# Time to Rein In Algorithms? Calls for algo certification growing louder **URL:** \ **Category:** General Q&A\ **Tags:** algorithmic-trading, xid\_169233, imported\ **Created:** [May 10, 2010, 3:18pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605 "2010-05-10T15:18:34Z")\ **Posts on this page:** 14\ **Page:** 1
**Author:** ![kaymahes\_x](https://avatars.discourse-cdn.com/v4/letter/k/c2a13f/32.png) [@kaymahes\_x](https://forum.fixtrading.org/u/kaymahes_x)\ **Post date:** [May 10, 2010, 3:18pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/1 "2010-05-10T15:18:34Z")
Imported from previous forum ---
**Author:** ![kaymahes\_x](https://avatars.discourse-cdn.com/v4/letter/k/c2a13f/32.png) [@kaymahes\_x](https://forum.fixtrading.org/u/kaymahes_x)\ **Post date:** [May 10, 2010, 3:18pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/2 "2010-05-10T15:18:34Z")
May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) While trading algorithms are regularly praised for their speed and efficiency, a growing contingent of professional traders, financial engineers, consultants and academics say they are being misused, or are faulty from the start. Their answer? Establish best practices and increase training to keep algorithms from doing more damage than good. The impetus for the movement is the fear that the availability of sophisticated trading strategies to a wider audience with varying levels of expertise could cause havoc in the markets–and already has, in the view of many. If the industry does not do something to fix the problem, then regulators may have to take action, with potentially disastrous results for the algorithmic trading business. “If you are a professional trading firm, you have to certify that you are following the FIX protocol in your trading activities. But there is no equivalent to that–a driver’s license, so to speak–for algorithmic trading,” pointed out Michel Debiche, a veteran of the proprietary trading desks of CIBC World Markets, Daiwa Securities America and Credit Suisse. Full article :- [http://www.securitiesindustry.com/news/-25303-1.html?zkPrintable=true](http://www.securitiesindustry.com/news/-25303-1.html?zkPrintable=true) Is there any Algo certification presently available in FIXProtocol ? If yes, can I have the URL of the relevant documents ? If no, is there any plan to create an algorithmic trading certification test suite ? Regards, K. Mahesh ---
**Author:** ![fplpo\_x](https://avatars.discourse-cdn.com/v4/letter/f/7ab992/32.png) [@fplpo\_x](https://forum.fixtrading.org/u/fplpo_x)\ **Post date:** [May 10, 2010, 4:06pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/3 "2010-05-10T16:06:53Z")
[original email was from John Greenan - [john.greenan@alignment-systems.com](mailto:john.greenan@alignment-systems.com) ] Isn’t this a case of what the Americans refer to as “Guns don’t kill people, people kill people”?? If you give access to the latest sniper / hunter-killer / guerilla / dominator etc algo to a user who has no understanding of the way that an algo works then really isn’t this an issue for the head of trading rather than the CTO??? Also, there’s a considerable element of talking your own book isn’t there: “Debiche, CEO and founder of Princeton, N.J.-based Quantia Capital Management, which advises buy-side firms and technology suppliers on quantitative trading systems … Debiche said that algorithms need to be certified…certification, ideally, would be handled by a third party not conflicted by the need to approve algorithms and start collecting commissions as soon as possible.” Oh, so maybe a firm which advises buy-side firms on quantitative trading systems??? > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > While trading algorithms are regularly praised for their speed and > efficiency, a growing contingent of professional traders, financial > engineers, consultants and academics say they are being misused, or are > faulty from the start. Their answer? Establish best practices and > increase training to keep algorithms from doing more damage than good. > > The impetus for the movement is the fear that the availability of > sophisticated trading strategies to a wider audience with varying levels > of expertise could cause havoc in the markets–and already has, in the > view of many. If the industry does not do something to fix the problem, > then regulators may have to take action, with potentially disastrous > results for the algorithmic trading business. > > “If you are a professional trading firm, you have to certify that you > are following the FIX protocol in your trading activities. But there is > no equivalent to that–a driver’s license, so to speak–for algorithmic > trading,” pointed out Michel Debiche, a veteran of the proprietary > trading desks of CIBC World Markets, Daiwa Securities America and > Credit Suisse. > > Full article :- > > [Resume Builder Online [Free] – Create A Professional Resume In 5 Minutes | ResumeBuild](http://www.securitiesindustry.com/news/-25303-1.html?zkPrintable=true) > > Is there any Algo certification presently available in FIXProtocol ? If > yes, can I have the URL of the relevant documents ? If no, is there any > plan to create an algorithmic trading certification test suite ? > > Regards, > K. Mahesh ---
