The High Frequency Trading Scam (2009) | Hacker News Reader

The High Frequency Trading Scam (2009)

seekingalpha.com

48 pointsaliston37 comments

One thing this commentator (and others) gets wrong is the emphasis on the place of flash orders in this process. What they seem to be missing is that flash isn't automatic, it is selected by the sender. In this example, the implication is that the buyers of broadcom were disadvantaged by flash orders, but if flash orders were involved at all, then they would have been selected by those very same buyers, who would presumably not have disadvantaged themselves. Flash is an attempt by the more technically sophisticated markets (DirectEdge and BATS) to provide a service to customers who were tired of slow execution on some exchanges causing their orders to be faded. Any time that I see someone complaining about this, I suspect that they either don't understand it, or that they are being paid by someone at the slower markets (e.g. NYSE) to spread FUD.

What is also missed is that most exchanges discontinued flash orders late in 2009. BATS and NASDAQ both discontinued the practice, Direct Edge still allows them (they are opt-in) and NYSE/ARCA never had them.

Wow this essay is totally off track.

There is no such thing as a "market price". To prove this to yourself, write a simple model of the stock market with buyers and sellers. Every order is a limit order.

What is the "market price"? Is it the last trade? Is it a weighted average of the last 10 trades? It doesn't actually exist.

At any price there will be a certain amount of liquidity. If you want to guarantee that you can buy a lot, then you have to pay extra b/c the counter party doesn't want to sell low so he hedges by adding a bit of spread.

The same applies to selling. If you want to sell a lot you have to discount a bit b/c the buyer will want a bit of insurance that he/she isn't overpaying.

The HFT algos do add liquidity b/c if the liquidity were there at the first price probed, then the algo would never really matter, the desired shares would have changed hands with some other counter party.

Clearly, in his example, there are people out there willing to sell at $26.10. In the old days, you could put out an order from 100,000 shares, saying you would pay up to $26.40, and you would get most of these shares at the $26.10. If there were enough people willing to sell that many shares at that price.

Now, even though there may be people out there willing to sell you the shares at 100,000, these programs can find out your top price, and they will but the shares at $26.10, and sell them to you at $26.39, thus making a profit of the difference. Providing no extra services to you, or the market. They have just stepped in as the middleman with no risk.

This is how the article describes it, and this is how I understood it. Is this wrong? I think you are just arguing semantics about the term market price.

Are the shares worth 26.40? Then you should be happy to get them for 26.39. If they aren't worth it, then don't make that offer.

I don't see why you should care who sells you the shares. You either want them or you don't.

The ethical issue is that someone is deriving additional profit through an asymmetrical advantage. It's kind of an interesting ethical question actually.

Personally it just feels wrong in some hard to define way to algorithmically determine the other sides maximum/minimum price through some process they don't have access to. I can certainly understand why you would do it but it just doesn't feel quite right.

I don't think I have an ethical problem with using small orders to sound out the other party. Where it tips over into being unethical for me is if those exploratory orders are cancelled rather than fulfilled. I'm surprised its even possible to cancel your offer once someone else has accepted it.

Well, keep in mind that this is just an intermediary before the trades hit the rest of the market. If the HFT had not been allowed the advanced access, then the transaction would not have been filled in the local exchange and would (or would not) have been filled by the larger market.

If another participant besides the HFT algo on the local exchange had been willing to take the other side of the initial price offered, then the HFT firm would not benefit from "probing" for the price.

And, what's to stop the entity buying a large quantity of shares from strategizing to account for HFT strategies?

The core reason that exchanges offer an advanced look at trades is to try to pull liquidity out of parties that create no transaction fee overhead before passing it to the rest of the market. The price will never be worse than what would have occurred if the trade had been allowed to go out to the rest of the market.

The author of the article is essentially just complaining that his own crude algorithm has been outsmarted.

It seems somewhat like people inflating bids on eBay by incrementing their max bid by odd amounts until they hit the other bidder's max.

Also, it is of note that such behavior is forbidden on eBay and many other marketplaces.

Not on stock exchange though.

There is no way to send an 'exploratory' order as you describe and is is, as you suspect, impossible to cancel an order that has already been executed.

The only way the example in the article would be if there were no other available sellers at a price below 26.41 (in which case I'm curious why an algo would bother sounding starting at the low end instead of the high end. The only ways to get such a wide market would be for all the other sellers to have disappeared (in which case the buyer is not disadvantaged by the HF algo because he is getting his order done rather than not) or for the HT algo to have made a guess that other people would be willing to pay a higher price in the future for BRCM and bought all of the shares available up to and above 26.40 which is the definition of a short-term trading strategy and, I should also add, full of risk that the price that you just pushed .30 will fall back down without you having a chance to get out hence making it a losing trade.

