General Betting
Welcome to Live View – Take the tour to learn more
Start Tour
There is currently 1 person viewing this thread.
1. If this were a bone fide punter, with an account balance of $600m, and they made a mistake, then they have to pay out, as we all have had to over the years for the mistakes we've made from time to time.
2. If, however, as seems more likely this is not an external punter (after all who realistically has an account in excess of $600m?!), then this must be down to some error on Betfair's part. If so what sort of error?
Given that Betfair are not a bookie, the only way I can see they made an error is if their cross-matching bot threw a wobbly, otherwise they must have been engaged in an activity for which on the face of it they are not licensed, and/or do not declare in their T&Cs. This begs three questions in my mind:
1. Can BF be engaged in any activity other than cross-matching in the Horse IR markets?
2. If not, and therefore the cross-matcher is to blame, how can BF guarantee that the cross-matcher is working correctly, and has done so historically? How do we know that we have really been matched at best price in the past?
3. AND/OR - has this little £600m at risk problem exposed something deeper and perhaps more sinister going on?
2. If, however, as seems more likely this is not an external punter (after all who realistically has an account in excess of $600m?!), then this must be down to some error on Betfair's part. If so what sort of error?
Given that Betfair are not a bookie, the only way I can see they made an error is if their cross-matching bot threw a wobbly, otherwise they must have been engaged in an activity for which on the face of it they are not licensed, and/or do not declare in their T&Cs. This begs three questions in my mind:
1. Can BF be engaged in any activity other than cross-matching in the Horse IR markets?
2. If not, and therefore the cross-matcher is to blame, how can BF guarantee that the cross-matcher is working correctly, and has done so historically? How do we know that we have really been matched at best price in the past?
3. AND/OR - has this little £600m at risk problem exposed something deeper and perhaps more sinister going on?
Show More
did we not have an integer overflow problem when the bat IDs reached a significant figure, I rememember all the IT guys laughing their c*cks off at error at the time.
Dec 28, 2011 -- 5:55PM, STEPTOES YARD wrote:
amazed more of that 29s didnt go
you could look at it the other way within 5mins of a wrong price appearing on run of mill irish race (guessing there was nothing special about it) almost £1m of real money was looking to grab it.
thanks for the Link brendan, I rem looking forward to seeing my peugeot going round the clock, but sadly I missed the magical moment
this what i dont understand it would surely have been better for betfair to say exceeded account balance then exceeded exposure limit, this implies the layer could cover the bet.
Betfair treated me abysmally a few years back, so this is nothing new.
I put a jackpot on at Cheltenham and premed some selections at 10p unit stake. the bet came to around £50 in total. Placed the bet and the bet confirmed. Logged out and settled down to watch the races on the telly.
Well, I managed to find all six winners and calculated that i'd scooped £5K. I told the missus that was her kitchen paid for.
On logging back into Betfair, there was no sign of the funds. On checking the statement, I noticed that my stake had been refunded some minutes after the bet was struck. It turns out that the minimum unit atake was 50p. I argued that they should have had software in place that prevented a unit stake of less than 50p being entered and morally they were obliged to pay out as the bet was accepted.
of course, I couldn't win and they ended up giving me £25 as a good will gesture - woopy ******* do.
I put a jackpot on at Cheltenham and premed some selections at 10p unit stake. the bet came to around £50 in total. Placed the bet and the bet confirmed. Logged out and settled down to watch the races on the telly.
Well, I managed to find all six winners and calculated that i'd scooped £5K. I told the missus that was her kitchen paid for.
On logging back into Betfair, there was no sign of the funds. On checking the statement, I noticed that my stake had been refunded some minutes after the bet was struck. It turns out that the minimum unit atake was 50p. I argued that they should have had software in place that prevented a unit stake of less than 50p being entered and morally they were obliged to pay out as the bet was accepted.
of course, I couldn't win and they ended up giving me £25 as a good will gesture - woopy ******* do.
another link with example of underflow/overflow in java with customer balances
http://weblogs.asp.net/dvravikanth/archive/2008/03/10/integer-overflow-amp-underflow-revisited.aspx
http://weblogs.asp.net/dvravikanth/archive/2008/03/10/integer-overflow-amp-underflow-revisited.aspx
Actually if you pause it at 0:09 it's £60 at 30, not £5 
someone may have found a way to over ride their exposure limit and decided to cash in by backing at a rival site - winning all ways up .
obviously they would have preferred today's horse in question to lose but still a few healthy ££££ in any case.
it will be an interesting outcome anyway
obviously they would have preferred today's horse in question to lose but still a few healthy ££££ in any case.
it will be an interesting outcome anyway
this customer up to now possibly raking in £100s,1000s then bot glitch or typing error(tis the season--eat/drink/be merry) is his undoing.
betfairs excuse fails the SMELL test
betfairs excuse fails the SMELL test
Brendanuk1 - That's a great find. I suspect you and Dstyle are getting close to what's happened.