**Author:** ![russellcurry\_x](https://avatars.discourse-cdn.com/v4/letter/r/d07c76/32.png) [@russellcurry\_x](https://forum.fixtrading.org/u/russellcurry_x)\ **Post date:** [May 10, 2010, 5:16pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/4 "2010-05-10T17:16:22Z")
It’s an interesting, and reasonable, question, but it doesn’t make sense to position this as a problem with the experience level of the users of the algorithms - it seems to me that this is more a question of needing better risk management capabilities on the side of the systems that are interacting with the algorithms. > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > While trading algorithms are regularly praised for their speed and > efficiency, a growing contingent of professional traders, financial > engineers, consultants and academics say they are being misused, or are > faulty from the start. Their answer? Establish best practices and > increase training to keep algorithms from doing more damage than good. > > The impetus for the movement is the fear that the availability of > sophisticated trading strategies to a wider audience with varying levels > of expertise could cause havoc in the markets–and already has, in the > view of many.(…) ---
**Author:** ![jaimeromanini\_x](https://avatars.discourse-cdn.com/v4/letter/j/ecd19e/32.png) [@jaimeromanini\_x](https://forum.fixtrading.org/u/jaimeromanini_x)\ **Post date:** [May 10, 2010, 5:58pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/5 "2010-05-10T17:58:01Z")
Indeed, in my point of view The algorithms must not be controlled by algorithms, but the market monitoring or risk management that should be a completely independent infrastructure. Regards Jaime Romanini > It’s an interesting, and reasonable, question, but it doesn’t make sense > to position this as a problem with the experience level of the users of > the algorithms - it seems to me that this is more a question of needing > better risk management capabilities on the side of the systems that are > interacting with the algorithms. > > > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > > > While trading algorithms are regularly praised for their speed and > > efficiency, a growing contingent of professional traders, financial > > engineers, consultants and academics say they are being misused, or > > are faulty from the start. Their answer? Establish best practices and > > increase training to keep algorithms from doing more damage than good. > > > > The impetus for the movement is the fear that the availability of > > sophisticated trading strategies to a wider audience with varying > > levels of expertise could cause havoc in the markets–and already has, > > in the view of many.(…) ---
**Author:** ![stephenedge63\_x](https://avatars.discourse-cdn.com/v4/letter/s/ebca7d/32.png) [@stephenedge63\_x](https://forum.fixtrading.org/u/stephenedge63_x)\ **Post date:** [May 11, 2010, 3:36am UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/6 "2010-05-11T03:36:51Z")
Considering the sheer volume of trades/orders generated by algos a tool to monitor what is going on will be necessary. It can’t be done by humans or who would bare the cost? I believe its the lack of tools and the evolution of this type of trading that is way ahead of the ability to monitor what these algos are doing. They are unpoliced and probably will be for sometime. > Indeed, in my point of view > > The algorithms must not be controlled by algorithms, but the market > monitoring or risk management that should be a completely independent > infrastructure. > > Regards > > Jaime Romanini > > > It’s an interesting, and reasonable, question, but it doesn’t make > > sense to position this as a problem with the experience level of the > > users of the algorithms - it seems to me that this is more a question > > of needing better risk management capabilities on the side of the > > systems that are interacting with the algorithms. > > > > > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > > > > > While trading algorithms are regularly praised for their speed and > > > efficiency, a growing contingent of professional traders, financial > > > engineers, consultants and academics say they are being misused, or > > > are faulty from the start. Their answer? Establish best practices > > > and increase training to keep algorithms from doing more damage > > > than good. > > > > > > The impetus for the movement is the fear that the availability of > > > sophisticated trading strategies to a wider audience with varying > > > levels of expertise could cause havoc in the markets–and already > > > has, in the view of many.(…) ---
**Author:** ![fplpo\_x](https://avatars.discourse-cdn.com/v4/letter/f/7ab992/32.png) [@fplpo\_x](https://forum.fixtrading.org/u/fplpo_x)\ **Post date:** [May 11, 2010, 4:48am UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/7 "2010-05-11T04:48:58Z")
[original email was from John Harris - [john.harris@bondmart.com](mailto:john.harris@bondmart.com) ] Did algorithms lie about the market value of assets on the balance sheets of big banks? Did algorithms create the “too big to fail doctrine” and saddle future generations of children, some not yet born, with debts they didn’t incur and can never repay? Did algorithms create fiat currencies or lever equity 100 fold? Did algorithms mandate the payout of 50% of bank revenues to people other than shareholders? Did algorithms create the absurd, childish notion of a “national market system” or approve bank holding company applications of Goldman Sachs and Morgan Stanley in two days, over a weekend? Did algorithms assign Aaa/AAA credit ratings to third derivatives of pools of stated-income loans? Did algorithms put the the full faith and credit of the people of the United States behind the liabilities of GE Capital, American Express, and a host of other fascist companies that have been allowed by corrupt politicians to make one-way bets on the backs of present and futures generations of taxpayers? Might it just be that all of us who love our families and children, who understand the economic importance of competition and free markets, have better objects of concern at the present moment than algorithmic trading? Sheesh. I guess the Romans proved that bread and circuses really do work. > Considering the sheer volume of trades/orders generated by algos a tool > to monitor what is going on will be necessary. It can’t be done by > humans or who would bare the