Traders Profit With Computers Set at High Speed | Hacker News Reader

Traders Profit With Computers Set at High Speed

nytimes.com

53 pointsmjfern22 comments

People want to know they have a legitimate shot at getting a fair deal.

Ironically, it's articles like this that give them doubt. They could've just as easily written an article praising high-frequency trading for getting most participants a fairer deal.

High-frequency trading has driven down costs. They're the main reason most people will only pay a 1c bid/ask spread for those Broadcom shares. Before computerized trading, they would've paid over 12x that. Additionally, as they push the systems to scale and force competition between markets, the per order and per share fees are driven down.

They've also made the market more efficient, so most people don't get screwed by some slime taking them for a quarter to a dollar per share doing arbitrage against another market, mismarking the trade, or maybe the worst, an order his friend is holding in their pocket. Those things are now typically limited to less than a cent. High-frequency traders are usually overjoyed to average 1/4 of a cent profit. You'd be amazed at how many losing trades they do. For most orders, most of the screwing happens between the trader and the broker, not the trader and other traders.

It's important to remember that there's two parties involved in every trade. One could easily say the Broadcom buyers described were upset they couldn't screw the sellers as badly anymore.

If you look at the details of the Broadcom trading given in the article, it's likely even more simple than what they conjectured. 56,000 shares of Broadcom really isn't that much. It's only 1/2% of the number of Broadcom shares traded daily.

The issue the buyers probably had was that MOST bids and offers in such stocks (>80%) are from computers running high-frequency trading strategies. If the computers just stop offering to sell more shares (the simplest reason being they hit their margin limit), you could get the same exact situation described.. Except now, there's nothing sinister at all going on. You just get to see how much they're normally helping you, and you'd write an article about how the SEC should increase the margins for high-frequency traders (which would be equally silly.. and meaningless since the large firms ignore the rules on that point anyway).

Disclaimer: I put bread on the table doing this stuff.

I don't understand how high frequency trading could be considered beneficial. In the article's example, HFT captures some profit from an unknown trend simply by being faster than the traditional participants. This directly benefits those who can get their software closest to the servers running the markets, to which the average person is limited. Insiders like Goldman Sachs can run their software directly on the routers connecting the markets, and thus benefit from insider access to a natural monopoly on speed.

To be clear, this has nothing to do with algorithmic trading. Using computers to detect trends or to plan strategic trades is natural and healthy for the market. I am also not suggesting that computerized trading and reduced fees are bad - they have made the market far more efficient.

> One could easily say the Broadcom buyers described were upset they couldn't screw the sellers as badly anymore.

But in this case, the Broadcom sellers AND buyers were both screwed by a third party, who bought at a price that hurt sellers and sold at a price that hurt buyers, on the order of milliseconds. This third party runs the same scheme all day, every day, taking money out of everyone's pocket without doing any real work. This is why people would distrust the market.

It's the same benefit as normal traders provide: by constantly trading, HFT engines are providing liquidity and cutting the bid/ask spread.

The "unfairness" here is limited to other traders who'd be trying to profit by selling the same liquidity, and now can't compete. But your occasional market order benefits.

How does this HFT practice provide additional liquidity? The only time they purchase shares is right before someone else is going to. It is the ultimate man in the middle, taking a little bit of every single transaction in an exchange. It raises the cost of doing business for everyone (except the HFT trader), and is a net inefficiency in the system, like energy lost to heat in a poor electrical conductor.

I know a little bit about trading, a decent bit about trading platforms, and not a lot about HFT in particular, but it was not my impression that HFT algos can literally MITM the market. They buy when they detect changes in demand.

I'm not sure MITM is accurate, but the article's graphic seems to imply that the machines have access to the trades before they are executed.

http://www.nytimes.com/imagepages/2009/07/24/business/0724-w...

I don't consider it an inefficiency as much as a parasitic loss. All those man-in-the-middle fragments are being scooped up and funneled to one entity.

You seem to be taking the article's statements as facts and treating the story as a representative subsample. I don't think either is the case.

Most high-frequency trading provides liquidity to a market, not takes it, especially in US stocks. That way, you not only possibly get some edge from the bid/ask spread, you also may collect "liquidity rebates" (see http://www.batstrading.com/FeeSchedule for one fee schedule for an exchange that's actually founded by some high-frequency traders). There's tons of high-frequency trading that's actually gross negative (they sell for less on average than they buy), but net positive (on average, they get paid more fees than they lose gross).

