However, the graph on the second page makes that argument clearly enough, so I think it's fair to summarize Moore's law as saying that the cheapest cost-per-transistor obtainable will drop exponentially in the way described on the graph on page 2. The discussion about density is just about mechanism, it is just the way that Moore explains how his observation is holding true and can continue to hold true. The fundamental observation that we call Moore's law can also hold with other mechanisms.
Moore's law is not about density, it is about the price per transistor. Doesn't matter how you bring the price down, the law will still hold if it does indeed go down as the law predicts. Density has been the most important component of that so far, that you are right on, but it doesn't have to continue to be true. If you can stack slices of transistors on top of each other, and keep making that process cheaper, that will help with Moore's law too.
No magic necessary, we're just not there yet. First of all, use all the spectrum all the time instead of carving it up like we do today. Just need QOS to make sure emergency communication can get through. Requires a cheap way for small devices to send and receive at any frequency, which I suspect we haven't got yet, but I don't know of a reason that it shouldn't be possible eventually. Second, add access points everywhere to get a better signal at lower strength and shared with only the people in your immediate vicinity. For example we could install an access point in every light bulb. After that, you could try directed channels where the wireless device knows where the access points are and sends a focused beam only in the direction of the access point. That way the channel is not shared.
If it's really necessary, you could set up several directed wireless channels from the same device, though that would require some precision aiming. If ALL the spectrum multiplied by however many channels can be kept separate is not enough for your application, then then yes that would be hitting a fundamental barrier, though this barrier depends on the volume of your device since that determines how many separate channels can be sent to an access point. Wires have a similar limitation since you can only attach so many individual strands to a device of a given surface area. In any case, the bandwidth of such a system would be tremendous - far beyond anything you can do on wires today. So it's not a physical limitation, it's a technology and price limitation. No magic necessary, just R&D. I agree that wireless will probably be behind perpetually, but it's possible that at some point we'll have so much bandwidth and so low latency that it doesn't matter.
OP doesn't assume guidance in his post. If the energy requirement is such that a present trait is not worth it in the sense that people without that trait survive better because they don't have to spend the energy, all else being equal evolution will tend to remove that trait from the gene pool precisely because "it is not worth the trade off", as the OP said. The determination of whether it is worth the trade off is made through practice instead of by some deliberate intelligence, but that determination is still being made none the less.
So speeding up trading, which by natures means reducing latency, stops being beneficial at some point? Lower latency by it's very nature no matter how low the latency allows for better price discovery, see below.
Not better, faster.
Faster trading, by very nature includes lower latency, encourages competition over smaller amounts of profit. If there is a spread of 10 points on a security; somebody that can be faster means they'll cut it down to 8 or 9 points. Somebody who gets a little faster will then undercut them yet again. When the spread is one, getting there faster means you get filled before the other guy. With a spread of 1 or 2 you've reached a point of significantly diminishing returns.
Say Alice offers to sell at 2$ and buy at 8$. Bob wants to undercut Alice. So Bob offers to sell at 3$ and buy at 7$. How is speed of transactions relevant to this situation, or are you thinking of something else?
Any time it settles down faster is an increase in efficiency and by very nature lends to better price discovery. Market makers on the floor only act on the activity already in the market. HFT only does it faster. So the activity of human MMs doesn't add any real information?
I'll give you that if we all trade 1000x faster using computers, then prices might settle down less than a second faster. That has never been in dispute. The question is how that is useful. It seems to me that there is no difference if you are willing to wait a second to trade. the point is exactly "HFT only does it faster". Faster isn't better for the world, unless that speed is somehow useful outside of the competition of getting there first.
I wasn't posing it as a strong argument. It's *something* beneficial, in which you claimed HFT only harms. You've moved your argument away from the original point.
Crashing your car has the benefit that then you'll be driving in a brand new car. So that is a benefit of crashing your car. Yet crashing your car is still harmful, and pointing out the benefit of having a new card to replace it doesn't negate that. So if someone tells you that crashing your car is just harmful, it's not a good point to say that he is forgetting about the benefit of having a new car, except perhaps to make him feel better.