cost? I believe its the lack of tools and > the evolution of this type of trading that is way ahead of the ability > to monitor what these algos are doing. They are unpoliced and probably > will be for sometime. > > > Indeed, in my point of view > > > > The algorithms must not be controlled by algorithms, but the market > > monitoring or risk management that should be a completely independent > > infrastructure. > > > > Regards > > > > Jaime Romanini > > > > > It’s an interesting, and reasonable, question, but it doesn’t make > > > sense to position this as a problem with the experience level of the > > > users of the algorithms - it seems to me that this is more a > > > question of needing better risk management capabilities on the side > > > of the systems that are interacting with the algorithms. > > > > > > > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > > > > > > > While trading algorithms are regularly praised for their speed and > > > > efficiency, a growing contingent of professional traders, > > > > financial engineers, consultants and academics say they are being > > > > misused, or are faulty from the start. Their answer? Establish > > > > best practices and increase training to keep algorithms from doing > > > > more damage than good. > > > > > > > > The impetus for the movement is the fear that the availability of > > > > sophisticated trading strategies to a wider audience with varying > > > > levels of expertise could cause havoc in the markets–and already > > > > has, in the view of many.(…) ---
**Author:** ![jim.northey](https://yyz2.discourse-cdn.com/flex010/user_avatar/forum.fixtrading.org/jim.northey/32/37_2.png) [@jim.northey](https://forum.fixtrading.org/u/jim.northey)\ **Post date:** [May 11, 2010, 5:43am UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/8 "2010-05-11T05:43:19Z")
Not to all the above. But when I looked at that stock chart from May 6th going into the next trading day - I thought where have I seen such a pattern before? Then I thought - yes - non-linear dynamics. Things are so complex and inter-related and all driven off of the same set of spurious signals with multiple levels of feedback loops that we have have created this incredibly system that is now exhibiting chaotic behaviors. Algos are just a part - HF, multiple venues. It is just an incredibly complex system. Maybe the SEC should surveil the system with software designed to track and monitor earthquakes and heart attacks. And John - nice to hear you chime in - The Romans have lead poisoning to blame for their decline - what is our equivalent? High fructose corn syrup and Facebook? 🙂 My two cents (before discounting) we have some good thoughts and observations on this thread and we do need to follow larger industry trends and begin to analyse the masses of trading data to understand the overall system behavior. Someone said it best forget where I heard it within the last two days, maybe the SEC should lay off a few dozen lawyers and hire some tech savvy people that understand complex system behavior. Heck with the calvary - call in the Santa Fe Institute! > Did algorithms lie about the market value of assets on the balance > sheets of big banks? > > Did algorithms create the “too big to fail doctrine” and saddle future > generations of children, some not yet born, with debts they didn’t incur > and can never repay? > > Did algorithms create fiat currencies or lever equity 100 fold? > > Did algorithms mandate the payout of 50% of bank revenues to people > other than shareholders? > > Did algorithms create the absurd, childish notion of a “national market > system” or approve bank holding company applications of Goldman Sachs > and Morgan Stanley in two days, over a weekend? > > Did algorithms assign Aaa/AAA credit ratings to third derivatives of > pools of stated-income loans? > > Did algorithms put the the full faith and credit of the people of the > United States behind the liabilities of GE Capital, American Express, > and a host of other fascist companies that have been allowed by corrupt > politicians to make one-way bets on the backs of present and futures > generations of taxpayers? > > Might it just be that all of us who love our families and children, > who understand the economic importance of competition and free > markets, have better objects of concern at the present moment than > algorithmic trading? > > Sheesh. > > I guess the Romans proved that bread and circuses really do work. > > > Considering the sheer volume of trades/orders generated by algos a > > tool to monitor what is going on will be necessary. It can’t be done > > by humans or who would bare the cost? I believe its the lack of tools > > and the evolution of this type of trading that is way ahead of the > > ability to monitor what these algos are doing. They are unpoliced and > > probably will be for sometime. > > > > > Indeed, in my point of view > > > > > > The algorithms must not be controlled by algorithms, but the market > > > monitoring or risk management that should be a completely > > > independent infrastructure. > > > > > > Regards > > > > > > Jaime Romanini > > > > > > > It’s an interesting, and reasonable, question, but it doesn’t make > > > > sense to position this as a problem with the experience level of > > > > the users of the algorithms - it seems to me that this is more a > > > > question of needing better risk management capabilities on the > > > > side of the systems that are interacting with the algorithms. > > > > > > > > > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > > > > > > > > > While trading algorithms are regularly praised for their speed > > > > > and efficiency, a growing contingent of professional traders, > > > > > financial engineers, consultants and academics say they are > > > > > being misused, or are faulty from the start. Their answer? > > > > > Establish best practices and increase training to keep > > > > > algorithms from doing more damage than good. > > > > > > > > > > The impetus for the movement is the fear that the availability > > > > > of sophisticated trading strategies to a wider audience with > > > > > varying levels of expertise could cause havoc in the markets– > > > > > and already has, in the view of many.(…) ---