Given what we know about the general level of muppetry at BF towers I can see that someone may well have stuffed up the treatment of a preferential customer.
Given what we know about the general level of muppetry at BF towers I can see that someone may well have stuffed up the treatment of a preferential customer.
great video that - love the way the guy is trading the fav and then must be thinking "what the f*ck" and quickly moves over and takes some 29, pauses, then must be thinking what the f*ck again, then goes for it massively just as the race ends
anyone know how much was traded at the purple place on this race
must have been some canny numbers being layed there
must have been some canny numbers being layed there
Richard LL 29 Dec 11 00:55
did we not have an integer overflow problem when the bat IDs reached a significant figure, I rememember all the IT guys laughing their c*cks off at error at the time.
that's exactly what happened.
did we not have an integer overflow problem when the bat IDs reached a significant figure, I rememember all the IT guys laughing their c*cks off at error at the time.
that's exactly what happened.
1424 could easily be cross matching.
Can anyone explain this?
I understand that the amount matched divided by 2 plus the amount unmatched at the end of the youtube clip above is equal to 1424 less than a 32 bit integer in binary.
So where is the missing 1424? What do you mean it could easily be "cross-matching"?
Can anyone explain this?
I understand that the amount matched divided by 2 plus the amount unmatched at the end of the youtube clip above is equal to 1424 less than a 32 bit integer in binary.
So where is the missing 1424? What do you mean it could easily be "cross-matching"?
Presumably on the real betfair screen there were prices available on the other runners due to cross matching. I'd assume that nearly all trading on this race was done on the winner (though I'm not a horse person and neither do I understand much of the API screen in the video) and that every time Betfair skims an ammount it goes missing rather than getting logged as bet.
It does seem plausible that 1424 could be that number.
It does seem plausible that 1424 could be that number.
So the API screen on the youtube clip does not totally reflect the real market, and the difference in both screens could be 1424??
When someone puts up a lay of 29, combined with other lays in the race, a back appears on the other runners to the inverse value - approximately thanks to betfair cross matching software.......... this would explain why the favourite gets shorter in price because the other runners hit 1000s before the favourite did and it tends towards the inverse of 29.
When anyone takes one of those backs, the money is distributed amongst the other runners. I don't know how this shows up on the market info and hopefully one of the regulars who looks into those things will know, but it seems plausible that it isn't assigned in the way we'd expect it to be - on top of that, there's often a little left over (that betfair keep) and who knows where that's assigned.
When anyone takes one of those backs, the money is distributed amongst the other runners. I don't know how this shows up on the market info and hopefully one of the regulars who looks into those things will know, but it seems plausible that it isn't assigned in the way we'd expect it to be - on top of that, there's often a little left over (that betfair keep) and who knows where that's assigned.
So your first paragraph explains why Mourad hovered around 1.03 and 1.04 for so long (1/33 and 1/25 respectively) or the inverse of 28/1.
But I don't understand the Mourad money being distrubted amongst other runners. If I (or my software if I owned any) take the 1.03 or 1.04, what happens then?
But I don't understand the Mourad money being distrubted amongst other runners. If I (or my software if I owned any) take the 1.03 or 1.04, what happens then?
If you lay the 1.04 with Betfair's skimmer (referred to inaccurately by the firm as 'best price execution') then you'll actually bet the appropriate amounts on all the other horses to make sure you've got what looks like a lay of 1.04
After you've got the price you've been promised, if there are pennies left over, they go in Betfair's coffers.
In a normal race, there would be lots of genuine people trying to bet 1.04 say, but in this instance it would almost all be as a result of the huge sum sitting at 29 creating the 1.04 (seems a bit bigger when I look at the video, but that's because you have to bet the second fave at 7/1 and so on)
After you've got the price you've been promised, if there are pennies left over, they go in Betfair's coffers.
In a normal race, there would be lots of genuine people trying to bet 1.04 say, but in this instance it would almost all be as a result of the huge sum sitting at 29 creating the 1.04 (seems a bit bigger when I look at the video, but that's because you have to bet the second fave at 7/1 and so on)
As a matter of interest does any one have the figures matched IR for Mourad?
On both the video and the screenshot the 29 lay goes after suspension after being there a good while during the race. Anyone hazard a guess as why its removed after suspension but left up as the horse clearly is going to win and when it actually crosses the line?
