Goldmans top-secret quant algorithm code stolen

[GammaJammer]

Google CEP (complex event processing) Aaron. Plenty there to keep you salivating. At some point in the next, ooh, say 12 months I predict a raft of threads on this and simiilar boards from $2k account spreadbetters asking whether they can get a news parsing algo plug-in for their MT4 or similar, so they can automatically do their little cods over payrolls rather than actually physically having to click the mouse to hand their not particularly hard earned dough straight back to the bookies.

 

[Technically Fundamental]

Thats exactly what I was thinking about Gamma.

 

[GammaJammer]

I know people at a couple of the major firms doing this stuff. Not quite doing us all out of a job yet but some interesting work being done nevertheless.

 

[gooseman]

it'll all go horribly wrong i tell ya

 

[Technically Fundamental]

i'd get used to it. pretty much every screen market has this form of "trader" in it somewhere. next time you get a fill think about why you're two ticks offside instantly ;-)

I dont get it. I just thought that was to do with actual bids/offers available from the data pools the broker uses (I've been told its several large banks :-S), assuming they hedge, which they dont with my positions so I presumed it was theft.

I'm intrigued now. What is the algo program doing? Instant Arbing?

 

[dcraig1]

Google CEP (complex event processing) Aaron. Plenty there to keep you salivating. At some point in the next, ooh, say 12 months I predict a raft of threads on this and simiilar boards from $2k account spreadbetters asking whether they can get a news parsing algo plug-in for their MT4 or similar, so they can automatically do their little cods over payrolls rather than actually physically having to click the mouse to hand their not particularly hard earned dough straight back to the bookies.

Esper is a free CEP framework. Probably a good place to start for anybody who wants to get their hands dirty.

Esper - Complex Event Processing

 

[Technically Fundamental]

I know people at a couple of the major firms doing this stuff. Not quite doing us all out of a job yet but some interesting work being done nevertheless.

Unles the macro markets react in a textbook way stimulae the formulae would need to be constantly updated to fit in with market conditions. Would be extremely hard to account for human interation too eh? How much would you lose if there was a war? Hmmm then again maybe if they incorporate some of that patriot act type technology to scan the news for names or keywords. I wouldnt be worried anyway. You guys seen how sh1t that "advanced" AI robot is? If thats the forefront then your jobs will be safe as houses lol.

All sounds a bit like a load of b0lklocks to claim a bit of tax relief on the R&D lol.


OR maybe you bankers are going the way of the miners and manufacturers!

 

[gooseman]

1987 aaron.

anyway-i think ou need to delve a bit deeper past SB markets.

 

[Technically Fundamental]

You mind beig a bit more specific Mr G? I can't handle the migranes anymore.

 

[gooseman]

causes of 1987 crash-look em up.

as for the looking pat SB-i have no idea what experience you have on exchange but if you know how they work and the order system works then you can quite easily see how algo's can get themselves between orders.

 

[Technically Fundamental]

Duuuuuuuuuuuuh. I tend to overcomplicate things. Look how I though t the algos worked lol.

 

[Hoggums]

I know people at a couple of the major firms doing this stuff. Not quite doing us all out of a job yet but some interesting work being done nevertheless.

Problem is - if it is successful - other banks/institutions will get wind and develop their own versions which will eventually render it useless. These are all self destroying once they become too popular, their behaviour changes the market behaviour which changes the fundamental conditions in which they have been programmed to trade in.

All of the quant code I've dealt with covers risk management rather than trading algos. In my opinion it's all a bit of a bodge based on assumptions. I sat there last year and watched these so called algos lose several million pounds a week trying to hedge the AUD & JPY moves. But - you have to protect against losses and there are no better systems at present than the complex math models available. Doesn't mean to say they are the be all and end all.

 

[gooseman]

i'm too apathetic to post meaningful stuff today

 

[GammaJammer]

haha probably true. I never said I ENDORSED it mate, just knew some people involved in that world. You should see these guys when they get together at networking events. Wild. We're talking bottles of Cristal, the works.

Or b.....

 

[gooseman]

i'll have a lemonade and my friend here wants an OJ. With no bits please.

 

[counter_violent]

CEP will play a bigger role in the future I recon...but perhaps the smart traders will use it as a filter of conditions....multi conditions....I know thats how I would use it.

Luckily my mate is working on this stuff....

 

[gooseman]

it will, like all technology/methodology grow until everyone thinks it is the holy grail of trading. then something out of the blue will happen, be it credit spreads, liquidity problems, iranian tactical strike, and then when everyone bolts for the same door it will cause havoc and everyone will start looking for a new system/holy grail.

 

[Trader333]

According to this It is far more serious than many have said with GS standing to lose millions:

Bloomberg.com: News

One of the key aspects of the code (which I read on TL), is:

The proprietary code lets the firm do “sophisticated, high- speed and high-volume trades on various stock and commodities markets,” prosecutors said in court papers. The trades generate “many millions of dollars” each year.

There has been work conducted to detect when these sorts of trades are going through which is sometimes known as detecting "Trade Intensity". Those who have developed models to detect when this happens are often able to capitalise on short term market reversals. It is an interesting science and appears to be quite profitable for those who have been able to do this.


Paul

 

[squall321]

karma - it's a bitch

 

[poborsky]

Looks like someone set the algo in reverse mode!