**Author:** ![edbuchwalter\_x](https://avatars.discourse-cdn.com/v4/letter/e/7993a0/32.png) [@edbuchwalter\_x](https://forum.fixtrading.org/u/edbuchwalter_x)\ **Post date:** [May 11, 2010, 7:58am UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/9 "2010-05-11T07:58:29Z")
> Not to all the above. > > But when I looked at that stock chart from May 6th going into the next > trading day - I thought where have I seen such a pattern before? Then I > thought - yes - non-linear dynamics. Things are so complex and inter- > related and all driven off of the same set of spurious signals with > multiple levels of feedback loops that we have have created this > incredibly system that is now exhibiting chaotic behaviors. Algos are > just a part - HF, multiple venues. It is just an incredibly complex > system. Maybe the SEC should surveil the system with software designed > to track and monitor earthquakes and heart attacks. > > And John - nice to hear you chime in - The Romans have lead poisoning to > blame for their decline - what is our equivalent? High fructose corn > syrup and Facebook? 🙂 > > My two cents (before discounting) we have some good thoughts and > observations on this thread and we do need to follow larger industry > trends and begin to analyse the masses of trading data to understand the > overall system behavior. > > Someone said it best forget where I heard it within the last two days, > maybe the SEC should lay off a few dozen lawyers and hire some tech > savvy people that understand complex system behavior. > > Heck with the calvary - call in the Santa Fe Institute! > > > Did algorithms lie about the market value of assets on the balance > > sheets of big banks? > > > > Did algorithms create the “too big to fail doctrine” and saddle future > > generations of children, some not yet born, with debts they didn’t > > incur and can never repay? > > > > Did algorithms create fiat currencies or lever equity 100 fold? > > > > Did algorithms mandate the payout of 50% of bank revenues to people > > other than shareholders? > > > > Did algorithms create the absurd, childish notion of a “national > > market system” or approve bank holding company applications of Goldman > > Sachs and Morgan Stanley in two days, over a weekend? > > > > Did algorithms assign Aaa/AAA credit ratings to third derivatives of > > pools of stated-income loans? > > > > Did algorithms put the the full faith and credit of the people of the > > United States behind the liabilities of GE Capital, American Express, > > and a host of other fascist companies that have been allowed by > > corrupt politicians to make one-way bets on the backs of present and > > futures generations of taxpayers? > > > > Might it just be that all of us who love our families and children, > > who understand the economic importance of competition and free > > markets, have better objects of concern at the present moment than > > algorithmic trading? > > > > Sheesh. > > > > I guess the Romans proved that bread and circuses really do work. > > > > > Considering the sheer volume of trades/orders generated by algos a > > > tool to monitor what is going on will be necessary. It can’t be done > > > by humans or who would bare the cost? I believe its the lack of > > > tools and the evolution of this type of trading that is way ahead of > > > the ability to monitor what these algos are doing. They are > > > unpoliced and probably will be for sometime. > > > > > > > Indeed, in my point of view > > > > > > > > The algorithms must not be controlled by algorithms, but the > > > > market monitoring or risk management that should be a completely > > > > independent infrastructure. > > > > > > > > Regards > > > > > > > > Jaime Romanini > > > > > > > > > It’s an interesting, and reasonable, question, but it doesn’t > > > > > make sense to position this as a problem with the experience > > > > > level of the users of the algorithms - it seems to me that this > > > > > is more a question of needing better risk management > > > > > capabilities on the side of the systems that are interacting > > > > > with the algorithms. > > > > > > > > > > > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > > > > > > > > > > > While trading algorithms are regularly praised for their speed > > > > > > and efficiency, a growing contingent of professional traders, > > > > > > financial engineers, consultants and academics say they are > > > > > > being misused, or are faulty from the start. Their answer? > > > > > > Establish best practices and increase training to keep > > > > > > algorithms from doing more damage than good. > > > > > > > > > > > > The impetus for the movement is the fear that the availability > > > > > > of sophisticated trading strategies to a wider audience with > > > > > > varying levels of expertise could cause havoc in the markets– > > > > > > and already has, in the view of many.(…) ---