That's what happens in the football markets with the main market making lumps. About 1 second before all suspensions the lumps disappear, as if they're directly connected to the suspend operation. I commented on this when it first started happening quite a few months ago, as it now appears that the main lumps are BF's, either directly or sub contracted.
The 29.0 lay money disappears in an identical manner.
The 29.0 lay money disappears in an identical manner.
trev, how can you relate one facet of an activity to company liability when disregarding the obvious large price on offer for a winning favourite(2nd). Even a 29 shot pre- race start would of shortened to under 3 a furlong out?
Isn't it something to do with "keep bets" ie the 29 was put into the market in-play and not as a keep bet therefore gets cleared on suspension?
"Even a 29 shot pre- race start would of shortened to under 3 a furlong out? "
This isn't always the case with large sums - usually when the big bet is a huge ask at a huge price rather than a lay - because often there just isn't the money to match it.
Happens in golf quite a bit where someone will ask for £500 at 500/1 or something and it should be 20,000/1, it takes forever to get matched.
This isn't always the case with large sums - usually when the big bet is a huge ask at a huge price rather than a lay - because often there just isn't the money to match it.
Happens in golf quite a bit where someone will ask for £500 at 500/1 or something and it should be 20,000/1, it takes forever to get matched.
Oh I see now, so the price stalling at 29 is due to the system busy matching up a potential 21 million pay-out relative to someones exsposure? all seems a bit odd really and open too topical debate regarding the conditions and mechanisms at work within the betfair API.
Well not so much the system, just the lack of backers with 2 million floating around.
of course, makes sense, but points the finger at mystery allowance of huge betting exsposure in-play?
was anyone else able to over-exspose their account?
We have identified the issue and replicated it in a test environment last night. A fix was applied overnight, and is now subject to rigorous testing. A further update on this will be made today.
So put in a fix, then let customers do the rigorous testing
So put in a fix, then let customers do the rigorous testing
I know, a true Christmas miracle... Betfair (1) have a test environment and (2) are running rigorous testing!
I almost fell off my chair when I read that.
It's got to be a rogue pointer hasn't it. Either from BF comm sw or from within Oracle - and my money's on the latter. But how can they explain that to most people on here? If they say it's a ddos people will worry about their funds, so they've gone for the middle ground.... a naughty customer who exposed a vulnerability in BF which BF have now found and closed.
and to those who tracked down this error in less than 24 hours (cough, cough) - in the name of god, direct them towards the farce that is Live Video!
I almost fell off my chair when I read that.
It's got to be a rogue pointer hasn't it. Either from BF comm sw or from within Oracle - and my money's on the latter. But how can they explain that to most people on here? If they say it's a ddos people will worry about their funds, so they've gone for the middle ground.... a naughty customer who exposed a vulnerability in BF which BF have now found and closed.
and to those who tracked down this error in less than 24 hours (cough, cough) - in the name of god, direct them towards the farce that is Live Video!
BF claim now to be in full possession of all the facts .... if so they are keeping them under their hat.
they knew everything soon after as it was obvious and clear technical fault, fixed it last night, tested it rigorously today, but failed to keep their chief spokesman briefed before going on national tv to do a Q & A session
just to state the obvious briefly, we're dealing with not one but 2 oddities here.
1 - the amount that appeared, and
2 - the price
that coupled with the fact that BF claim to have suspended the relevant account to me fairly clearly suggest that this was a deliberate exploit of a hole someone found. in which case, it also seems likely that
1 - the account involved had very little money in it
2 - the actual money was being made elsewhere.
I would therefore suggest that the accounts involved in the first trades made on purple at silly prices will be the ones to look at. anyone know how much got matched over there?
if the above is correct, whoever was involved would know it was a one-shot deal. so they'd need to make as much as possible from the one coup. which makes it a bit odd they didn't go for a higher profile race.
what's also odd is, if this is basically the same problem as already occured with bet IDs, why did no one think to test this scenario when that was being fixed?
apologies if this is old ground. not had time to follow all threads on this closely.
1 - the amount that appeared, and
2 - the price
that coupled with the fact that BF claim to have suspended the relevant account to me fairly clearly suggest that this was a deliberate exploit of a hole someone found. in which case, it also seems likely that
1 - the account involved had very little money in it
2 - the actual money was being made elsewhere.
I would therefore suggest that the accounts involved in the first trades made on purple at silly prices will be the ones to look at. anyone know how much got matched over there?
if the above is correct, whoever was involved would know it was a one-shot deal. so they'd need to make as much as possible from the one coup. which makes it a bit odd they didn't go for a higher profile race.
what's also odd is, if this is basically the same problem as already occured with bet IDs, why did no one think to test this scenario when that was being fixed?
apologies if this is old ground. not had time to follow all threads on this closely.