If I understand you, you are basically saying that the main effect of HFT is that the people who bought Broadcom paid more to the people who sold Broadcom, with a small percentage going to the HFT middlemen? Thus, HFT decreases the advantage of people who have more knowledge of the market (the buyers in this case) over those with less knowledge (sellers). Correct?

> They've also made the market more efficient, so most people don't get screwed by some slime taking them for a quarter to a dollar per share doing arbitrage against another market, mismarking the trade, or maybe the worst, an order his friend is holding in their pocket.

If someone is a "slime" to do arbitrage against another market, how come it is suddenly okay when HFT does it?

Yeah, I mean, if anything, I much prefer those acting on knowledge to those acting simply on just observing moves of others. Decreasing the incentives of the smart money doesn't seem to do anyone any favors.

If I understand you

Sorry, probably my error. The main takeaway should've been that the article is skewed, inaccurately presenting traditional traders just as victims, and algorithmic traders as entirely bad.

how come it is suddenly okay when HFT does it?

I didn't say that. Show me a clean portion of the financial world, and I'll be shocked.

On a side note, any tips for the hoards of eager younguns dying to get into the biz?

Invent a body swapping machine. I know plenty dieing to get out.

If you must, stake out a table at Starbucks for a month, and read every post on elitetrader.com, nuclearphynance.com, and seekingalpha.com. That'll at least teach you enough to know what the steps you should take are. I just don't recommend many of them, except, for the right individual, breast implants. In some cases, they can be a great investment.

I left a startup for a regular hedgefund first; I think having a lot of outside interest in finance helped. Then the HFT firm hired me because now I had some finance experience, and I knew C++. Another approach would be to look at the career pages of getco, citadel,etc, and diff your resume against what they want.

Edit: but really, why are you dying to get in? I would guess startups are a better bet for most readers of HN, anyway.

The "slow" traders had off-market places to find liquidity for a long time. Posit does a block crossing every half hour, with reports back to the exchenge.

If fairness is such a concern, they should just do a once-a-day crossing. That'd be much more fair, but much less liquid.

The part of this article that concerned me is the rapid bid/cancel scheme. Traditionally, the cost of finding out about price elasticity in the market is that you have to sell/buy some product and see how the market responds.

This kind of rapid bid/cancel would be the equivalent of placing a bid on Ebay to see if your competitor's max auto-bid will top it, then retracting your bid. Do this on a regular basis, and you can figure out exactly how much you'd have to bid to beat out your opponents, without putting a cent at risk.

A policy that any order must remain in force for at least one full second (long enough for standard electronic trading services to act on it) would probably limit this behavior.

While markets are supposed to ensure transparency by showing orders to everyone simultaneously, a loophole in regulations allows marketplaces like Nasdaq to show traders some orders ahead of everyone else in exchange for a fee.

That's just wrong, plain and simple, as is the colocation of certain privileged servers next to (and I mean literally right next to, in the same rooms, even) exchange computers. The exchanges should not be able to profit by offering certain customers higher quality access to data that is supposed to be equally available to all, it seriously raises doubts about the fairness of the system as a whole.

FWIW, the blog ZeroHedge (www.zerohedge.com) has been following this topic in much more detail for months.

The fact that some banks like GS are allowed to have their servers and other equipment co-located at the exchange while most others are prohibited is suspicious albeit not damning.

I think the bigger issue here are the dark pools that continue to grow largely due to HFT on the exchanges. When 75%+ of the volume is coming from HFT, a lot of transaction end up occurring off the map reducing transparency and brining along its consequences.

I wonder what the effect of making quick bid/cancel or increasing the cost of canceling a bid would do to the market. My suspicion is that it would make HFT less useful, but, as all things in game theory, results can be surprising.

Not that this is practical, but I wonder what the effect would be of adding a random (milliseond scale) delay to the transmission of orders. Maybe that would force a longer-term outlook (seconds instead of milliseconds).

I suppose the effect would just be for firms to find ways around the delay.

Wasn't there a proposal floating around to make all orders valid for one second, effectively killing these types of trades? I'm all for that. No individual investor can cancel a trade that quickly, how come it's okay for GS and all the other traders to do it?

Do we need stock exchanges exclusive for Value Investors?

Reply on news.ycombinator.com