This is why following procedures matters



  • Recently, Knight Capital goofed, and lost $400+MM by putting in many errant program-generated trades.

    It turns out that they installed new software, but inadvertently ran old software - in parallel. The old trading software multiplied trades by 1000, effectively amplifying the program trading logic of the new trading software, but out of proportion to what the new algorithms were trying to do. 

    I know for a fact that Knight has proper procedures to make sure the correct software is installed in the correct way, with the correct approvals.

    This is what happens when someone decides to just bypass the system for a quick change.

    The only difference is that this time, it was with real money; it almost took the firm down, and it became very public.



  • Crappy thing is that the person who decided to bypass the normal testing procedure is prbly going to find a way to deflect the blame somewhere else.



  • @this_code_sucks said:

    Crappy thing is that the person who decided to bypass the normal testing procedure is prbly going to find a way to deflect the blame somewhere else start working alongside snoofle next week.
    RTFY



  • @snoofle said:

    I know for a fact that Knight has proper procedures to make sure the correct software is installed in the correct way, with the correct approvals.

    At first when the problem happened but no information had been published your theory was that it was a problem with AI making bad trading decisions. But now that they released the details of what actually happened (and that has nothing to do with AI) you *know* that they have correct procedures and the problem is someone deciding to "bypass the system for a quick change".

    George Carlin said it all



  • @Speakerphone Dude said:

    @snoofle said:

    I know for a fact that Knight has proper procedures to make sure the correct software is installed in the correct way, with the correct approvals.

    At first when the problem happened but no information had been published your theory was that it was a problem with AI making bad trading decisions. But now that they released the details of what actually happened (and that has nothing to do with AI) you *know* that they have correct procedures and the problem is someone deciding to "bypass the system for a quick change".

    George Carlin said it all

    They always did have good procedures. I don't work there (although I know several who do), and didn't know what actually happened (hence: theory). it could easily have been AI run amok (see: flash crash, which I pointed out in my original post). As it turns out, it was old AI (that wasn't supposed to be running) modifying new AI without the knowledge of new AI.

    That's not AI. It's a lack of "I". And not following procedures.



  • @snoofle said:

    @Speakerphone Dude said:

    @snoofle said:

    I know for a fact that Knight has proper procedures to make sure the correct software is installed in the correct way, with the correct approvals.

    At first when the problem happened but no information had been published your theory was that it was a problem with AI making bad trading decisions. But now that they released the details of what actually happened (and that has nothing to do with AI) you *know* that they have correct procedures and the problem is someone deciding to "bypass the system for a quick change".

    George Carlin said it all

    They always did have good procedures. I don't work there (although I know several who do), and didn't know what actually happened (hence: theory). it could easily have been AI run amok (see: flash crash, which I pointed out in my original post). As it turns out, it was old AI (that wasn't supposed to be running) modifying new AI without the knowledge of new AI.

    That's not AI. It's a lack of "I". And not following procedures.

    You were wrong about "flash crash" then and you are still wrong now. This situation has nothing in common with "flash crash" and high-frequency trading other than taking place in financial services; it's bad automation, which is totally different. If you want a similar situation, look at the RBS IT failure.

    If you have actual experience in the financial services industry you would know that this kind of FUD about AI and flash crash is a shame. It's infotainment, nothing more, and all it will do is damage the reputation of excellent systems that have been running for decades. Saying that automated trading is bad is like saying that automated assembly lines are evil and that computers and cars should be crafted by hand. This is ridiculous.

    Knight fucked up. They pay the price for it. End of story, no need for collective psychosis about Skynet taking over Wall Street.



  • @Speakerphone Dude said:

    saying that automated assembly lines are evil and that computers and cars should be crafted by hand
     

    Wow this is like a meta-car-metaphor.



  • @Speakerphone Dude said:

    ...
    You seem to be taking plausible explanations for what, in my personal experience, is far too frequent an occurrence, way too personally.

    Yes, Knight screwed up. They should eat their mistake, and they will.

    But AI has also been responsible for numerous unintended mountains of trades, although not at this magnitude, and not always publicized. It can and does go wrong, far more frequently than people know. Yes, referencing skynet was a bit of an exaggeration, but automated trading CAN be bad - when it goes wrong. Like any other automated task.

    OBTW; you are quick to point out "collective psychosis" after the fact. But until the details came out, nobody, including me, knew what actually happened. I was merely speculating, and after worknig in the trading business for more than 20 years, more often than not, I'd be right. In this case, I wasn't. Big deal.

    Lighten up, Dude!



  • Speakerphone Dude, shut up and quit threadsh--ing. Your personal attacks are getting boring, and they're not even intended to be constructive like blakeyrat's. Post some WTFs for a change.

    With that out of the way...
    @snoofle said:

    it was old AI (that wasn't supposed to be running) modifying new AI without the knowledge of new AI.

    I wonder which of Knight's AI's would win a game of Core War?



  • @TwelveBaud said:

    Speakerphone Dude, shut up and quit threadsh--ing. Your personal attacks are getting boring, and they're not even intended to be constructive like blakeyrat's. Post some WTFs for a change.
     