Would you go against an odds on fav if it's a one shot deal ?
Would be surprised the customer hasnt been hung out to dry if there was anything at all malicious in it. Police etc. this hasnt happened.
They wouldn't hang him it out dry, they want to hush this as quickly as possible. It shows a flaw in their software, whether deliberately hacked or not, which they want to hush to stop people thinking that betfair may be unreliable. Unreliabiity would cause most damage to their share price.
At some point during the almost inevitable third party investigation I would like to understand a lot more about the "customer in question". My opening question: has this "customer" ever worked for Betfair?
And Brendanuk1, I'm sure it's not necessary to remind everyone that if someone were of the view something malicious has happened here and they had reasons to suspect a crime had been committed, they would of course be duty bound to inform the police.
And Brendanuk1, I'm sure it's not necessary to remind everyone that if someone were of the view something malicious has happened here and they had reasons to suspect a crime had been committed, they would of course be duty bound to inform the police.
depends when the money came on and what appeared the likely outcome at the time frames.
arguably it works either way, though - depending what you expect BF to do if the horse loses.
arguably it works either way, though - depending what you expect BF to do if the horse loses.
I read the money came on less than a minute into the race when the horse was at the back.ATR had the race on a split screen at the time.
It would be interesting to see the figures matched on the purple place for the race.Not sure if they have been posted anywhere ?
It would be interesting to see the figures matched on the purple place for the race.Not sure if they have been posted anywhere ?
. Perform a number of bets pre-race on betfair to achieve a goal that nobody notices and as such nobody comments on.
.. ?
. Utilise split screen viewing (has to be minismised though), similar colours and starting out near the back?
.. ??
. Coordinate bets with Purple?
.. ???
. Align all other horses, jockeys, owners, trainers, and courses so said horse wins.
.. ????
. Place bet inrunning to void inrunning market.
.. ?????
. Collect
.. ??????
Take a bow
.. ?
. Utilise split screen viewing (has to be minismised though), similar colours and starting out near the back?
.. ??
. Coordinate bets with Purple?
.. ???
. Align all other horses, jockeys, owners, trainers, and courses so said horse wins.
.. ????
. Place bet inrunning to void inrunning market.
.. ?????
. Collect
.. ??????
Take a bow
So the theory is that the BF lay was just to increase the price of the **** horse at the purple? The Sting is a great film, still watch it when it's repeated even after all this time :)
A couple of questions that have already been mentioned are:
What would have happened if the horse had lost?
Why not choose a race with greater liquidity/amount of punters to take the lay money on BF and lay it off at a silly price on purple for you to back?
Has anyone actually confirmed yet how much was matched at purple?
A couple of questions that have already been mentioned are:
What would have happened if the horse had lost?
Why not choose a race with greater liquidity/amount of punters to take the lay money on BF and lay it off at a silly price on purple for you to back?
Has anyone actually confirmed yet how much was matched at purple?
Please see my post re smoking gun.
If you look at nearly's post at 17:48, why were the bets matched below 29.00, when Betfair guarantee best execution, and 29.00 was clearly available at the time the bets were matched?
If you look at nearly's post at 17:48, why were the bets matched below 29.00, when Betfair guarantee best execution, and 29.00 was clearly available at the time the bets were matched?
hmmm. latest update still doesn't explain how it happened (although it does state the full value of the lay was the int32 max in pence), in fact it's just a reiteration of what's been previously said.
perhaps they don't want people to know.
"devastated" is a bit of a strong word as well.
perhaps they don't want people to know.
"devastated" is a bit of a strong word as well.
but well done shifting the blame onto bots. evil evil bots.
less than 1k in their account though, so a recreational bot?
can we assume that the missing 1424 was cross matched? wonder what proportion of that is then lost to the cross matcher?
can we assume that the missing 1424 was cross matched? wonder what proportion of that is then lost to the cross matcher?
Isn't it time that Betfair banned these bots...they have caused nothing but bother.
how does a bot "develop" a fault, exactly?
They can't ban them. They could do away with the API but programmers would just screen scrape. They could limit the information available to them though.
Because a programme can scrape the necessary information from their web page and submit bets automatically. The betfair server just receives a block of binary code and has no way of knowing whether it was submitted by a manual user and generated by the web page or was submitted by a user programme. The way to combat it is by limiting the information available to bots.
well you're wrong then.
you could only ban on suspected activity.
you could make it very difficult to operate a bot, but ultimately all communciation between your PC and betfair's servers which can be performed via a website, can be sent programatically.
you could only ban on suspected activity.
you could make it very difficult to operate a bot, but ultimately all communciation between your PC and betfair's servers which can be performed via a website, can be sent programatically.