**Author:** ![robertsmythe\_x](https://avatars.discourse-cdn.com/v4/letter/r/e19adc/32.png) [@robertsmythe\_x](https://forum.fixtrading.org/u/robertsmythe_x)\ **Post date:** [May 11, 2010, 12:50pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/10 "2010-05-11T12:50:42Z")
> I agree with Jim Northey. What happened last week appears to be caused by the interaction of several complex processes. No useful information came out of the SEC and exchanges on Monday so this brings me back to the old management principle. If you cannot measure it you cannot manage it. Not to all the above. > But when I looked at that stock chart from May 6th going into the next > trading day - I thought where have I seen such a pattern before? Then I > thought - yes - non-linear dynamics. Things are so complex and inter- > related and all driven off of the same set of spurious signals with > multiple levels of feedback loops that we have have created this > incredibly system that is now exhibiting chaotic behaviors. Algos are > just a part - HF, multiple venues. It is just an incredibly complex > system. Maybe the SEC should surveil the system with software designed > to track and monitor earthquakes and heart attacks. > > And John - nice to hear you chime in - The Romans have lead poisoning to > blame for their decline - what is our equivalent? High fructose corn > syrup and Facebook? 🙂 > > My two cents (before discounting) we have some good thoughts and > observations on this thread and we do need to follow larger industry > trends and begin to analyse the masses of trading data to understand the > overall system behavior. > > Someone said it best forget where I heard it within the last two days, > maybe the SEC should lay off a few dozen lawyers and hire some tech > savvy people that understand complex system behavior. > > Heck with the calvary - call in the Santa Fe Institute! > > > Did algorithms lie about the market value of assets on the balance > > sheets of big banks? > > > > Did algorithms create the “too big to fail doctrine” and saddle future > > generations of children, some not yet born, with debts they didn’t > > incur and can never repay? > > > > Did algorithms create fiat currencies or lever equity 100 fold? > > > > Did algorithms mandate the payout of 50% of bank revenues to people > > other than shareholders? > > > > Did algorithms create the absurd, childish notion of a “national > > market system” or approve bank holding company applications of Goldman > > Sachs and Morgan Stanley in two days, over a weekend? > > > > Did algorithms assign Aaa/AAA credit ratings to third derivatives of > > pools of stated-income loans? > > > > Did algorithms put the the full faith and credit of the people of the > > United States behind the liabilities of GE Capital, American Express, > > and a host of other fascist companies that have been allowed by > > corrupt politicians to make one-way bets on the backs of present and > > futures generations of taxpayers? > > > > Might it just be that all of us who love our families and children, > > who understand the economic importance of competition and free > > markets, have better objects of concern at the present moment than > > algorithmic trading? > > > > Sheesh. > > > > I guess the Romans proved that bread and circuses really do work. > > > > > Considering the sheer volume of trades/orders generated by algos a > > > tool to monitor what is going on will be necessary. It can’t be done > > > by humans or who would bare the cost? I believe its the lack of > > > tools and the evolution of this type of trading that is way ahead of > > > the ability to monitor what these algos are doing. They are > > > unpoliced and probably will be for sometime. > > > > > > > Indeed, in my point of view > > > > > > > > The algorithms must not be controlled by algorithms, but the > > > > market monitoring or risk management that should be a completely > > > > independent infrastructure. > > > > > > > > Regards > > > > > > > > Jaime Romanini > > > > > > > > > It’s an interesting, and reasonable, question, but it doesn’t > > > > > make sense to position this as a problem with the experience > > > > > level of the users of the algorithms - it seems to me that this > > > > > is more a question of needing better risk management > > > > > capabilities on the side of the systems that are interacting > > > > > with the algorithms. > > > > > > > > > > > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > > > > > > > > > > > While trading algorithms are regularly praised for their speed > > > > > > and efficiency, a growing contingent of professional traders, > > > > > > financial engineers, consultants and academics say they are > > > > > > being misused, or are faulty from the start. Their answer? > > > > > > Establish best practices and increase training to keep > > > > > > algorithms from doing more damage than good. > > > > > > > > > > > > The impetus for the movement is the fear that the availability > > > > > > of sophisticated trading strategies to a wider audience with > > > > > > varying levels of expertise could cause havoc in the markets– > > > > > > and already has, in the view of many.(…) ---
**Author:** ![fplpo\_x](https://avatars.discourse-cdn.com/v4/letter/f/7ab992/32.png) [@fplpo\_x](https://forum.fixtrading.org/u/fplpo_x)\ **Post date:** [May 14, 2010, 2:00pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/11 "2010-05-14T14:00:48Z")
[original email was from John Greenan - [john.greenan@alignment-systems.com](mailto:john.greenan@alignment-systems.com) ] Excellent point Jim. One question - but has anyone in the CEP world got any points to make in favour of CEP systems as market surveillance? I’ve not heard much from the firms in that space. Maybe because the existing surveillance systems failed? > Not to all the above. > > But when I looked at that stock chart from May 6th going into the next > trading day - I thought where have I seen such a pattern before? Then I > thought - yes - non-linear dynamics. Things are so complex and inter- > related and all driven off of the same set of spurious signals with > multiple levels of feedback loops that we have have created this > incredibly system that is now exhibiting chaotic behaviors. Algos are > just a part - HF, multiple venues. It is just an incredibly complex > system. Maybe the SEC should surveil the system with software designed > to track and monitor earthquakes and heart attacks. > > And John - nice to hear you chime in - The Romans have lead poisoning to > blame for their decline - what is our equivalent? High fructose corn > syrup and Facebook? 