The article is misleading. Please don't go around saying "AI beats top progamers at Starcraft". For one, the player they mention, Oriol, was not a progamer. The article does not say so either, but articles quoting the article seem to. They say

There seems to be confusion about the name of the player. The player that the UCSC article refers to, =DoGo=, indeed participated in WCG 2001 finals for Spain, but his name was Antonio Crespo Gomez.

Who knows what the context really was? Maybe the developers asked him to try a specific build order in order to see how the computer would respond. Maybe he did legitimately lose one game... out of a hundred, the only reason being he was forced to only use his mouse.

In any case, the biggest complaint I have is that he was a good player back in 2001. That was before the invention of mutalisk micro, the macro-oriented plays led by iloveoov, the micro-oriented strategies pursued by Boxer -- basically, before people figured out how to really play the game. The AI researchers undoubtedly utilized a lot of modern strategies. Also, no matter how good someone used to be, it's hard to be as good ten years down the line, even with crammed practice. I'd bet anyone that no computer in the next couple decades would beat actual progamers of today (by which I mean an A-team member of a progaming team in Korea, in a best-of-five).

Don't take this post as bashing the research -- I think it's amazing what they've been able to do. Just don't compare it to Deep Blue vs. Kasparov, because it's closer to Deep Blue vs. Middle School Chess Team Captain.

My last sentence may have been a cryptic.

What I meant was that a lot of continuous market making is managing fleeting imbalances of supply and demand that do not really reflect the stock's fundamental value. As a result, specialists really are not helped by fundamental analysis. They're just hoping to win cents here and there, and that's their edge.

You make good points. Market makers were providing plenty of liquidity before. However, I challenge you to find data suggesting that the bid-ask spread has, in general, gotten worse because of computerized market making. We should not discourage firms from providing the fairest price for investors, i.e. provide the tightest spread. These firms are competing for investor business, and that's how spreads get tightened (if you're not at the inside, you don't get the trade).

The spreads for the most liquid symbols are now at their minimum tick sizes, one cent. You can't improve on that, it's not mathematically possible. The argument can be made though, that the minimum tick size should be less than one cent. On stocks like Citigroup, it would probably save investors a lot. That's more of a regulatory issue though.

I don't disagree that market makers / specialists always existed. In fact, I emphatically agree! HFT market making (note that I'm not including predatory HFT algos) is basically that -- computerized market making.

Also, to your last comment -- why wouldn't you want market makers to get as close to the market as possible? That's a good thing! It's like Staples upgrading their technology to provide you with paper for cheaper. It's a win-win scenario. Instead of paying $25 for MS because the spread was $20-$25, the fastest market making firm now provides it for $23.35-$23.36, so you get your fill at $23.36. I'd say that's good for investors, isn't it? In the end, that's really the goal of all market makers -- to make the tightest spreads possible.

Exactly. They worked fine because market makers have been around for hundreds of years. High frequency market makers are just that... computerized market making.

Think of it this way. Let's say you want to make a trade. Buy 100 shares of Google. What would you do? You would probably call up your broker and ask him to buy your shares. The next part is what investors don't see, which is why the concept of market making is usually missed. The basic idea is... who is going to trade against you? If you're buying 100 shares, that means someone needs to sell 100 shares. What if no one is willing to sell 100 shares? Then you have to wait. Maybe you can wait a few minutes. Maybe an hour. If the stock is illiquid, you might have to wait many hours. If the stock is rallying, no one might want to sell to you for a long time.

That's where market makers come in. They don't own any of the stock, but they're willing to take the other side of your trade. As I mentioned, market makers make a continuous market possible. They're standing there ready to take either side of a trade. That allows you to think of a trade you want, enter it, and get it executed at a good price within the second.

All businesses are willing to stay in business if they can make money. Doesn't mean they're not also providing a valuable service.

"I want to sell for at least $5." What you're saying here is basically, "Give me the best price you can sell for, and sell it there." You're completely welcome to put in a limit order to sell at $5.00 or $5.01, depending on your internal algorithm for what's a "good price". You think your broker is giving you shitty fills? Well, then place the limit order yourself! No one is forcing you to trade at $5.00! Note the ambiguity of "best price". A price good now may be a terrible price tomorrow. Analogy: Let's say you're trying to sell a house at $500,000, but you can't find anyone right now to sell to. But then, a house flipper shows up and takes your offer. Later, you find that the flipper sold it for $550,000. Do you blame the flipper? You could have waited longer and sold it yourself for $550,000. You chose not to because you wanted to sell it immediately. On the other hand, if the value went down to $450,000, you'd probably be very happy, wouldn't you? The point is, you paid the price for selling it now. If you want a "better price", you can always sell later, but you have to be willing to take the risk.

I don't disagree. The auction-based system, with perhaps some complicated rule-set, may eliminate latency wars to a large extent. However, investors give up something in return. If the auction happens every hour, then investors cannot trade until those hours.

The rhetoric present here reminds me of news articles I've read. It's really a shame how much the media has spoonfed ill-conceived ideas into everyone's heads. Ignorance is bliss, I suppose. I'm going to (as I already mentioned) rule out HFT firms that are not market makers and specifically look for unfair advantages. That said, specialists did the role of high frequency market making back in the day. They posted bids and offers for anyone to trade against all day long. Now if you were a specialist, how would you provide continuous price discovery? Valuations of the company might help, but really, if I calculate the company's valuation two seconds from now, will it change from what it was two seconds ago? My point is that using fundamental analysis fails when you're a specialist worried about making a market all day long. You care about bids and offers down to the cent because the only money you're making is if someone sells to you and hopefully someone else comes in and buys from you. The role of the specialist is to balance supply and demand. If there is a lot of demand, the specialist will move his market up. If there is a lot of supply, the specialist will move his market down. It's as simple as that. The faster you move, the tighter bid-ask you can provide.

