Interesting Bofe Speech On The Impacts Of Hft

[wonderpup]

Not sure if this has been posted- some of it goes over my head but still worth a read- this quote gives the idea;

In 2003, a US trading firm became insolvent in 16 seconds when an employee inadvertently turned an algorithm on.

It took the company 47 minutes to realise it had gone bust.

The quick and the dead

http://www.bankofengland.co.uk/publications/speeches/2011/speech509.pdf

[interestrateripoff]

Not sure if this has been posted- some of it goes over my head but still worth a read- this quote gives the idea;

The quick and the dead

http://www.bankofengland.co.uk/publications/speeches/2011/speech509.pdf

That quote is just brilliant.

[fluffy666]

All this speech does is qualify in their language what I already knew some years ago as to what the problems are in this activity.

A Tobin tax would end HFT very, very quickly.

Hence the extreme lobbying against such a tax. Heaven forbid they'd have to hold stocks for over a second..

[A.steve]

I strongly suspect that HFT is a red-herring... as interesting a subject as it is, and as amusing/interesting as the article proved to be.

Here's my reasoning: I can see three distinct ways to make money by trading:

• Brokerage - placing yourself as an intermediary to the market and charging a spread.
  
• Arbitrage - re-packaging instruments and re-selling.
  
• Taking (optionally hedged) speculative positions.
  

Brokerage costs fall when technology is employed - more customers can be serviced at a lower cost. HFT in the context of brokerage simply increases supply of brokerage services and drives down prices... This is good for the customer/investor - but has little systemic consequence as, even if the cost of brokerage were zero, that wouldn't greatly affect the market.

Arbitrage is slightly different - here the crux of the matter is to establish that no risks are hidden when the instruments are re-packaged. This issue of risk estimation is a subject in its own right - but is not significantly complicated by HFT. In fact, where risk is modelled by calculus, HFT's biggest influence is to increase liquidity and reduce one of extent of one of the discrepancies of risk models from reality. HFT doesn't make the risk models work - they'd be right or wrong independent of HFT.

Large speculative positions - especially leveraged ones - are where real risks lie. Subjectively, one might think the argument that HFT can't be regulated effectively - because the regulators can't keep up - might seem to make sense... but it is nonsense... the regulator doesn't need to keep up - the regulator just needs to ensure that HFT counter-parties have sufficient assigned capital in margin accounts... then it doesn't matter if an HFT company goes belly-up in a femto-second. What's even more curious about this sort of trading is that in order to make profit from a speculative position, it must be held until prices have moved... trading faster doesn't necessarily yield more profit... Faster round-trip forecasting feedback-loops in trading models will help - but the speed of trading itself will have limited effect on both risk and the bottom line.

Pointing the finger at HFT is to ignore the real problems that are as present with HFT as other trading activities:

• Inadequate counter-party risk management.
  
• Inadequate or non-assigned capital in margin accounts.
  
• Systemic over-leverage destabilising all asset prices.
  

Perhaps some focus should be put on ********-in, ********-out risk models that suggest that risks have entirely vanished for some market participants?

To focus on "AI v AI" - I think the situation is fairly simple... automated trading is just a tool... it's not game changing. It's all about establishing the best market intelligence... the better the input to your models, the better the results. This has a drastically larger impact than the AI techniques you may employ. Curiously, this might be seen as a truth about intelligence itself... where what matters most is a person's focus and the information they've collected on a subject. I think it's no coincidence that a large proportion of the human brain is associated 'just' with vision.

Edited August 1, 2011 by A.steve

[Bloo Loo]

A Tobin tax would end HFT very, very quickly.

Hence the extreme lobbying against such a tax. Heaven forbid they'd have to hold stocks for over a second..

or they could force them to have a paper trail for each and every trade.

[fluffy666]

It would but there are realms the Tobin tax wont reach where the technology and knowledge learned from HFT can be applied and have disasterous effect on everyday humans which I will call Financial Facism. FF is already happening right now to some extent and people dont realise it.

It would be a start.. as would a closing down or ringfencing of the whole 'offshore finance industry'.

[Bloo Loo]

I strongly suspect that HFT is a red-herring... as interesting a subject as it is, and as amusing/interesting as the article proved to be.

Here's my reasoning: I can see three distinct ways to make money by trading:

• Brokerage - placing yourself as an intermediary to the market and charging a spread.
• Arbitrage - re-packaging instruments and re-selling.
• Taking (optionally hedged) speculative positions.