You are missing the point. A few years ago, I wouldn't have had to pay so much, now I do. How is that good, that I now have to pay more for each transaction, while these companies make the difference?

It means I will buy less shares of other companies. I will have less money, for the same goods.

I think this is a bit off-the-mark. The point is not really a moral one from the point of view of the person getting the worse deal. Sounds illogical, but it's not: as an earlier poster said, the limit went in at 26.40, it doesn't matter to you why the market is more efficient (or whatever you want to call it), you were happy to buy at 26.39. From your point of view for that individual trade there is no case that you were scammed.

The problem is the larger perspective of the market structure being weighted in favour of certain participants. A particular class of investor is given a specific advantage by the supposedly neutral exchange, and they use it so systematically exploit an edge that doesn't exist for you or anyone else. That is wrong.

As anyone involved in bidding war with limited resources would tell you, "happy" is not a fitting word here. There is a reason why shill bidding (old-school name for what HFT does) is often prohibited or illegal.

I don't think this can possibly be right, though, unless I'm not understanding the claim, exactly - any limit orders in the queue at any price between $26.10 and $26.38 would be matched against the $26.40 limit order before an order at $26.40 or $26.45 would be, so how the hell could the HFT algo even probe your top price if the market is "full" in-between?

I think it more applies to situations where other liquidity providers would step in offering more shares somewhere near $26.10 if they had a chance to react, but the HFT algo beats them to the punch, buying up every share under $26.39 and then selling it to the big buyer at that price before anyone else can add liquidity.

Again, though, I have no idea how they could possibly figure out that demand cuts off at $26.40 with a full book under that price...I suspect there's something off with the explanation this article offered, it doesn't quite make sense to me except in highly illiquid markets.

Clearly, in his example, there are people out there willing to sell at $26.10

I am not sure where you are getting this. I don't believe it to be true based on my understanding of the mechanisms involved.

Doesn't sound like a scam to me. It's as if you own an apple stand, and customers periodically come up to you and say they would like to buy some apples. You offer to sell 1 apple at 20 cents, and they agree. Then you offer another apple at 25 cents, and they agree. Eventually you get to 40 cents, and they disagree. At that point, you offer all the apples they could possibly want at 39 cents, and you make a lot of money.

Your advantage wasn't speed. Your advantage is that the other side was a bad negotiator.

At the very least, they could have placed an all-or-nothing order. Also, this analysis assumed the seller had no competition.

It's more like an apple vendor was offering an apple for 20 cents and someone stepped in between you and the vendor and offered it for 21 cents. You have no idea that the person offering the apple for 21 cents is not the real vendor, he's the only person you can buy from. That's what the millisecond HFT systems are doing.

The prescient part of the 2009 article: "If you're wondering how Goldman Sachs and other "big banks and hedge funds" made all their money this last quarter, now you know."

Actually, if you have a limit to buy at 25 cents and a seller still exists at 20 cents, the market maker will fill your order at the lower price first.

But, and someone correct me if I'm wrong, not if the HFT apple trader buys from the apple seller at the lower price and then resells it.

If course the HFT trader wouldn't sell unless it could profit by selling. They are HFT traders, not market makers.

Obviously if you are buying shares sold by an HFT trader, you are not buying them from a market maker, so you are getting a better price than the market maker is willing to offer.

HFT is a technique that's used by many people, including exchanges, hedge funds and market makers. You're right though - the exchange is obliged to fill the consumer at the best available price.

I meant market makers in the role of traditional market maker (which could be HFT but is not HFT as a strategy, merely as a speed measurement).

Thanks for correction. On HFT - in what sense is it ever anything other than a speed measurement?

Well, lately people are using HFT to describe specific sorts of strategies... mainly strategies that rely on per-trade kickbacks, etc., not just high speed.

If the HFT trader offers a lower price, and you offer a higher price, the exchange will fill your order first.

This is ancient, wrong, and out of date.

Ah for the days when Karl was mcs.net. A finer ISP there was not to be had. I wish he were still doing that rather than whatever it is that he is doing these days, with boldface and intensity.

Seriously, don't poke the yamhead. You'll be sorry.

I read this article when it was first published, and it was what first sparked my interest in the exchange side of the equities world. There's a lot of junk information on the internet on these topics, and as a result they're difficult to read up on. A common error I've seen is articles that imply that flash orders and high-frequency-trading are the same thing, which is incorrect.

