News for hackers. Stuff that matters.
Re:Michael Lewis: shilling for the buyside? by tptacek on Friday April 04, 2014 @11:54AM (#7532216)
Two other concepts you really want to understand before forming an opinion about HFT: "large block trades" and "bid-ask spread".

Lewis has a drive-by mention of bid-ask spreads in the book (he concedes that they've been decimated, while asserting without evidence that HFT has created a new "shadow spread"), and, thus far, no mention whatsoever of the mechanics of large-block trades. In Lewis' universe, a hedge fund can unload 100,000 shares of XYZ just as easily as you can sell 10 shares of AAPL.

Re:Michael Lewis: shilling for the buyside? by 001sky on Friday April 04, 2014 @12:56PM (#7532704)
tptacek -- Not to nitpick but isn't this comment is like someone accusing you of having a "drive by" understanding of computer security? He knows what a block trade is, and how to quote a market. Lewis worked on a bond trading desk, and coined the phrase "equities in dallas". So it's not CFA 1 here for him.

To your point: what's missing is the hedging costs. The reason block trades trade out of the market is not just that they are "bespoke", but that there is the inventory hedge risk. It may take some time to build up a position (during which prices can move adversely). And this may or may not need a hedge to protect the holding period (which someone needs to pay for).

Various types of trading strategies have very different risk profiles when compared to required inventory hold/periods etc. And this is a hidden source of economics or cost. Hedging doesn't always scale well, either.

So its very hard (especially as a lay-person) to make inferences about many things unless you are taking into account both profits and risk. As I'm sure you know.

Re:Michael Lewis: shilling for the buyside? by tptacek on Friday April 04, 2014 @01:04PM (#7532754)
I've reread Liar's Poker at least 5 times (it's what I read when I don't want to think about what I'm reading next, because you can pick it up, turn to a random page, read in either direction, and enjoy yourself for a few minutes).

Read reviews of the book from traders and you will see lots of people scratching their heads the same way I am.

It is confusing that a former bond salesman, who wrote several famous vignettes about how Salomon got its clients into jams by selling them blocks of bonds and having the market move out from under them, would believe the things he appears to believe about RBC's role in the market.

In Lewis' universe, a $2MM/yr sell-side broker accepts a 100,000 share block order from a hedge fund, consults a single blotter screen, enters a 100,000 share order, and is startled and outraged when that order isn't as easy to place as an Amazon one-click order. Egads! The broker has to actually work for his money! Kill the HFTs!

I don't understand what inventory hedging has to do with the point I'm making. Hedge trades are trades. Large hedge trades involve block shopping.

Re:Michael Lewis: shilling for the buyside? by kylebrown on Friday April 04, 2014 @02:24PM (#7533318)
Even Manoj Narang conceded that he noticed the same thing as Brad Katsuyama around the same time: lower fill rates. The lower fill rates only led to scratching of heads. It was the underlying cause of lower fill rates which led outrage (in Brad's case).

The underlying cause is the "perverse incentive" (Lewis's words) of the taker-rebate (reverse of the usual maker-rebate taker-fee). Is there a sensible reason to such a fee structure? Or is it just useful for HFT firms to bait orders and get an informational edge they can exploit against mom & pop's retirement fund?

The same should be asked about esoteric order types (Hide-n-Slide, Post-only, etc). Its cleary a complex landscape with middle ground, but it would be surprising if there weren't cabals and conspiracies hiding in obscure corners.

Re:Michael Lewis: shilling for the buyside? by 001sky on Friday April 04, 2014 @01:23PM (#7532892)
Large hedge trades involve block shopping

ELI5 version is that you need to pay for (1) working capital. This scales with the size of the block (capital cost). You also need to pay for (2) some lead-gen fees to get the block (service cost). You also have to pay for (3) a hedge on economics while your lead-gen process play's out. The cost of this hedge scales with size (just like capital cost), but also with vol (of underlying) and time (speed of lead gen process).

If you don't cut the deal to cover your hedge (or put one on) you run the risk of massive losses from mino movemnets in the market (ie, if you have a <1% cent spread on 5 mill position and the stock spikes 10%).

So a trade's profits:

(A) scale+fees = Revenues

(B) capital+hedge = Costs

Both the hedge and the capital are proportionate to time.

