Speaking at a London conference on Tuesday, Donal Byrne, chief executive of Corvil, a high-speed trading technology company, caused a ripple of audible incredulity throughout the room when he suggested that trading speeds could be reduced to picoseconds in the not too distant future.
Josh who (a) reviews everything and (b) has a decent eye for garbage – made the obvious point. A nanosecond is a very short time. Light travels about 30cm in a nanosecond – so if you want nanosecond trading your computers need to in the same box – and probably your chips need to be on top of each. And even then it is problematic as the data for matching the trade will need to travel through the chips many times to be processed.
Light takes about 20 picoseconds to travel through a silicon chip. Presumably it has to do that a few times to complete a trade. So picosecond trading is an Einstein speed-of-light counterexample and Donal Byrne and his company are in line for a Nobel Prize.
Josh is more gentle than me – so I am gonna say it: Donal Byrne is a technologist speaking garbage.
When the leaders in the field start speaking easily identifiable egregious crap you know the game is nearing the end. And so it is for high frequency trading. If there were any decent return on capital for technology to make trading faster that return is likely to be pretty thin. (I would not invest in any fund that talks about that as its "edge".)
Of course we can all get smarter rather than faster. Including it seems Mr Byrne...
John
17 comments:
FYI.. Corvil are not a trading firm. They get their business measuring latency in networks and systems.
Wisdom can sometimes take a lifetime, in which pinosecond execution may not be of much use.
Cheers
Yeah - and the CEO thinks he can measure the time at which it takes light to travel the width of a grain of ground pepper.
And he thinks this matters.
It might matter whether you put ina 10m cable or a 5m cable (say 15 nanoseconds). But picoseconds.
He is selling bullshit.
J
I don't believe in Picoseconds, but there might be quantum entanglement in the future... ;-)
Pico-seconds is nonsense of course, embarrassing to the speaker, but so also is talk of nanoseconds (except, arguably, delays of multiple hundred nanos and even then it would need to be compounded by many consecutive events to have the tiniest hope of being relevant).
But at some point you need to forget about the technology and think about the economics ... what mistake is someone making that there is even a conceivably a "pico" second race to pick him off? If I am a sucker (sorry, "alpha-less individual investor") and my foolishness is worth so much money, eventually there should be some trading venue or middleman that (statistically speaking) restores some of these obvious profits to me. (Today we call this "price improvement".)
Basically, it seems to me, we will (I hope) see this is not viable on economic grounds well before speed-of-light limitations show their head.
It's also interesting he is talking in Europe. The very best few of US equity exchanges are at a point where about 50 microseconds is a reliably measurable edge. In Europe? Anywhere in the world including US for any non-equities? At least one order of magnitude behind and in Europe often two orders so. Major future exchanges - again, more often two orders than one. (And yet capitalism still functions, how about that?)
Well, maybe it's 50 megapicoseconds, LOL.
Way back when, I got one of Adm. Hopper's nanoseconds. It's a roughly foot long piece of telephone wire that she gave out at lectures. She also had a handy example for picoseconds, a pepper flake.
So yeah, not expecting picosecond trade execution any time soon.
"Donal Byrne is a technologist speaking garbage"
or eveb shorter:
Donal = Dogbaert
I can say this as an Irishman, I think Donal was being a bit 'flaithiúilach' looking for a bit of a sound byte / get the name out. It got picked up by a few spam, sorry, news services so seems it worked.
Corvil do some serious physics / have a lot of phds in the room so I'm sure he's getting a bit of a slagging, but it's hardly in the egregious space. He used all the right weasel words as any CEO / market bot would.
Just his 15 femtoseconds of fame!
Have you noticed that microprocessor speeds have hit the buffers? Ive watched them rise from 25MHz in the first Windows 3.1 machines to 3GHz - at which point, if I've got my sums right, light (and therefore limit for electrons) can travel only 10cm in one clock cycle. Starting to hit physical constraints as far as clock speeds are concerned.
Back to the China Agritech topic:
I am process engineer and therefore interested in their claimed production.
First point:
I see packing/botteling machines, BUT NO SYNTHESIS of the fertilizer. Only downstream machinery and no upstream machinery!
Second point:
Picture of botteling machine
http://www.chinaagritechinc.com/2/2.JPG
I was looking for that Ximgda company. It doesn't exist. I entered the machinery type ZCG-12D in Google. Et voila
http://image.cn.made-in-china.com/2f0j01jBSTHGEKsfca/%E5%BE%AE%E7%94%B5%E8%84%91%E7%81%8C%E8%A3%85%E6%9C%BA+%28ZCG%E2%80%9410%E2%80%9412D%E5%9E%8B%29.jpg
Same machine, other manufacturer. This manufacturer seems to exist somehow. Interesting is, that in the CAGC machine the lower part of the manufacturer is cut off (where the name should be) but another writing is on the window. The typing to the left and right of "ZCG-12D" is more or less the same (I don't know Chinese).
Strange is this picture:
http://www.chinaagritechinc.com/anhui/7.jpg
You see open conveyor belts. You use open conveyor belts for stone and or large particels, and not for dusty products. You would have a massive dust-explosion danger with an open system. I wanted to check the product safety sheets, but there are no such sheets available. Maybe I should ask for one :-)
Don't you measure the speed of your car in Angstroem per week?
@Laban It's heat and energy issues. Higher clock speed == more energy == more heat dissipation requirements.
Its being rubbished in the industry I actually got sen the link by Giles Nelson one of the inventors of Algo trading through the Progress Apama platform. Having sold a few of these in the OZ market raw speed is a relatively minor consideration, its not the speed of the trade and its cognition and complexity of the event that causes the trade. For instance in the FX markets a algo trading platform might straddle many physically separate exchanges and through event recognition across these markets creating a virtualised single market . The old systems used to persist data and access via an index, trading technology now persists the index and runs the data across it. This is the change thats allowed the massive increases in trading cognition and speed. if you ever want to a more in dept piece on it I can put you in touch with some of the leading thinkers.
ctindale
Anonymous said...
I don't believe in Picoseconds, but there might be quantum entanglement in the future... ;-)
March 8, 2011 8:23 AM
There is quantum entanglement presently, Anonymous.
What if you can make a fairly accurate prediction of what will happen in 1ms? Then "picosecond" HFT would be possible in that you could time orders to be placed and executed within picoseconds of your prediction.
Actually John, since signals do not propagate in circuits at the 30 cm/nanosecond that they would in free space but instead move even more slowly, something closer to 20 cm/nanosecond (the "velocity factor" of an on-chip trace or between-chip circuit-board trace, viewed as a transmission line, is around 2/3), everything you said is not only true, it's true in spades!
-JeffC (former electrical-engineering professor)
Post a Comment