I don't really understand where you're coming from. It clearly does not seem like you are in the industry, because your comments sound prejudiced without basis. In any case, you should think about this: what percentage of the investing public actually knows about market microstructure? I'm not talking bonds or stocks. I'm talking about the inside bid, inside ask, spread, etc. What percentage of the news articles out there say "HFT's are stealing your money" and actually provide evidence? From what I've seen, the best argument seems that "we don't know what they're doing and they're very secretive, so they must be doing something illegal."
So really, what evidence do you have that market makers don't help investors? As I mentioned, I'm sure that there are firms out there who abuse the market and make money off of unfair advantages such as flash trades. Unfortunately, all of these firms have been bundled up into "HFT" and the media has had a field day beating it up. Investors love to find someone to blame for the stock market going down. Whoever makes money, must be taking mine!
We're doing what old specialists used to do, except fairer. Specialists were pretty much monopolists in certain stocks. Without competition, the spreads naturally were huge.
For your information, we actually do talk about how much of the daily volume we're doing and what fraction of liquidity we provide. We're very proud of the fact that we are almost always on the inside bid and ask (the spread for which is often the minimum tick size, 1 cent) for many of the products we trade.

I've been an avid follower of /. for some time now. I've gained a lot of insight from reader responses, which are generally well thought-out, mature, and reasonable. On the topic of market microstructure, however, I feel /. falls woefully short. I cringe when I read comments that sound like something from Zero Hedge.

I work in HFT. I make markets. Obviously, there is an incentive for me to talk about all the good things HFT brings to the world. However, I also believe that we serve a function in the market. Perhaps not vital, but still a service nonetheless.

What do market makers add to the market? They're willing to stand on the other side of your trade. They serve a vital function to the market and we can trace them back to the specialist days on the floor. Let's all agree to start from there.

What do HFTs add to the market? Now this is where you have a large divide in opinion, and rightly so. Some HFT firms will engage in predatory behavior that is unfortunate, including quote stuffing and price manipulation. I am not writing to absolve all the bad things that many HFT firms do. However, in my view, ideally, HFT market makers add these factors: immediacy and continuity.

As an investor, you can go up to a trading terminal at any time in the day and someone (most likely an HFT firm) will be there to take the other side. That is immediacy. You also have access to price discovery that is happening every fraction of a second. That is continuity. These are ideal situations, and not every HFT adds these values. Firms that only remove liquidity are often not providing immediacy. Firms that manipulate prices are usually not providing continuity.

If you think, "HFT's will run at the sign of chaos!" I agree with you. The better, smarter, and faster firms will continue to stay in the market, but only up to a certain point. Why should anyone stand in the way when a big institution sells 75,000 ES contracts? We trade and provide liquidity so long as it's profitable. If you have a problem with that, you have a problem with capitalism. How do you possibly incentivize participants to absorb tail-end risk?

If you think, "But investors don't care about 30 microseconds!" I agree with you. The short reaction times are there so that we can manage risk. It indirectly adds value to the investor because it allows us to manage risk better, which allows us to provide really tight markets. Think about it. We're standing there for anyone in the world to trade against all the time. Adverse selection is the name of the game. Back in the specialist days, spreads were sometimes in dollars. Now they are in pennies, and in many liquid stocks, exactly a penny. I assure you -- if we ever move to a system that taxes each trade or throttles latencies, you will see spreads widen out immensely because it's harder to manage risk. If you impose a limit on the minimum life of a quote, you will see spreads widen because there's risk in standing in the middle of the highway for too long.

If you think, "But company values don't change every 30 microseconds!" I agree with you again. It's the possibility that they could change that necessitates high reaction speeds. Company valuations are stable -- on average. But once in a while, some information is leaked that damages the company's reputation or some big institution decides to buy a ton, and you're left with a huge position that's going against you. Should we stand there and absorb that flow even when it's not profitable?

The last point is probably the biggest factor in a gating system where trades only take effect every N seconds. You can only update your position every N seconds, so as a market maker, you're essentially putting out a lot more risk. Some firms will be smart about risk management and be able to provide tighter spreads and make money for themselves. Some firms will not and they will go out of business. Almost certainly, firms would widen spreads to protect themselves from adverse selection.

The gating system also will not eliminate the arms race. Let's suppose we have an auction system that matches buyers and sellers every 2 minutes. Who gets matched first? If the system is based on price-time priority, then there is an incentive to be at the front of the line. Not only do you want to be at the front, but you also want to be on the right line. That sets up demand for faster computers and faster connections. If the system is pro-rata, which matches large orders first, it raises the barrier to entry for smaller companies with less capital. There was a comment here about randomized matching. It would still not completely remove the arms race. You still want to be able to cancel orders in case bad things happen and whoever cancels as close to the deadline can process the most information.

Maybe there is a reasonable solution, but please do not simply throw out ideas like "let's throttle speeds" and "let's tax traders" and think that's all there is to it. Market makers would love to provide liquidity and get paid for it. We don't widen spreads or decrease liquidity out of spite -- we get out if we're losing lots of money.