You are falling back on old arguments in which you have not provided any evidence. How does HFT harm? Nothing backing it up but what you believe. Please backup what you are saying, I'm very seriously interested in discovering the truth. The transatlantic cable is a large investment in which banks are lining up to pay. It's not all HFT, in fact much will be MFT that needs to react FAST with a few trades. You pose an interesting theory but again have nothing to back it up. Also the current pipe is clogged up something fierce. I wonder what the costs of laying a new pipe just for bandwidth would cost? Subtract that from the actual cost of the pipe and then you have your HFT difference, in which you have to subtract out the MFT stuff and other low latency applications.
I think you might not understand what I wrote. If HFT doesn't provide significant benefit to the world, yet significant money gets allocated to pursue HFT, then the harm to the world is the opportunity cost of those investments. That is an a-priori logical argument, it doesn't need anything to back it up, it's just logic. To put it more simply, suppose there were only two investments in the world, investment A and investment B. Investment A gives a return of 5% but it only benefits you. Investment B gives a return of 4%, but it also benefits the world (positive externality) by 20%. The smart money goes to A, and the harm of that is 19% of the value. So 19% is the opportunity cost for the world of allowing A. In this case A could be HFT or insider trading or something that pollutes. Obviously I just made these numbers up to illustrate the idea of opportunity
I do imagine that if the market was limited to trading to once a day, then increasing speed to once an hour could provide some benefit. That is not what we are discussing though - we are discussing going from a minute to a second to a millisecond to a nanosecond and so on. The difference is that a day and an hour makes a real difference to human being trying to do something even if that human doesn't care about being first. The difference of a minute to second doesn't mean as much, and a second to a millisecond doesn't matter much at all other than for being first.
Trickle down technology happens because money is flowing to HFT and HFT requires technology. If the money did not go to HFT they would go somewhere else and I don't see why that could not equally result in technology development. This is the broken window fallacy. In any case I doubt that a very large amount of the money flowing into HFT results in open technology that gives benefits beyond the HFT firms where the technology is developed, since HFT firms are in direct competition and revealing their technology puts them at a disadvantage. To put it pointedly, I doubt that putting money into HFT is a cost-effective way to improve Linux. If that were the goal, it would be much more efficient to impose a tax on trading and use the money to fund Linux development. So I don't think that trickle down technology is a strong argument in favor of HFT.
How does faster trading lead to a lower spread? I understand that computerized trading leads to lower costs which would reduce spread in a competitive market and I understand that increased competition would lower spread. How does faster trading do that, if I'm willing to wait a little bit to buy or sell?
It seems to me that HFT can't really help with price discovery. At most, it seems to me, it could make the price settle down less than a second sooner than it otherwise would, which is pointless. In any case, for the price to reflect something real, which is how price discovery leads to a benefit, the price has to reflect something real about the world. HFT acts only on the activity already in the market, so it doesn't add any new information, and if it does add information, say by reading the news very quickly, all that happens is that this information gets into the market less than a second sooner than it otherwise would. Is that wrong?
The harm that HFT does is that it wastes resources that could otherwise be allocated in an efficient way. Consider the recent Slashdot story about a transatlantic cable laid down to transfer information a few milliseconds faster between exchanges. I believe that a few milliseconds faster trades between exchanges doesn't benefit the world. If I'm right here, one effect of HFT is to extract money from the stock market without providing a benefit and then pouring it into projects like this cable that also doesn't do anything useful. Another way of saying it is that the harm of HFT is the opportunity cost of the resources that HFT extracts from the market.
It doesn't, the advantage comes in incremental movements where a few get ahead of everybody else. It's the nature of competition. You pose the question in such a way as to bias the outcome. It does not reflect real world mechanics. The whole point of trading faster is to be further in front. Similar to racing cars, those that come in 1st win more than those that come in 2nd, 3rd, 4th... If you made all of them equally faster there's no point, but if the one in 4th makes some improvements and then is able to take 1st, he wins.
How does the HFT being further in front benefit the world, though? I think it only benefits the HFT trader. I agree that trading faster is a good idea for the people who are able to be faster than everyone else by doing so. I agree on this in the same way that I agree that insider trading grants a benefit to the person doing it. I also think that it is reasonable to think that investment in HFT has the side effect of developing technology that other people can benefit from such as Linux kernel patches. I also agree that care would be needed in developing a system to stop HFT, and just saying that trading can only occur at for example 1 second intervals may not be the best way to go. I also agree that if computers can trade as well as humans, then replacing some of the humans with computers is a good thing.
