Comment by tptacek | Hacker News Reader

Comment by tptacek | Hacker News Reader

Parent Full thread

One of the more disingenuous things about Lewis' book (caveat: I'm only halfway through, because I'm listening to the audiobook and, on the advice of another reviewer, switched to the book _Dark Pools_ in midstream) is the rhetorical sleight of hand he plays with the term "front-running".

In Lewis' universe, the price movement that happens when an informed trader tries to piecemeal-fill a 100,000 share order is "front-running".

But in Larry Harris' universe†, not only is price movement during block shopping not "front-running", it is the fundamental reason for the existence of sell-side firms; in other words, the challenge of stealthily shopping 100,000 shares of thinly-traded product XYZ is the reason that Katsuyama made $2MM/yr.

Lewis posits that the natural state of affairs is that there's an NBBO for XYZ, and that Katsuyama should expect his equities desk to collect fat fees simply by accepting a block order from a client, pushing a button, and unloading it at the current best bid, whereupon at the completion of this order the price will change to reflect the altered supply/demand condition of XYZ.

But that is crazy. The presence of a 100,000 order changes the supply/demand conditions for XYZ. For RBC to expect to sell 100,000 shares at the current bid, it must expect to get something for nothing: to take the price hit for the increased demand of XYZ out of the hide of whoever is holding it right now.

Before HFT mania, "front-running" meant something very specific: it was a problem of agency. A client/broker relationship existed, and the broker exploited it by trading in front of their client's order. The client had a right to expect the broker to act in their best interests. But the markets as a whole aren't a positive-sum system; in the absence of a contractual relationship, nobody has an obligation to act in the interests of anyone else. A dollar you make on the market is a dollar someone else didn't make.

† https://news.ycombinator.com/item?id=7526783

The specific example given in the book is that an existing ask sized at 100,000 is already sitting on the order book when the buy order goes in at the ask price. The ask then disappears before the order executes. This is very different from the market moving in response to an order that has no market fill.

The 100,000 shares are sitting on the order book at different exchanges. It's not a single block being traded with a single order. That's a critical distinction.

If it were a single order in the book, I'm not sure I see the advantage HFTs would have conceptually had. They're not sniffing individual orders and then outrunning them before the original order hits the exchange; they're exploiting the individual piecemeal orders that are stealthily trying to move the block.

They're hunting down the orders of informed traders who are trying to exploit their own knowledge of the true price of products, and correcting the price on the market before those informed traders can take their knowledge out of everyone else's hide.

Yes exactly. A lot of people don't seem to understand this.

The institutional orders get baited to BATS BYX with "taker rebates" (on BZX the fee structure is the normal maker-rebate taker-fee, but on BYX it is reversed). It hits BYX first because the broker's router is programmed to send it there first for that rebate, but its only for a minimal amount of shares. And that's a signal which the HFT firms use to pull or front-run equivalent offers on the various other exchanges.

The other big scam seems to be HFT firms buying a broker's order flow - for hundreds of millions of dollars. For example, Schwab sold its order flow to UBS for 8-year contract in 2005, for $285 million. It was massively underpriced, Schwab is said to have left a billion dollars on the table. And UBS, in turn, sold Schwab's order flow to Citadel for an undisclosed sum. Why is the right to execute a broker's order flow so valuable? Nobody seems to have a good answer, but the obvious reason is that its easy scalping.

I see a lot of heat generated in this discussion, but I don't see an answer to your second paragraph. That's the answer I want to see.

It is the easy scalping he mentioned.

A market maker can only stay in business if they sell as much as they buy and if the average sale price is above their average buy price. They lose money if they provide liquidity in the way of market moving forces.

Market makers try to avoid informed and "toxic" order flow and broker flow is guaranteed source of small uninformed orders that are most likely going to be crossing the spread.

The answer is retail flow has no information - it's the ideal flow for professional market makers to trade against

While it's true that retail flow has no information, it's also true that a broker will obtain better executions if it does not sell order flow to HFTs.

http://www.elitetrader.com/vb/showthread.php?t=245029

https://institutions.interactivebrokers.com/en/index.php?f=1...

I will refer you to page 96 of the book. The specific example is 100 shares of MSFT asking $30.00 and 10,000 shares asking $30.01. Because of Reg NMS, your broker is required to take down the 100 shares first ("NBBO"), allowing the HFTs to then start buying the $30.01 offer and front run the remainder of your bid. The problem with this requirement, according to Lewis, is the NBBO is priced off SIP rather than direct feeds. I don't disagree with the literal point that you are making, however, there is no need for this kind of intermediation. Why should natural traders pay intermediaries for this function? Why shouldn't they coordinate on an exchange where they don't face this tax?

RegNMS requires exchanges and trading centers to honor the Order Protection rule to route trades to the trading center with the best listed bid/ask price.

I don't think RegNMS requires a broker-dealer like RBC to route its order to any particular exchange.

Brokers already have a duty of best execution to their clients. They are allowed to use their judgement to send orders to any given exchange regardless of the NBBO. That's what enabled RBC to implement "Thor".

I'm not an expert but I believe you are incorrect. What you are suggesting is that the broker could perform a "trade-through", which was allowed under the old ITS rules. Under the Order Protection rule, trade-throughs of even 100 shares are explicitly prohibited. Reg NMS is very strict and that's why it is so controversial. My understanding of Thor is that it does not trade through, but it executes at all relevant exchanges near-simultaneously by delaying order routing to lower-latency exchanges. This prevents HFTs from sniffing your order at one exchange and beating you to the next.

I too raise my "not an expert" flag and await corrections from 'yummyfajitas or 'kasey_junk.

There are a number of possibilities depending on whether you send in a routable order (it's a flag you can set -- route or don't route, and there can be fees associated with routing). What a sophisticated trader would do is send two orders simultaneously taking out the small 100 lot at 30.01 and the 10,000 at 30. The exchange with the larger but worse price order will accept your order if you set the ISO flag, indicating you have sent orders to clear out better priced orders at any market centers displaying a better protected quote.

That's correct. As long as the exchange to which the order is routed is presenting NBBO at the time the order is executed, then the broker is compliant with Reg NMS.

Also note, if that exchange is not offering NBBO, it is bound by Reg NMS to route onward to the actual venue that has NBBO. The exchanges are also bound by Reg NMS, not just broker dealers.

Interesting side note: Here's a link to an Amazon review of _Dark Pools_. The review claims the name Brad Katsuyama and opens the review with "As an insider in the world of electronic trading..."

http://www.amazon.com/review/R2HRZNW5661QVR/ref=cm_cr_dp_tit...

I'm liking it a lot so far; it's written by an author who clearly thinks HFT is sinister, but who also seems to think the code behind it is really cool. (Maybe I'm biased, because that describes me pretty well too.)

I quite like Dark Pools even with the obvious bias. I will say that occasionally it gets technical aspects of things completely wrong, but that is pretty easy to do.

Also, as with any discussion about HFT look into the people who are making accusations against it. Most of the time they are ex execution traders made redundant by HFT or failed at HFT themselves.

Basically, I really enjoyed the history part of the book and found the expose part riddled with mistakes, obvious biases and vague terms that could be interpreted in many different ways.

Reply on news.ycombinator.com