Brokerage costs fall when technology is employed - more customers can be serviced at a lower cost. HFT in the context of brokerage simply increases supply of brokerage services and drives down prices... This is good for the customer/investor - but has little systemic consequence as, even if the cost of brokerage were zero, that wouldn't greatly affect the market.

Arbitrage is slightly different - here the crux of the matter is to establish that no risks are hidden when the instruments are re-packaged. This issue of risk estimation is a subject in its own right - but is not significantly complicated by HFT. In fact, where risk is modelled by calculus, HFT's biggest influence is to increase liquidity and reduce one of extent of one of the discrepancies of risk models from reality. HFT doesn't make the risk models work - they'd be right or wrong independent of HFT.

Large speculative positions - especially leveraged ones - are where real risks lie. Subjectively, one might think the argument that HFT can't be regulated effectively - because the regulators can't keep up - might seem to make sense... but it is nonsense... the regulator doesn't need to keep up - the regulator just needs to ensure that HFT counter-parties have sufficient assigned capital in margin accounts... then it doesn't matter if an HFT company goes belly-up in a femto-second. What's even more curious about this sort of trading is that in order to make profit from a speculative position, it must be held until prices have moved... trading faster doesn't necessarily yield more profit... Faster round-trip forecasting feedback-loops in trading models will help - but the speed of trading itself will have limited effect on both risk and the bottom line.

Pointing the finger at HFT is to ignore the real problems that are as present with HFT as other trading activities:

• Inadequate counter-party risk management.
• Inadequate or non-assigned capital in margin accounts.
• Systemic over-leverage destabilising all asset prices.

Perhaps some focus should be put on ********-in, ********-out risk models that suggest that risks have entirely vanished for some market participants?

regulation seems to be the problem....to get a better price, algos "stuff" the market with quotes and requests, none of which they intend to fulfil. in a second these, quotes disappear, but of course, the "market thinks there is demand and a price found accordingly...except, it isnt and a bankd of these guys can buy and sell between each other and make a mint...

Totally illegal, of course.

but no-one does a thing about it.

[Timak]

But I thought financial markets were for raising capital for new business ventures and allowing people to invest in the long term?

HFT is damaging- for example if I say to a broker "buy me 1000 Tesco shares at a maximum price of £3 each" then in the past he might have secured all of these for an average of 290p but by probing my maximum bid price with HFT it guarantees I end up paying 300p each. I lose out by £100 because a computer has found out the highest I'm prepared to pay and ensured I pay it. There were willing human sellers at 290p but a computer has sat between us and creamed off the profit without either of us asking it to get involved.

[goldbug9999]

But I thought financial markets were for raising capital for new business ventures and allowing people to invest in the long term?

HFT is damaging- for example if I say to a broker "buy me 1000 Tesco shares at a maximum price of £3 each" then in the past he might have secured all of these for an average of 290p but by probing my maximum bid price with HFT it guarantees I end up paying 300p each. I lose out by £100 because a computer has found out the highest I'm prepared to pay and ensured I pay it. There were willing human sellers at 290p but a computer has sat between us and creamed off the profit without either of us asking it to get involved.

Personally think that trades should be settled like the board game Risk - offers and bids should have to be placed 24 hours in advance and are settled once a day

[interestrateripoff]

Personally think that trades should be settled like the board game Risk - offers and bids should have to be placed 24 hours in advance and are settled once a day

All trades should involve 100% cash and everything should be settled immediately, you will then discover the true price.

[A.steve]

regulation seems to be the problem....to get a better price, algos "stuff" the market with quotes and requests, none of which they intend to fulfil. in a second these, quotes disappear, but of course, the "market thinks there is demand and a price found accordingly...except, it isnt and a bankd of these guys can buy and sell between each other and make a mint...

Totally illegal, of course.

but no-one does a thing about it.

Yes - that sort of behaviour does throw a spanner in the works... My only argument is that such practices are just as bent if you trade every hour; every minute; every second; every millisecond or every microsecond. HFT is orthogonal to market abuse.

What you're describing is a problem with the way in which exchanges are managed. Essentially, exchanges should make bids (and offers) binding - at least for a significant period - say an hour - and require the commitment of capital when the bid is placed. If they were to do this, prices would be a lot more stable. Of course, there remains potential for abuse where the spread of bids represents considerable insider information - but that's, again, orthogonal.

We seem to agree that the real problems are existing dishonest behaviour - not HFT itself, per se.

