Reaction to Michael Lewis's book and "60 Minutes" interview

[Rum Runner]

Yes, you may be correct in one sense. But "your" market is not a "functional" one in the sense that the National Best Bid and National Best Offer are illusory then. And if you were designing a market structure from the ground up, "your" market would not be the obvious design choice.

How do you know what my design choice is? You are assuming I want it the status quo. If you really want to know. And I know you do. Here is what I'd want:

1) No dark pools
2) no internalization
3) one central exchange
4) basic limit and market orders only
5) Fix the SIP

However, we all know that won't happen. So I'd settle for...

1) only lit venues
2) no internalization
3) no sub pennies
4) basic order types
6) allow locked/crossed markets
7) fix the SIP

 

[StarDust9182]

At the 2 minute mark of the CNBC tape between Katsuyama and Lewis there is an odd exchange (IMO) to a basic question - how do you price trades in your price matching engine. (That is where the market exists, not on the tape.)There are two answers given to the same question and one answer is avoided strangely. It seems to be the heart of the argument although I don't understand precisely what was said or implied. Can someone else explain it better?

There is a SIP and a direct feed as different entities. So is that a slow price tape (US) and a fast price tape (HFT, colocaters)? If so, then the latency matching is clearly a kind of front running isn't it? And couldn't the slow tape re-write time since only the results appear without intermediate steps shown?

I really liked the point that liquidity is not the same thing as huge volume. The hardest thing about the whole discussion is that terms are bantered about and misused. It is hard to follow the heart of the argument.

My research has uncovered this gem:

Rule 603(a)(2) requires that any SRO, broker, or dealer that distributes market information must do so on terms that are not unreasonably discriminatory. These requirements prohibit, for example, a market from making its "core data" (i.e., data that it is required to provide to a Network processor) available to vendors on a more timely basis than it makes available the core data to a Network processor.

from this document : http://www.nanex.net/aqck2/4442.html

So is the defense against front-running that computers are not human beings and so can't commit crimes? If the computers are convicted and sent to jail would they be out in milliseconds?

 

[wpfund]

At the 2 minute mark of the CNBC tape between Katsuyama and Lewis there is an odd exchange (IMO) to a basic question - how do you price trades in your price matching engine. (That is where the market exists, not on the tape.)There are two answers given to the same question and one answer is avoided strangely. It seems to be the heart of the argument although I don't understand precisely what was said or implied. Can someone else explain it better?

There is a SIP and a direct feed as different entities. So is that a slow price tape (US) and a fast price tape (HFT, colocaters)? If so, then the latency matching is clearly a kind of front running isn't it? And couldn't the slow tape re-write time since only the results appear without intermediate steps shown?

I really liked the point that liquidity is not the same thing as huge volume. The hardest thing about the whole discussion is that terms are bantered about and misused. It is hard to follow the heart of the argument.

Took me awhile to find this "debate" online (perhaps because it's 24hrs old), so posting a link to it: http://video.cnbc.com/gallery/?video=3000263252

 

[StarDust9182]

There wasn't much substance in that cnbc freeforall.

I disagree. I made a trade some 5 years ago when I hit the bid and was only filled for 2000 shares of the 14000 showing and the bid-ask moved. I phoned in and asked how this could possibly be and we went over the exchange published tape together at the precise time. It was a holiday eve and I had been watching the bid ask for some time and so off the exchange tape we went over bid and ask at the exact time (remember I was partially filled). I was astounded to find that my memory of what had occurred watching level 2 and the tape DIDN'T MATCH! The fellow didn't seem to believe me, but was unable to explain how this could possibly happen. Something was definitely wrong.

Katsuyama describes the same thing happening to Royal Bank traders. I have seen this although it is really hard to prove.

After that experience, I had a habit of capturing a screen shot (for proof) and got several odd things that occur documented. Thinking about these odd things has definitely helped my trading.

 

[stock777]

I've seen so many odd things over the years, I only notice it when it's normal.