🙂 > > My two cents (before discounting) we have some good thoughts and > observations on this thread and we do need to follow larger industry > trends and begin to analyse the masses of trading data to understand the > overall system behavior. > > Someone said it best forget where I heard it within the last two days, > maybe the SEC should lay off a few dozen lawyers and hire some tech > savvy people that understand complex system behavior. > > Heck with the calvary - call in the Santa Fe Institute! > > > Did algorithms lie about the market value of assets on the balance > > sheets of big banks? > > > > Did algorithms create the “too big to fail doctrine” and saddle future > > generations of children, some not yet born, with debts they didn’t > > incur and can never repay? > > > > Did algorithms create fiat currencies or lever equity 100 fold? > > > > Did algorithms mandate the payout of 50% of bank revenues to people > > other than shareholders? > > > > Did algorithms create the absurd, childish notion of a “national > > market system” or approve bank holding company applications of Goldman > > Sachs and Morgan Stanley in two days, over a weekend? > > > > Did algorithms assign Aaa/AAA credit ratings to third derivatives of > > pools of stated-income loans? > > > > Did algorithms put the the full faith and credit of the people of the > > United States behind the liabilities of GE Capital, American Express, > > and a host of other fascist companies that have been allowed by > > corrupt politicians to make one-way bets on the backs of present and > > futures generations of taxpayers? > > > > Might it just be that all of us who love our families and children, > > who understand the economic importance of competition and free > > markets, have better objects of concern at the present moment than > > algorithmic trading? > > > > Sheesh. > > > > I guess the Romans proved that bread and circuses really do work. > > > > > Considering the sheer volume of trades/orders generated by algos a > > > tool to monitor what is going on will be necessary. It can’t be done > > > by humans or who would bare the cost? I believe its the lack of > > > tools and the evolution of this type of trading that is way ahead of > > > the ability to monitor what these algos are doing. They are > > > unpoliced and probably will be for sometime. > > > > > > > Indeed, in my point of view > > > > > > > > The algorithms must not be controlled by algorithms, but the > > > > market monitoring or risk management that should be a completely > > > > independent infrastructure. > > > > > > > > Regards > > > > > > > > Jaime Romanini > > > > > > > > > It’s an interesting, and reasonable, question, but it doesn’t > > > > > make sense to position this as a problem with the experience > > > > > level of the users of the algorithms - it seems to me that this > > > > > is more a question of needing better risk management > > > > > capabilities on the side of the systems that are interacting > > > > > with the algorithms. > > > > > > > > > > > May 7, 2010 Katherine Heires @ [securitiesindustry.com](http://securitiesindustry.com) > > > > > > > > > > > > While trading algorithms are regularly praised for their speed > > > > > > and efficiency, a growing contingent of professional traders, > > > > > > financial engineers, consultants and academics say they are > > > > > > being misused, or are faulty from the start. Their answer? > > > > > > Establish best practices and increase training to keep > > > > > > algorithms from doing more damage than good. > > > > > > > > > > > > The impetus for the movement is the fear that the availability > > > > > > of sophisticated trading strategies to a wider audience with > > > > > > varying levels of expertise could cause havoc in the markets– > > > > > > and already has, in the view of many.(…) ---
**Author:** ![kaymahes\_x](https://avatars.discourse-cdn.com/v4/letter/k/c2a13f/32.png) [@kaymahes\_x](https://forum.fixtrading.org/u/kaymahes_x)\ **Post date:** [June 11, 2010, 2:27pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/12 "2010-06-11T14:27:08Z")
Algos Gone Wild on the OSE By Steve on Jun 10, 2010 On June 1st during the first few minutes of the morning session at the Osaka Securities Exchange (OSE) a prop desk at Deutsche Securities sent several orders to sell the Nikkei 225 and the mini Nikkei 225 futures. These orders were “repeatedly sent to the exchange” by “an internal trading systems error” according to a press release issued by David Hyatt President & CEO of the firm. The statement continued “a limited number of orders [were] repeatedly sent”. However, according to a Bloomberg news release there were 980,000 orders sent. Read full article at [http://www.asiaetrading.com/algos-gone-wild-on-the-ose/](http://www.asiaetrading.com/algos-gone-wild-on-the-ose/) ---
**Author:** ![tayloedraughon\_x](https://avatars.discourse-cdn.com/v4/letter/t/d6d6ee/32.png) [@tayloedraughon\_x](https://forum.fixtrading.org/u/tayloedraughon_x)\ **Post date:** [June 11, 2010, 7:38pm UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/13 "2010-06-11T19:38:18Z")
Any trading system could replay messages. Most APIs, and FIX in particular, could replay orders into the sell side. This is a reminder that all FIX developers and integrations should pay attention to the Resend, Poss Dupe and Poss Resend flags in your trading systems. When you are asked (at a technical or business level) to replay messages make sure you properly know what needs to happen. Orders become stale over time. Perhaps if you are asked to retransmit an order you should apply business logic to skip stale orders. This is a great discussion to have with your eTrading management team. Keep in