Edited August 1, 2011 by A.steve

[Bloo Loo]

Like everything in life, its not just one thing, but a combination of factors.

When I say AI, I mean a number of algorithms which are seemingly intelligent but not something that could pass a Turing test for example, but they have the fundamental rules programmed into them that a trader would use and would describe or classify as experience.

You can already predict the market moves just by quantify sentiment on twitter for example http://www.wired.com...r-crystal-ball/, but twitter is just one source of data, and its a variable factor in that sometimes Twitter will be wrong becuase the twitter users are not up to date with developments or announcements that can also affect stock prices. Quantify the knowledge of twitter users ie knowing how much a tweeter knows is hard but not impossible if you can measure some of their activities.

Lets just say hypothetically general MSM is putting out positive sentiment for Company A. People will generally get this news (data) and some will tweet their feelings (more data). This can create a positive snowball effect as more people read this positive sentiment and more people decide based on their own circumstances to load up on Company A's stocks, pushing the price up. Now lets just Company A get hit with a problem, it could come from internally, ie a new qtrly managment report shows a problem, they are currently bound to report things at set times, or lets say Company A gets hit with a fine from a Govt/Financial body of sorts. This can affect the company in a negative way however how quickly do those twitter users get to hear about that news? It depends on how much the MSM cover it ie the more data feeds expressing a negative feedback of Company A will increase the likely hood of a drop in share price, its back to that snowball effect, so depending on how hooked up you are to not just the normal financial feeds but also everyday feeds in teh rest of the world also determines how much the stock price will rise or fall.

So in some instances it pays to listen to twitter but in other instances it pays to listen to the financial data feeds to react quickly and then gain a short position before the tweeters catch on their sentiment changes.

The bottom line is the more data you can process and quantify the better. Even words from MSM sources can be quantified to some degree of accuracy within a range, becuase if general sentiment across society is positive, a negative for company A will not be as damaging as say a society with negative sentiment. For example, take Gabriell Gifford she got shot yet that pushed up Gun sales.

http://www.guardian....rizona-shooting The US are pro gun but if that happened here in say the UK Gun sales would fall.

The effects are variable and no human trader can quantify this as fast as a computer program.

Examples are all around us, be it houseprices, popularity of TV programs or Films. Film Critics are market manipulators for example, Restuarant Reviewers are market manipulators for example. Its all around us, even we can be viewed and quantified as market manipulators and the MSM are one big market manipulators, so the ability to measure this data is the key to success.

As I say numerous times, Chaos Theory is just the unmeasured. The internet just makes its easier to monitor chaos imo.

people cant make a million tweets a second...

[Harry Sacks]

This stuff isn't actually real is it?

[A.steve]

When I say AI, I mean a number of algorithms which are seemingly intelligent but not something that could pass a Turing test for example, but they have the fundamental rules programmed into them that a trader would use and would describe or classify as experience.

It's probably best I don't bang on about AI too much - but I have to mention that I'm not aware of any AI tech that passes the Turing test - and, in any case, that test tells us much more about human misconception than it does about intelligence - artificial or otherwise. Rule based systems are usually referred to as "Classical AI".

The bottom line is the more data you can process and quantify the better. Even words from MSM sources can be quantified to some degree of accuracy within a range, becuase if general sentiment across society is positive, a negative for company A will not be as damaging as say a society with negative sentiment.

This might seem overly pedantic, but I draw a big distinction between processing more data and processing data more effectively. In my opinion, the latter almost always trumps the former. Quality and quantity is the ideal - but it's a mistake to sacrifice (or shift focus away from) quality for quantity. The Twitter data is a great example... it's saying something - but the interpretation of its message is rather difficult. The biggest problem I see is that, in all trading circumstances, there's a winner and a loser (to some extent) - prices go up, those with long positions benefit; prices go down those with short positions benefit. If you don't know if you're recording the sentiment of people with long or short biases - you're recording white noise - you may as well fish out Mystic Meg - or read some entrails.

The effects are variable and no human trader can quantify this as fast as a computer program.

I argue that the human trader is quantifying - they're (mis)using the computer program - it's just a tool. The program doesn't care if it wins or loses - it's all bits - there's no "I'm-Alive-Johnny-5" computer program. It's all GIGO.

Examples are all around us, be it houseprices, popularity of TV programs or Films. Film Critics are market manipulators for example, Restuarant Reviewers are market manipulators for example. Its all around us, even we can be viewed and quantified as market manipulators and the MSM are one big market manipulators, so the ability to measure this data is the key to success.