    "Automatic programs began issuing and canceling tiny
    orders within milliseconds to determine how much the
    slower traders were willing to pay

I originally understood this to mean that parties were 'painting the tape' by causing trades to appear with prices different to what the actual market price is. But that's not it.

In the scenario described, it's not trades that are being issued and cancelled, but orders. As in, an opportunity to trade if the other party matches.

I don't think it would be possible to coordinate a situation of sniffing out the best price as described here. Markets have queues, and if you want your order to be filled then you put it in the queue where it sits until you cancel it or it gets filled. It can only be filled when it's at the front of the queue. So participants in the market compete for queue spots at each price point from the moment the market opens.

That would be exchange dependent, but that just emphasises the point. If one exchange did this and another didn't then the exchange that didn't would get more business because their prices would be better.

    A couple of years ago if you entered a limit order for
    $26.40 with the market at $26.10 odds are excellent that
    most of your order would have filled down near where the
    market was when you entered the order - $26.10. Today,
    odds are excellent that most of your order will fill at
    $26.39, and the HFT firms will claim this is an
    "efficient market."

Citation required? I don't see how this can be true. If the market is at 26.10, then there will be orders in the queue at price points around that. If the market is scared, then the spread will be wide, but that's to be expected and accounted for by the fear, not by conspiracy.

Despite all these problems, the criticism of the flash order culture looks sound. A flash data stream could give a party the ability to cancel stuff they have in the queue earlier than they otherwise would, and to get to the queue faster on new information. It seems wrong that there should be a certain type of information to be available only to certain parties. I think the major exchanges have stopped doing flash orders now though. There was a lot of press about the SEC reviewing it after this article last year.

I don't know if it's the case but if you paid a higher rate of commission/fees on flash orders then it might be the case that the general market is getting slightly worse prices but at the advantage of lower fees. Which could make it beneficial for general market participants. As I say though, I don't know if this is the case.

Flash orders have lower commissions, and always give the best available price[1] to the buyer.

The mechanics: best asking price is 10.0 on INET, 9.90 on ARCA. You place a buy order on INET. INET is obligated to route your order to ARCA (where the cheapest price is available) rather than filling it at 10.0 locally, and you get charged 9.90 + routing fee [2].

If INET flashes your order, it gives an HFT trader on INET the opportunity to fill your order at 9.90. In that case, you pay 9.90 for the trade and avoid the routing fee.

[1] A better price may be available on a darkpool, but the exchange would not have routed your order to the darkpool since the exchange did not know about it.

[2] There are flags you can set so that your order is canceled rather than routed.

a very one sided article there and fails to mention the algos used by your broker to minimize information leakage of your trades... if their large enough. what gets me is the amount of money HFT market makes make off your trade is tiny compared to the fist full of cash your broker is taking from you..

I'm not sure, but I think the logic for the latter is that by sending a straight limit order, you would consistently be moving the bid price up by continuing to fill the highest priced bid or filling the highest priced ask, up to the limit.

Of course, this would be true even if there were no algo trading... so... I dunno!

Oh, it's also worth noting that this is not an issue for the individual investor because most individuals buy stocks in very small blocks, usually less than the minimum increment size that shares trade at on the open exchanges. Larger, institutional investors use algorithms from their trading partners.

I guess if you're a millionaire personal investor you're SOL though. But usually people with that much money have a Wealth Management firm to take care of their trading for them.

> most individuals buy stocks in very small blocks, > usually less than the minimum increment size that shares > trade at on the open exchanges

There is no minimum increment size. You can trade a single share at NASDAQ, BATS, ARCA, wherever. 100 shares orders are very common, and that is lower than the size of most retail orders. 100 shares at $20 is only $2k.

I just thought of an analogy:

If what happens in the article is correct, it is very similar to what auto-bidding shill bots do on auction sites, eBay, they obtain information on the highest price you were willing to pay for an item.

Is this the price we(investors) are willing to pay for 0.03sec liquidity? I am not sure it is worth it.

Interesting, given the date (July 2009). Is there a presumed connection with last week's 1000 point drop? A mis-order to sell at any price meets the HFT scam that rapidly, though temporarily, drops to the extreme price (in this case, $0.00 or $0.01).

The connection is not so simple. People who do HFT are painfully aware that orders get broken, and so all exited the market before the drop really began going. The 0.01 orders are always hanging around.

However the fact that people have come to depend on HFT to help make markets means that when they disappear unexpectedly the market is smaller than it would have been, which makes large price swings much more likely.

People need to adjust to the presence of HFT. For many traders it breaks the model they may have used successfully for some time and they just don't want to go out and update their model.

As typically happens, they go crying to the government to stop the evil villains from forcing them to change the way they think about a changing world.

Reply on news.ycombinator.com