mind – Pre trade risk systems are often upstream of the re-transmission logic. When FIX asks for a retransmission it happens within the FIX engine. Many trading systems don’t re-check pre-trade controls before a re-transmission. On the sell side – when receiving an order with Poss Dupe or Poss Resend, you cannot assume that you haven’t received the order. Remember that those flags should be checked and appropriate action taken. By doing these two steps your algo system will be fortified against a repeat of this problem. Thanks Tayloe > Algos Gone Wild on the OSE > By Steve on Jun 10, 2010 > > On June 1st during the first few minutes of the morning session at the Osaka Securities Exchange (OSE) a prop desk at Deutsche Securities sent several orders to sell the Nikkei 225 and the mini Nikkei 225 futures. These orders were “repeatedly sent to the exchange” by “an internal trading systems error” according to a press release issued by David Hyatt President & CEO of the firm. The statement continued “a limited number of orders [were] repeatedly sent”. However, according to a Bloomberg news release there were 980,000 orders sent. > > Read full article at > > [http://www.asiaetrading.com/algos-gone-wild-on-the-ose/](http://www.asiaetrading.com/algos-gone-wild-on-the-ose/) ---
**Author:** ![kaymahes\_x](https://avatars.discourse-cdn.com/v4/letter/k/c2a13f/32.png) [@kaymahes\_x](https://forum.fixtrading.org/u/kaymahes_x)\ **Post date:** [June 12, 2010, 3:46am UTC](https://forum.fixtrading.org/t/time-to-rein-in-algorithms-calls-for-algo-certification-growing-louder/1605/14 "2010-06-12T03:46:39Z")
Last weekend was a major software release at BuySide MAX. Monday Market open started with live tests using ZVZZT on Acct 0000000000 and “all hands on deck” with no incidents. On Thursday morning, within a few minutes of trading, alarm bells started going off in the MAX organization’s CEO’s office because of the huge loss in milliseconds of trading when large buying of GE to cover short positions pushed up GE prices and selling large volume of GM pushed down GM prices. Now a senior management committee in incident review meeting with trader, developer, tester, FIX analyst etc. [Snip of chat Transcript from Tuesday evening USA EST electronic chat session between BuySide trader=MAX & SellSide Broker=NATX] MAX (4:08 PM): Hi GM NATX (4:09 PM): Hi GE At your order sir. [End Tuesday snip] [Snip Thursday morning USA EST] MAX (9:04 AM): Hi GE NATX (9:04 AM): Hi GM At your order sir. [End snip] Dev : See the chat texts have GM and GE Ticker symbols and FIX keyword Order. This gives OneTriggersOther FIX Order for Buy GM by using money thats received by Selling / short selling / naked short selling GE. Trader: B\*\*\*\*\*IT, GM means “Good morning” and GE means “Good Evening”. You know where Broker NATX is located ? In Asia. Look at the timestamp 4 PM, its evening for me and morning for him. Dev : But how come you are trading a US security using a broker in Asia? Trader : Its their operations which are run from offshore, its a NY-US brokerage firm. I did not want to Trade in GM or GE at the close of market. Dev : OK. Sorry. Trader: B\*\*\*\*\*IT FIX keyword Order, I wish NATX had said “at your service”. Dev : Then InterpretChatSignalsAsDayTradingOrders.java would have called a POJO Web service to trigger portfolio(s) rebalancing using these ticker symbols. CIO to Dev: You wrote this “Interpret Chat Signals as Day Trading Orders” routine ? Dev : Acually many MultiThreaded overloaded and overridden methods in InterpretChatSignalsAsDayTradingOrders.java non abstract class. CIO : May I ask why you wrote it? Dev : Managers kept asking me to appy Artificial intelligence in trading systems functionality / Algorithimic trading, a senior manager commented “To apply artificial intelligence, natural intelligence is needed, Haa Haaa”. My design was “If chat text has FIX Keywords and / or values, it indicates the chatters are interested in those FIX / trading actions, use information from dynamic market data feed & RIXML / New feeds, do Complex-Event-processing which would lead to a partial rebalance that can be triggered on available portfolio(s) and or position closing can be done and present the CEP results of analysis (of data from feeds into the REBAL routines) to the Trader(s) so (t)(s)he(y) can make an “intelligent” choice of trading using that result set or manually adjust the New Order List quantities / prices / order type(s) / TIFs etc to own preferences or discard the list of generated orders altogether.Anyway they have manual order entry screen and XML data upload.” CIO : How does your program decide Order quantity, Market or Limit order, Limit price etc.? Dev : As per FIX.4.2 spec, for a New Order Single ^35=D^, OrderQty 38 is a Req’d=N ie its an optional field. CIO : So ? Dev : The program looked up its AI dictionary of action to be taken when a piece of information is absent and implemented the solution to defer assigning a concrete value to the missing field to a later component using ChainOfResponsebility GOF design pattern. CIO : Keep going. Dev : Absence of Quantity was interpretted as Buy or sell / shortSell as many shares as available in the market by the front end trading engine FANTOMS FIX Automated Neural Trading & Order Management System. Its runs a background Thread and uses method generateBuySell to keep creating the actual orders using small quantities of MaxFloor and keeps scrapping the market. It also listens to IOIs from Market and triggers rebalances to generate orders. Broker is setup to not send these auto orders into the market as IOIs. CIO : May we ask what your program uses to decide OrderType 40? Dev : I analyzed the FIX / electronic trading logs for the last many years since we started using electronic trading and found that ``` 1. More than 95% of the orders were market orders which came from automated rebalancing portfolio management engines which did Cost basis analysis. 