As I say numerous times, Chaos Theory is just the unmeasured. The internet just makes its easier to monitor chaos imo.

Measurement and analysis - one without the other is worthless. One needs to decide what to monitor and how to interpret the results... and that can't be automated.... though there's plenty of advantage in having the right tools to hand - and using them sensibly.

people cant make a million tweets a second...

Sounds like a challenge to me.

Edited August 1, 2011 by A.steve

[A.steve]

This stuff isn't actually real is it?

That's a curious statement. In what sense do you mean it?

[A.steve]

What makes you think a human isnt defined by its chemical make up? Chemicals are rules defined by nature which depending on how they are interlinked with each other will spawn different responses (data). Event diet can alter responses, drugs like alcohol can spawn different behaviours yet we know if you add certain chemicals to a body you can alter the output.

With regard to the Classical AI, its semantics, I'm more interested in the outcome not so much in how something is labelled, they all boil down to the same thing eventually, I have the same arguements with some programmers about the definition of an application server or sql server with some types of technology for example.

I'm amused. Maybe it's not obvious - but, like the majority of sane people, I do believe that humans are constrained by their material composition. I don't think a human is defined that way, however... if I did, I'd consider someone who makes mango smoothies comparable with a serial killer - as both commit violence on biological matter. The distinction, for me, is simple - people are special because I'm one - and I don't care if that's hypocrisy from a mango's perspective. I don't assume people to be only as important as their intelligent output - if I did, many I've met would compare unfavourably with cow-pats. I have no practical difficulty distinguishing man-made technology from people - I hold people responsible for their actions (which defines my ethics and morality) - but I don't hold machines or programs accountable - because that makes no sense from a human perspective.

If a machine "is trading" - and I'm "it's counter party" - I'm not trading with the machine - I'm trading with the person (or people) who the machine represents. Similarly, if I were to smash open a Coke machine and steal a can, I'd be stealing from the people who own business - not from a robot - the robot has no standing in law.

As for your "its semantics" - I can't help myself but respond. I have to admit, I crack-up every time someone says a phrase like "it's only semantics" - which happens scarily often. Almost universally, the phrase is used by someone who doesn't know what "semantics" actually means - and, if you quiz them, you get a definition for syntax not semantics. Semantics is the meaning of syntax. It makes sense to dismiss something as "only syntax" - should I draw a distinction between brown and beige - or contrast damp with moist... but to dismiss something as "only semantics" is beyond ridiculous. An example of "only semantics" might be if a king told a warrior to throw his spear into the air - and he impaled the prince... throwing his spear into the heir. Clarity in communication brings its own rewards.

In order to quantify something you need to analyse it otherwise how can you quantify it? And when I say quantify I mean measuring the data and weighing up teh significance of the data.

Quantification, to my mind is about establishing a measure - a number (possibly including margins for error) subject to a unit. Analysis includes comparison between quantities - and consideration of the relevance of the units to the problem at hand. Quantification does not demand any consideration of relevance. I can accurately quantify the number of cows in my local field over time - but it requires analysis to decide if these quantified facts justify buying IBM's shares. Quantification doesn't involve establishing the significance or relevance of the data - that's analysis.

... a system that can bear this in mind will always outperform a human trader becuase a human trader has slow data feeds via the slow interfaces we have namely our senses and the ability to manipulate interfaces like keyboards.

The system will never out-perform a human trader - because the human trader is not competing with the system - he's competing with the person who's using the system. Conceptually, it's always two humans who are duelling - and the system is merely a weapon. It might be effective - it might not. It is a mistake to assume that an automated system will always out-manoeuvre the unarmed man - because automation takes time - and the best strategies for the unarmed man include shifting the context of the battle faster than the automated system can be amended - or identifying weaknesses in the automated system - and automating a response to exploit that flaw (such as what Bletchley did to the Enigma Machine.) Tools are exceptionally important - but it is important to recognise that you're always competing with another person... whatever toys they've brought.

Another example, if you dont talk to a baby how can it learn to speak?

A baby(the brain) is designed to have data feeds via the senses and associate different patterns to other patterns, in this case we might show it a picture of a cat and say the word cat. The brain associates the sound pattern (data) with the visual pattern(data).

Do you see where I am coming from?

Possibly not. "A baby is designed..." - I never had you down as having that religious conviction.

(BTW - I think you've diverged somewhat from the original subject...)