    Was just gonna say, I hate blakeyrat's guts, but Speakerphone Dude just really makes me long for some blakeyrants.



  • @pbean said:

    @TwelveBaud said:

    Speakerphone Dude, shut up and quit threadsh--ing. Your personal attacks are getting boring, and they're not even intended to be constructive like blakeyrat's. Post some WTFs for a change.
     

    Was just gonna say, I hate blakeyrat's guts, but Speakerphone Dude just really makes me long for some blakeyrants.

    There is a difference between doing personal attacks and refusing to treat snoofle as the Steve Jobs of this forum. He is a big boy and like everyone else if he makes a comment that appears to be wrong it is fair game to make a fuss about it. And if he behaves like Uncle Snoofle entertaining the children with tales of IT bravery and FUD about Wall Street while the content is weak, it is also fair game to comment on it.

    Also the two of you are doing exactly what you are complaining about; you are like two old women bitching over the phone about people who are bitching over the phone. So if you are not happy with my comments just go back to brown-nosing snoofle; you can even print his avatar (or blakeyrat's) and use it as jerk-off material for all I care.



  • Thank you for contributing your insights to this forum, Speakerphone. You have challenged us and given us much to think about in ways that will surely improve our lives and careers.



  • @Speakerphone Dude said:

    ...

    Knight fucked up. They pay the price for it. End of story, no need for collective psychosis about Skynet taking over Wall Street.

     

    You can laud the virtues of automated trading all you want, but the simple fact remains: the flash crash wouldn't have happened without automated trading.  Trying to argue that the occassional automated system going haywire is worth the benefits that automated systems bring is different than what you did, which is to claim that anybody pointing out a particular danger of automated systems is clearly an idiot who doesn't know anything about the financial industry.

     



  • @Speakerphone Dude said:

    treat snoofle as the S**** J*** of this forum

    Nobody deserves that treatment and you should insult people using such strong words, shame on you sir, shame on you

    CIFY


  • Accept no imitations.



  • @ShatteredArm said:

    You can laud the virtues of automated trading all you want, but the simple fact remains: the flash crash wouldn't have happened without automated trading.

    While that's technically true, it misses the point: virtually nobody objects to automated trading; people object to flash trading.

    Saying flash trading can't exist without automated training is, while true, not particularly relevant to people's complaints. For example, the market could simply enforce a maximum trade rate somehow (say, each trade station gets an hourly quota of 20 trades, no more) and that would stop flash trading without hurting automated trading at all.

    In any case, it's dumb to object to automated trading because it's not going to go away. Even if you have to register a thumbprint to prove you're human before making a trade, there's nothing to stop a human from just getting instructions from the automated trading system. All you've done is inconvenienced everybody without changing the status quo. You might as well complain about the freeway system-- it ain't going away, you better learn to cope.



  • @ShatteredArm said:

    @Speakerphone Dude said:

    ...

    Knight fucked up. They pay the price for it. End of story, no need for collective psychosis about Skynet taking over Wall Street.

     

    You can laud the virtues of automated trading all you want, but the simple fact remains: the flash crash wouldn't have happened without automated trading.  Trying to argue that the occassional automated system going haywire is worth the benefits that automated systems bring is different than what you did, which is to claim that anybody pointing out a particular danger of automated systems is clearly an idiot who doesn't know anything about the financial industry.

     

    Automated trading is not the problem. Bad automated trading is. The comment which hinted at micro-trading and hedge funds AI activities was a red herring; it was misleading and off-topic because it's not the type of activity that Knight is doing. It falls in the same category as the news reporter saying that Angry Birds could be weaponized because it's done with LUA (or something).

    I'm not saying that Uncle Snoofle is an idiot but he should know better than feed the children this kind of FUD. The fact that the financial services industry is full of WTF does not make it less unfair.



  •  @Speakerphone Dude said:

    Automated trading is not the problem. Bad automated trading is. The comment which hinted at micro-trading and hedge funds AI activities was a red herring; it was misleading and off-topic because it's not the type of activity that Knight is doing. It falls in the same category as the news reporter saying that Angry Birds could be weaponized because it's done with LUA (or something).

    I'm not saying that Uncle Snoofle is an idiot but he should know better than feed the children this kind of FUD. The fact that the financial services industry is full of WTF does not make it less unfair.

     You do realize this is the internet right?  I found nothing like a crazed rant in snoofles original post.  Instead it sounded more like caution from someone who actually works for those companies and thus understands how they operate, specualting on a possible cause.  

    And besides this is a whole different thread, although related, on a different topic with a different point behind it.   I really don't think there is a reason to get angry.

     



  • @galgorah said:

    I really don't think there is a reason to get angry.

    @galgorah said:
    You do realize this is the internet right? 

    You can find the answer in your own post



  • @snoofle said:

    They always did have good procedures. I don't work there (although I know several who do), and didn't know what actually happened (hence: theory). it could easily have been AI run amok (see: flash crash, which I pointed out in my original post). As it turns out, it was old AI (that wasn't supposed to be running) modifying new AI without the knowledge of new AI.

    That's not AI. It's a lack of "I". And not following procedures.

    You do realize that AI is short for Artifical Idiot?


Log in to reply