 

[kinggyppo]

this is worth checking out, around minute 16 Naroj talks about direct feed vs consolidated
feed, assuming there is an edge in the direct feed, can anyone comment on how the SIP
feed could/should be fixed updated. I saw one other poster mention this.

http://www.businessweek.com/videos/2014-04-02/katsuyama-narang-lewis-debate-speed-trading

 

[athlonmank8]

its all in this lengthy ass article, they essentially put a giant cable in a box to slow down access, the idea being that it puts a buyer on equal footing with a high frequency firm.
They have introduced a 350 millisecond delay into their system. To me its the idea that if you are looking at 300,000 shares offered on GM that if you want to buy them the quote won't disappear. This seems to be the main gripe about HFT is the order canceling or vanishing liquidity. If you are a buy side firm this could be an advantage, that said if you are buying 300,000 GM for a sleepy mutual fund what's the difference between this and parking a fill or kill limit order on an exchange?

http://www.nytimes.com/2014/04/06/magazine/flash-boys-michael-lewis.html

Exactly

 

[athlonmank8]

My research has uncovered this gem:

Rule 603(a)(2) requires that any SRO, broker, or dealer that distributes market information must do so on terms that are not unreasonably discriminatory. These requirements prohibit, for example, a market from making its "core data" (i.e., data that it is required to provide to a Network processor) available to vendors on a more timely basis than it makes available the core data to a Network processor.

from this document : http://www.nanex.net/aqck2/4442.html

So is the defense against front-running that computers are not human beings and so can't commit crimes? If the computers are convicted and sent to jail would they be out in milliseconds?

lol

 

[baglunch]

How do you know what my design choice is? You are assuming I want it the status quo. If you really want to know. And I know you do. Here is what I'd want:

1) No dark pools
2) no internalization
3) one central exchange
4) basic limit and market orders only
5) Fix the SIP

However, we all know that won't happen. So I'd settle for...

1) only lit venues
2) no internalization
3) no sub pennies
4) basic order types
6) allow locked/crossed markets
7) fix the SIP

I agree with most of this, definitely no dark pools or internalized flow. What you describe is how the US equity market was in the 80s and 90s before REG NMS, multi listed products, or direct feeds. The only downside was that specialists controlled the liquidity and you had to be part of the club to get hit, but far from broken the way it is now. Don't blame the players, blame the SEC for breaking the game.

While the sip is improving through enhancements the problem is that the market is still too fragmented and that fragmentation is what makes the sip unusable even if it was fixed of its latency issues. I had hope with cboe shutting down the cbsx exchange that there would be one less place to distract us but soon enough iex will be added to the fray. Maybe the market will be better off if the SEC forced all of the exchanges that acquire other exchanges to consolidate, no one needs 3 exchanges under the same roof that have the same model and only differ in fees. NYSE/amex/arca, bats z/bats y/direct edge a/ direct edge x, Nasdaq/psx/bx can consolidate into 3 markets and the other indy markets can just go away. Just like how ICE bought NYSE and is cleaning house by shutting down liffe-us and selling Euronext.

Also if there is one central exchange you don't need to fix the sip nor do you need a sip at all.

 

[StarDust9182]

I agree with most of this, definitely no dark pools or internalized flow. What you describe is how the US equity market was in the 80s and 90s before REG NMS, multi listed products, or direct feeds. The only downside was that specialists controlled the liquidity and you had to be part of the club to get hit, but far from broken the way it is now. Don't blame the players, blame the SEC for breaking the game.

While the sip is improving through enhancements the problem is that the market is still too fragmented and that fragmentation is what makes the sip unusable even if it was fixed of its latency issues. I had hope with cboe shutting down the cbsx exchange that there would be one less place to distract us but soon enough iex will be added to the fray. Maybe the market will be better off if the SEC forced all of the exchanges that acquire other exchanges to consolidate, no one needs 3 exchanges under the same roof that have the same model and only differ in fees. NYSE/amex/arca, bats z/bats y/direct edge a/ direct edge x, Nasdaq/psx/bx can consolidate into 3 markets and the other indy markets can just go away. Just like how ICE bought NYSE and is cleaning house by shutting down liffe-us and selling Euronext.

Also if there is one central exchange you don't need to fix the sip nor do you need a sip at all.

Could you explain what is bad about dark pools and internalization? (I have no position on these items.) Didn't both arise because of the same front running issue we are talking about here - just done by humans originally?