[A.steve]

I've tried to explain my pov but language isnt my speciallity so maybe all I can do is just demonstrate it. Watch this space.

I might have grasped what you meant - I aimed to go further by highlighting some distinctions I think to be important that you didn't seem to be making. I realised that I was being somewhat pedantic, but hoped you'd see the humour rather than assume I was making a personal gybe. I guess I was also brought my personal pet-peeve to the table - i.e. that I hate it when officials shirk their responsibilities and blame their tools... when it's their responsibility to make sure they are using the tools correctly.

Edited August 1, 2011 by A.steve

[Harry Sacks]

That's a curious statement. In what sense do you mean it?

Well, how can an investment instrument be held for less than a minute and a profit taken allowing for the spread?

[A.steve]

Well, how can an investment instrument be held for less than a minute and a profit taken allowing for the spread?

That's something I identified as a problem with the concept of HFT when I first heard about it, as well.

It definitely can be done... effectively because HFT is one mechanism by which a spread can be established (and widened). An exchange fills the purpose of matching buyers with sellers and skimming a narrow (almost risk-free) spread from that.The exchange will pay the sellers as little as they can get away with, while charging the buyers as much as they can get away with. Their big decision is to decide where their prices will be to maximise the spread when weighted by trade volumes... then to match buyers and sellers accordingly. The exchange will happily buy and sell at any price as long as it sells for more than it buys (and vice-versa) - hence prices will be volatile - but the Exchange doesn't want the risk of 'holding stock' - they don't have capital allocated to do that... so they need to operate transactionally - effectively buying and selling simultaneously to eliminate risk of loss.

HFT can exploit this where an HFT trader understands how buyers are matched with sellers... and has a decent model for the distribution of bids. The HFT trader - if they can predict this distribution of bids and offers, from the information available to them from the exchange, they can effectively front-run real customers. Say Joe-Bloggs a thousand miles from the exchange wants to buy and will pay up to $100.2 per share - and the prices are between $99.8 and $100.2... with a spread of $0.1 at the exchange. The HFT trader can anticipate that Joe-Bloggs will still be offering $100.2 per share in the next trade-matching, so he can bid up-to $100 in this round - hold the asset until the next round - then sell, but only at $100.2 - skimming a small but low-risk profit.

The interesting question is this... Why do exchanges allow this... as it damages their ability to profit from the spreads to which they used to have exclusive access? The answers are several-fold. One answer is that the exchanges can charge lucrative up-front fees for HFT access - and they can charge further fees for treating HFT traders' trades preferentially - and they can charge again for high-frequency market intelligence data. All of these allow them to hedge their risk that over the next year the market will be less lucrative for them than they should otherwise expect. Another benefit to the exchange is that an HFT trader does put capital at risk - and this capital provides liquidity... so their valued investment clients will be able to place larger trades with them at the spot-price... because HFT traders will soak up any bursty demand... all without either the exchange taking risks with its own capital - or leaking information about large orders before they complete... netting large fees for the exchange in the process.

Essentially, I believe HFT exists, but I don't think it increases systemic risk - and I don't think it's as sexy as its reputation. HFT traders exist because, on balance, the exchanges are willing to suffer narrower profit margins from spreads in exchange for added liquidity to support big clients who want to place big orders. If the big clients wised-up they'd avoid placing big orders... and make smaller trades at random times to get the best prices... but that would require effort... and who can be assed if it's someone else's money anyway?

[erranta]

I strongly suspect that HFT is a red-herring... as interesting a subject as it is, and as amusing/interesting as the article proved to be.

Here's my reasoning: I can see three distinct ways to make money by trading:

• Brokerage - placing yourself as an intermediary to the market and charging a spread.
• Arbitrage - re-packaging instruments and re-selling.
• Taking (optionally hedged) speculative positions.

I read a leaked article that mentioned Goldman Sachs and a handful of 'select' banks pay top notch to get first dibs on new tech like Intel/AMD processors along with the US military long before they are released onto public markets

Edited August 1, 2011 by erranta

[wonderpup]

Isn't the real problem with HFT the degree to which the 'market' becomes a self referential closed-circuit in which decisions are determined not by the real world fundamentals but simply in reference to the behaviour of other traders ( or their algorithms)?

What these computers are not doing is buying and selling on the basis of a knowledge of the real world- they are are simply reacting to the behaviours of other players, who in turn watch them.

The 'flash crash' could be interpreted as a kind of neurosis in nervous system of the market, as it's points of reference become ever more Incestously meshed. At some point the whole process becomes useless as real world tool for price discovery- or it has a nervous breakdown.