TLDR when you talk about HFT you are optimizing on time; and thus talking about (B) as much as (A) in terms of profits. Most of the optimization is on (B).

What's unique here is that (B) is blind to the casual observer and to some extent, the market.

But analytically, it influences profits and drives behaviour/incentive structure.

If you dis-regard it out of hand, you're missing an important part of the picture.

Re:Michael Lewis: shilling for the buyside? by foobarqux on Friday April 04, 2014 @02:21PM (#7533298)
I think you're simplifying too much. The example being repeated in relation to the book and IEX is you have a 100,000 order, 25,000 displayed at the same price, say $10, on each of 4 exchanges.

You might think you can fill 100,000 @ $10 and allegedly many people did. In fact if the exchanges were consolidated into one you would be filled.

It seems to me a displayed order should be able to be traded and not behave in a seemingly prescient way by disappearing when the order is made.

In an ideal world would the latency between exchanges serve any beneficial purpose?

Re:Michael Lewis: shilling for the buyside? by tptacek on Friday April 04, 2014 @02:24PM (#7533322)
Think about CAP theory, and the markets as a distributed system. How is the ideal you're looking for achievable?
Re:Michael Lewis: shilling for the buyside? by foobarqux on Friday April 04, 2014 @02:31PM (#7533364)
It's not achievable (although maybe periodic batch auctions would help), it's a thought experiment as to why this particular scenario isn't quite the same as reacting to a large order in principal. Specifically you cannot decide to remove orders during the initial transaction of a big buyer or seller.
Re:Michael Lewis: shilling for the buyside? by apaprocki on Friday April 04, 2014 @12:05PM (#7532309)
And just in case it isn't entirely clear, one example of 'algo' trading (algo != HFT) is developing an algorithm to sell smaller blocks of shares over a period of time to "hide" the fact that one participant is unloading 100,000 shares so the the price is not driven down too hard, thereby reducing the overall return from selling the shares.
Re:Michael Lewis: shilling for the buyside? by kasey_junk on Friday April 04, 2014 @12:53PM (#7532678)
Just to nitpick a touch. One of the real problems with discussions about HFT is that there is no generally accepted definition of what the term means. In general it is some mix of "computers make trading decisions without human intervention", a high number of trades throughout the trading day, & low latency requirements.

The reality is that electronic trading systems have different values for all of those variables. I've worked on a trading system that only traded 4 times a week, had no human operator, but extremely tight latency requirements.

I've also worked on a system that did very high volume of trades every day, that was watched like a hawk by a team of human traders, and didn't much care about latency.

Re:Michael Lewis: shilling for the buyside? by wglb on Friday April 04, 2014 @01:50PM (#7533079)
One definition I have heard is "multiple trades per day", but likely the common definition, for the liquidity-providing definition, is "multiple trades per millisecond".
Re:Michael Lewis: shilling for the buyside? by apaprocki on Friday April 04, 2014 @02:17PM (#7533278)
HFT as you read it in most stories/comments is trying to imply is what Mark Cuban is hinting at in his blog post that also made the front page. Being fast or having low-latency does not make you a bad person. Using those advantages to front-run books or other "tricks" are what people imagine when they think of HFT with a negative connotation.

A firm simply unloading thousands of shares in 100 block increments in a number of milliseconds with very low latency technically would be a bunch of "high frequency" trades, but that type of activity is not what is being debated.

Re:Michael Lewis: shilling for the buyside? by wglb on Friday April 04, 2014 @03:01PM (#7533556)
HFT does not front-run trades. This is an apparently common misconception.

Mark Cuban misunderstands a lot of things, including Dennis Rodman and High Frequency Trading.

Re:Michael Lewis: shilling for the buyside? by tptacek on Friday April 04, 2014 @03:04PM (#7533572)
Since you two are both personal friends of mine can I just point out that (a) you both know a bunch of stuff about markets and (b) Andrew works in R&D at Bloomberg. :)
Re:Michael Lewis: shilling for the buyside? by apaprocki on Friday April 04, 2014 @04:06PM (#7533977)
I should have said "allegedly" :) I wasn't making a statement that they do/don't, just that those actions are what are considered the evil behavior.
Re:Michael Lewis: shilling for the buyside? by wglb on Friday April 04, 2014 @04:34PM (#7534158)
Ah, got it. Sorry if I was overly blunt.