Where I don't agree is that having the opportunity to beat a competitor is a benefit to the world. Beating a competitor is simply allocating funds inefficiently if the basis on which the competitor is beat does not produce value. If an efficient farmer puts an inefficient farmer out of business, value is produced by giving the opportunity to profit from production to the more efficient producer. If a fast trader beats a slow trader, I don't see that kind of benefit. In fact I see harm, since, as you acknowledge, fast trading does not in itself produce value over slow trading, and perhaps the slow trader is actually doing smarter things, so he should be beating the fast trader.
In short, if I were a HFT I would want HFT to continue for my own sake. At the same time, if I were a market regulater, I would want HFT to stop because it only benefits HFT traders by extracting resources from the market and allocating them inefficiently. That is the exact same reason that I would want insider trading to stop. The function of the market is to allocate resources efficiently, and both HFT and insider trading are harmful because they disrupt that.
Your NDA discussion must have been with someone else. I have said nothing about NDAs. As for cars, exactly, the whole point is that making cars faster is useful, just as growing crops faster is useful. If we could grow crops at twice the speed, we'd get twice the food for the same effort. When the inefficient farmers go out of business because they can't compete with the efficient ones, that means that food is being produced more efficiently. Great. The problem is that trading faster doesn't result in something of value for the world that trading slower would not also accomplish. Rewarding fast trading only has the result of wasting resources in trading ever faster. Again, I ask you, if everyone who now trades in less than a second could suddenly trade in less than a nanosecond, at no additional cost, how does that help anyone? I say it doesn't help anyone because the only point of fast trading is to be faster than the other people, it doesn't matter how fast you are going - only that the other people are slower. Is that wrong? If trading is 1000 times faster for everyone, that helps no one to get ahead, so the outcome is the same. So making sub second trading faster doesn't do anything useful. Do you also think that moving from millisecond to nanosecond trading is pointless? (except for the people who ahead, obviously)
So building cars faster than humans with computerized machinery has no intrinsic benefits? That's the logic you are utilizing.
That's the opposite of my logic. I'm saying that sub-second trading is not intrinsically useful. Faster cars are intrinsically useful. In any case, you are talking about computerized trading, which is necessary for HFT but you can have computerized trading that is not HFT. In a market where you can only trade at a slow speed, you can still have computers involved or even to do the trading itself. It just won't matter as much how fast your computers and connection are.
Substitute "government" for "voters". Now realize that those same voters are also the customers being competed for.
The problem with HFT isn't just that it is too fast for a human to pull the brakes in case of something going wrong. The basic problem with HFT is that it is profitable without creating anything. So the HFT profits are just extracted from everyone else while contributing nothing of value. The best I've ever heard HFT apologists come up with is that HFT provides liquidity to the rest of the market. But a HFT trader won't buy something to hold on to for more than a fraction of a second (or by definition it wouldn't be HFT), if things go according to plan, so if I'm trading at a reasonable human speed of a few actions per minute, that does nothing to help me sell or buy. All that occurred was that some amount of money is now in the pocket of the HFT, and no benefit was provided to anyone else trading at human speed that I can see.
The problem here is that trading faster than humanly possible is pointless in itself - there is no intrinsic benefit. The only reason HFT traders want to trade at superhuman speed is that they themselves gain a benefit from trading faster than everyone else. Suppose tomorrow that we could somehow make everyone able to trade 1000x faster. Does that add anything of value to the world? I don't see it - but feel free to enlighten me (yeah, if it goes from a week to minute, then yes, that could be a benefit, but we are talking about sub-second trades here). Suppose tomorrow we make trades faster than 1 second impossible. I don't see anything has been lost to the world there either. Contrast to something actually useful like agriculture. If we could make agriculture 1000x faster, that would have a tremendous impact on the world, because agriculture is useful. The only effect of HFT I can see is to distribute wealth and effort in an inefficient way.
It's much like insider trading - as an insider you can do something that gives you an advantage in the market, yet the thing you can do isn't actually useful to the world. All it does is redistribute wealth to you even though you didn't do anything useful. That is not a good thing.
Good point.
I don't know the exact people in question so I can't know what they did. I can only imagine an "ideal bean counter" who would indeed take something like that into account. I doubt that the difference in expected value is enough to justify a 50 year copyright, in fact I'm pretty sure it isn't.