Just for clarification purposes, the paper was talking not about whether these things "actually have" free will, but whether they (in a precise sense) can reasonably believe that they do. They administer this "Turing test" to themselves.

It is slower, more prone to error, and requires humans to do work that we can just automate. Because of that, humans doing the task are just out-competed by computers programmed to do it. That said: algorithms can be kind of dumb sometimes, so humans still do it but just not as much percentage-wise, and you can still put up asks and bids, of course.

This is how it works, it is just done with computers now instead of humans, and over Ethernet instead of with sound waves.

It can be. Some exchanges run their own algorithms to help provide liquidity. It is just a separation of duties issue--some people are good at making really smart algorithms and want to do that, other people don't want to deal with the risk and algorithmic competition, so they just run the exchange. Also, as an exchange, you may run into increased regulatory issues if you're trading on your own exchange. I don't know. But if so, I could imagine some exchanges just not wanting to deal with that.

You're absolutely right. I was, however, talking about the hypothetical situation where we don't have any middlemen. But you're right, there can be other regular people who are using limit orders. Part of the issue is just in defining what, exactly, is a "middleman" in this context.

In your example, you say you are moving to buy a share of stock, and there are plenty of asks available. Some of those people are selling because they want to turn around and buy it back on the bids. I'd call those middlemen. Others are selling for other reasons, e.g. they are thinking longer term and don't intend to buy back soon or ever. Without the middlemen, you'd only have the latter. In general, those people are not traders, they're not all updating their orders all the time and outbidding one another, so there won't be as many competitive offers.

This means you're unlikely to get as good a price as you would with the middlemen--there just aren't as many offers to choose from. It sounds backwards that middlemen actually reduce your costs, but it is because they provide increased competition. People take the best prices available first, so there's demand for better prices. That's why human market makers could make a living in the first place, and it is why computers are now able to make money doing the same thing (albeit with much tinier spreads).

I know, it's just that I think a 10% spread gives the misleading impression that these traders are gouging people. In practice, they can't get a 10% margin, unless we're talking about some market with very low trading volume, because competition will see these great opportunities and just come in with better offers.

No, if the asks are at $10.501 and the bids are at $10.499, that would be a $0.002 spread. If it were the other way around, the exchange would *immediately* fill some orders, even before telling others about the overlap. A "proper" exchange should never give you an order book with an ask less than or equal to a bid, i.e. it should always leave a positive gap (the spread).

It's not better. In fact, that would suck. But that's not what happens. Instead, a trade will occur at the price of whichever offer was entered first. So if the bid was already at $10.501 (using your numbers) and then suddenly someone wants to sell at $10.499, instead of putting it on the asks-list, the exchange would just fill the sell order at $10.501. In this case the buyer pays $10.501 (the maximum he was willing to pay) and the seller gets $10.501 even though she would have accepted as low as $10.499.

The article is about computational time, i.e. computational complexity / number of operations, not actual physical running time. The number of computations a machine must perform to determine the output of another arbitrary machine (a description of which is given as input) must be, asymptotically speaking, at least as high as the simulated machine took, with the best possible algorithm being to just simulate it.

Nice try at shifting the burden of proof. You're the one wanting it outlawed. It is your job to prove that it actually "steals money from people".

With human market makers. They used to be known as specialists. People literally call out prices in a process known as open outcry.

Because without HFT middlemen, you have human middlemen. Human time is in higher demand than computer time. Computers automate the process and reduce costs. You can't trade without the middlemen unless you actually want everyone to be forced to sit and wait in line for a buyer when they want to sell or for a seller when they want to buy. One of those options is going to have a queue at any given point in time.

Presently, you can indeed wait if you so choose. However, middlemen provide you another option: sell directly to them, instantly, for a small fee. Market makers compete to give the best price, i.e. take the smallest fee. The exchange automatically matches you with the guy offering the best price. This used to be all done with humans, so the transaction costs were higher.

Because the traders aren't there at the exact same time, so they can't just trade with one another.

With that kind of spread, it isn't. This is why in actual exchanges the HFTs will compete to move the bid up to $10.499 and the ask down to $10.501. Now you can sell at $10.499 if you choose, or you can put an offer at $10.501 and wait for a buyer and hope the price doesn't fall while you're waiting. When the spread is that small though, many would rather pay the middleman to save them time from babysitting their order to make sure it fills.

Certainly! Off the top of my head, I'm sure that kind of improvement would have medical applications.

But as for microseconds instead of seconds: you do that because of competition. Your competition can do it in 1 second, but you can do it in half a second, so you're the one getting the overseas exchanges corrected. The overseas traders pay you because you got the information there first. Then it becomes an arms race.

You might think it is a pointless arms race, and claim it'd be better to just regulate it away. But aside from the political difficulty of getting all countries to agree to do your way, imagine you're one of those overseas traders being asked to willingly take bad prices for a second or two all the time. Sure, most of the prices individually are not that bad, but when you're a brokerage making millions of trades per day on behalf of millions of clients, it adds up, and then you've gotta pass those costs on to your customers.

And my implication was that it doesn't matter if it is liquid "enough" for the common person. You don't ban something just because it doesn't cater to the common person. You'd ban it because it actually hurt people. Neither you nor the OP provided any evidence that it actually hurts people, just a bare assertion.

I'm not sure what you mean. You can't put yourself in the middle of a market order. And if they are never an endpoint for a trade, they can never make a trade. All trades have a counterparty; if they make a trade, they must be a counterparty at some point, by definition.

The only thing I can think of is that you must be talking about limit orders, in which case yes, market makers can use limit orders to position themselves in the bid-ask spread. There is another style of trading which simply gives a single quote and everyone trades against that with a fee. That has the same effect as bid-ask (with the fee causing the spread) but you can probably see that it gives the market maker less control.

You wouldn't even need the SEC to do it. So long as it is legal and you comply with the regulations, just open up an exchange yourself based on this principle, and see whether traders like the idea or not.

I would like to point out one thing, though--it doesn't eliminate all middlemen. It just makes you the only middleman on your exchange, with nobody else competing with you. Not sure traders would like that.

Agreed. But, in your example, don't you think the datacenter company would rather grow to support more spots instead of cutting off paying customers? If feasible, they'll go that route. If they can't, they'll have to raise their prices and cut people off, until people start moving to other exchanges (i.e., the datacenter's competitors) who are in lower demand but charge less.

Today, market makers are HFTers. If you took out all HFT, you'd have a lot fewer orders. Ultimately, everyone would be paying more to trade, and people would have incentive to mirror HFT-like strategies manually as best they can manage. Seems like a giant waste of labor when we can just automate it.

Heh, good catch. I meant to say the rail transportation industry. My apologies.

I would not recommend buying high and selling low as a long-term strategy.

Yeah, it does. With or without HFT. Can you see how increased competition to provide you better prices results in less money "having to come from somewhere" ?

And I suppose you are the only person to whom people should be catering...?