2. The remaining less than 5% of Limit orders were manually created by traders without using an electronic rebalance. Since I was only trigerring an partial rebalance using an existing tested module, I let the rebalancing module assign the order type. Order Type market did not require Price tag and one lesser tag in the FIX message improved performance. Also market orders trade faster than limit orders, so I thought traders would like it better. ``` CIO Nods his head in sign of hopelessness of situation. Trader: {Keeps fuming} This is why I say these dangerous tools should NOT be made available to all people. CIO : And what may I ask is a dangerous tool ? A weapon of Mass desctruction. Haa Haaa. Trader: Complex Event processing and Algorithmic trading, Background multithreading in Java (I dont like anything behind my back), Class heirarchies etc. Did you read the recent “Time to Reign in Algorithims” news article ? and the post “Dark Liquidity Pools — Friend or Foe?” ? I do not like dark liquid pools, reminds me of blobs from crude oil spills. or “Algos Gone Wild on the OSE” ? Its a Wild Wide World ☹ www._._/ CIO : Yes I read them. I had also read an article in the past about class hierarchies dividing ancient Indian society into 4 classes and how that was bad for that society as a whole. I dont know how we the people would clean up this dark liquidity pool’s mess or tame the wild Electronic financial monsters. CCO: Now Dev, explain to me why you think your automated trading program is not in violation of this organization’s policies regarding automated software systems. CLO : To quote this organization’s policy “Human action would always be the final step before any electronic trade communication is done and details regarding all compliance overriding / human actions shall be sent electronically to legal department for acceptance.” CTO : Dev, has this rule be adhered to ? Dev : Yes, the GM and GE orders were presented to the Trader and logs prove he overrode compliance and sent those trades to FANTOMS which released orders for these securities into the markets in waves. FIX Analyst : Tag 105 Wave No String Deprecated in FIX.4.2 Are we actually using Wave trading given that its not “in the latest waves”? Thats why I keep reccomending that we upgrade to latest FIX version. CEO : Trader, is it true that you over-rode compliance and sent the orders to FIX broker ? Trader : What have I done different / new compared to what I have been doing since I joined this trading post as an apprentice trader fresh out of college with an MBA degree in finance with specialization in modern electronic trading ? Every time the portfolio management system rebalances to create buy / sell orders, there will always be some compliance failures, senior traders have advised me “the idea is to keep the ball rolling by doing trading, don’t get stuck up in some silly compliance rules. Get the big picture, look at the dollars, not the fractions of cents.” CEO : Just doing something u r told to do without fully understanding the implications of what u r doing is simply unforgivable. Period. Trader : Sorry but this is life. CEO : Messaging support has provided the FIXML sent to LEGACA (Legal Committee Approval) application and logs indicate these compliance checks overriding was approved by Legal. CCO : Its true that the Legal Committee gave approval for these orders. Chief Systems architect : All electronic orders comming out of REBAL and having compliance failures overridden by trades are automatically marked approved because the cost basis calculation would have passed before REBAL generates orders. CCO : Is this scenario “Interpret chat message contents as FIX keywords / values to generate FIX actions” in the testcases suite? Tester: Yes, we have test cases, but in last release this was untested because we testers are not given access to chat applications. CEO : Why were u not given access to chat applications when traders have access to it ? CTO : Due to organization’s “Need to have & use” policy. I believe everybody should have access to everything, but this organization’s IT policies were written much before I took office here and I am only implementing the policies. Dev : Actually all of us in technology do not have access to chat applications. I wrote the auto trader code and since I did not have access to chat applications, it was tested using stubs and skeletons. Trader : Could you not find any real human testers, u had to use skeletons ? So disgusting to imagine using a program tested by a skeleton… Dev : Tested by a human using a skeleton. Trader : Skeleton wherever, whenever doing whatever… CIO : Thats why I keep saying lets outsource / offshore and use real humans sitting somewhere else instead of skeletons here. Trader : There is beauty in symmetry but there is no beauty in cemetery. [Pause] The CEO concluded that global action is required to stop these weapons of mass destruction with various bogus names like Artificial intelligence, Expert systems, Complex event processing, Dynamic heuristics, Background MutiThreading, Class hierarchy, C, MVC, C#, Java, Phython, Jython etc. are the real culprits. All should be done to stop the spread of these dangerous tools to inexperienced users who could create a global financial catastrophe and destroy the worldly way of life as we know of it presently by destroying the financial structures of the world. CIO : Somebody had rightly said - To err is human, but to really foul up large numbers, you need computers ;-( Implemented improperly, Agile model became Fragile model. Note :- All names of people / companies, ids, places, incidents, etc bear no relationship any entity living or dead. All softwares are real and copyright owned by their respective organizations / companies.