[A.steve]

Isn't the real problem with HFT the degree to which the 'market' becomes a self referential closed-circuit in which decisions are determined not by the real world fundamentals but simply in reference to the behaviour of other traders ( or their algorithms)?

From the perspective of society, it makes sense to be worried that this might be happening... but - I'd argue - it's not only HFT that is making its decisions solely on the basis of market data... a whole bunch of non-HFT operations are doing exactly the same thing - just more slowly. The phrase 'technical analysis' suggests decision making based solely upon market data as opposed to a fundamental understanding of the assets - and this is endemic throughout markets... if it were not, we'd not see bubbles like the Tech bubble of the late 90s, or the housing bubble of the noughties.

HFT does not run counter to these trends - but it is far from the biggest contributor to the problem.

Paradoxically, given public and political sentiment, it is low frequency trading (on margin) that poses a far greater risk.

[Harry Sacks]

That's something I identified as a problem with the concept of HFT when I first heard about it, as well.

It definitely can be done... effectively because HFT is one mechanism by which a spread can be established (and widened). An exchange fills the purpose of matching buyers with sellers and skimming a narrow (almost risk-free) spread from that.The exchange will pay the sellers as little as they can get away with, while charging the buyers as much as they can get away with. Their big decision is to decide where their prices will be to maximise the spread when weighted by trade volumes... then to match buyers and sellers accordingly. The exchange will happily buy and sell at any price as long as it sells for more than it buys (and vice-versa) - hence prices will be volatile - but the Exchange doesn't want the risk of 'holding stock' - they don't have capital allocated to do that... so they need to operate transactionally - effectively buying and selling simultaneously to eliminate risk of loss.

HFT can exploit this where an HFT trader understands how buyers are matched with sellers... and has a decent model for the distribution of bids. The HFT trader - if they can predict this distribution of bids and offers, from the information available to them from the exchange, they can effectively front-run real customers. Say Joe-Bloggs a thousand miles from the exchange wants to buy and will pay up to $100.2 per share - and the prices are between $99.8 and $100.2... with a spread of $0.1 at the exchange. The HFT trader can anticipate that Joe-Bloggs will still be offering $100.2 per share in the next trade-matching, so he can bid up-to $100 in this round - hold the asset until the next round - then sell, but only at $100.2 - skimming a small but low-risk profit.

The interesting question is this... Why do exchanges allow this... as it damages their ability to profit from the spreads to which they used to have exclusive access? The answers are several-fold. One answer is that the exchanges can charge lucrative up-front fees for HFT access - and they can charge further fees for treating HFT traders' trades preferentially - and they can charge again for high-frequency market intelligence data. All of these allow them to hedge their risk that over the next year the market will be less lucrative for them than they should otherwise expect. Another benefit to the exchange is that an HFT trader does put capital at risk - and this capital provides liquidity... so their valued investment clients will be able to place larger trades with them at the spot-price... because HFT traders will soak up any bursty demand... all without either the exchange taking risks with its own capital - or leaking information about large orders before they complete... netting large fees for the exchange in the process.

Essentially, I believe HFT exists, but I don't think it increases systemic risk - and I don't think it's as sexy as its reputation. HFT traders exist because, on balance, the exchanges are willing to suffer narrower profit margins from spreads in exchange for added liquidity to support big clients who want to place big orders. If the big clients wised-up they'd avoid placing big orders... and make smaller trades at random times to get the best prices... but that would require effort... and who can be assed if it's someone else's money anyway?

Thanks.

[interestrateripoff]

In 2003, a US trading firm became insolvent in 16 seconds when an employee inadvertently turned an algorithm on.

It took the company 47 minutes to realise it had gone bust.

http://www.bankofengland.co.uk/publications/speeches/2011/speech509.pdf

What happened to this firm where all the trades rolled back?

Also I've been wondering if this "inadvertently" switched on algorithm had worked would it have been turned off? It seems a one way bet was made if this firm didn't go bust.

I also have a problem with the fact it's presented as an accident rather than something which didn't work as intended ie generate a profit.

[Harry Sacks]

What happened to this firm where all the trades rolled back?

Also I've been wondering if this "inadvertently" switched on algorithm had worked would it have been turned off? It seems a one way bet was made if this firm didn't go bust.

I also have a problem with the fact it's presented as an accident rather than something which didn't work as intended ie generate a profit.

Does it matter if it's true these days?