On the question of how a bot can 'develop' a fault, I expect they changed it and the change introduced a bug.
Other causes are possible though. I was once running a bot (small stakes, just for fun) and it crashed. So I restarted it, not realising the crash had corrupted its MySQL database. The loss of a database meant it had total amnesia. It would place a bet, not know it had placed it, place it again ... until my account was empty
Fun times.
Other causes are possible though. I was once running a bot (small stakes, just for fun) and it crashed. So I restarted it, not realising the crash had corrupted its MySQL database. The loss of a database meant it had total amnesia. It would place a bet, not know it had placed it, place it again ... until my account was empty
Fun times.
STEPTOES YARD 28 Dec 11 23:55
amazed more of that 29s didnt go
Actually I'm amazed that so much was taken.
If I'd been playing the race, and seen that lay offered in-running, I would have assumed someone had spotted the horse go the wrong side of a marker or lose its weight cloth and I hadn't. Especially in a race shown on split screen.
Yes, on reflection the price offered didn't really match that scenario, but in the hurly burly of a horse race on a split screen, I would have just let the opportunity go.
For comparison, when Inzamam-Ul-Haq refused to come out after tea in the Oval Test vs England, double figure prices were offered about an England victory for a lot longer than this horse race took to run, yet weren't taken, and that was as big a gift. But, again, only with hindsight.
amazed more of that 29s didnt go
Actually I'm amazed that so much was taken.
If I'd been playing the race, and seen that lay offered in-running, I would have assumed someone had spotted the horse go the wrong side of a marker or lose its weight cloth and I hadn't. Especially in a race shown on split screen.
Yes, on reflection the price offered didn't really match that scenario, but in the hurly burly of a horse race on a split screen, I would have just let the opportunity go.
For comparison, when Inzamam-Ul-Haq refused to come out after tea in the Oval Test vs England, double figure prices were offered about an England victory for a lot longer than this horse race took to run, yet weren't taken, and that was as big a gift. But, again, only with hindsight.
exactly what i did screaming.
I could see the prices, but i've got slow pics and I wasn't playing, so didn't know the colors and i just sat there thinking 'God that poor commentator's gonna get some stick when all this is over!"
I could see the prices, but i've got slow pics and I wasn't playing, so didn't know the colors and i just sat there thinking 'God that poor commentator's gonna get some stick when all this is over!"
Are we take it then that the magic 2^31 number is what led to this mess?
Betfair have now disclosed the Lay amount.
Which just happened to be 2^31 pence.
Well done to whoever here spotted that first.
Which just happened to be 2^31 pence.
Well done to whoever here spotted that first.
Not being a techie, is that sommething important in a tech sense ?
Someone has probably already explained it for your ilk.
It is (or was) common for signed integers to be store in 32 bits.
The largest positive number was therefore 2^31
Exceeding that makes the number negative.
It is (or was) common for signed integers to be store in 32 bits.
The largest positive number was therefore 2^31
Exceeding that makes the number negative.
Maybe I should rephrase then.
Not being a techie, is this 32 bit revelation or whatever it is, be of the slightest future concern to me, as a dinosaurus " manual" punter ?
Not being a techie, is this 32 bit revelation or whatever it is, be of the slightest future concern to me, as a dinosaurus " manual" punter ?
If the bet size was 2^31+1 pence = 21,474,836.50 pounds
a 32 bit signed integer would interpret that as
-1 penny
and presumably that times 28 would be considered below the account available funds.
a 32 bit signed integer would interpret that as
-1 penny
and presumably that times 28 would be considered below the account available funds.
Thanks for your repeated attempts to try to put me more in the picture on this.
I get what you are saying literally.
But what are the consequences or future risks for me in all this ?
None ?
It's all been recognized glitch wise and has now been satisfactorily fixed ?
Or it's just the tip of the iceberg regarding other potential, unknown techie risks out there ?
I get what you are saying literally.
But what are the consequences or future risks for me in all this ?
None ?
It's all been recognized glitch wise and has now been satisfactorily fixed ?
Or it's just the tip of the iceberg regarding other potential, unknown techie risks out there ?
Fixing this appears to trivial. Once the problem is known.
But we can only speculate as to future impact.
Betfair's full disclosure is garbage.
So if it's not giving us the full or true story, how can any of us say what other risks there may be out there?
But we can only speculate as to future impact.
Betfair's full disclosure is garbage.
So if it's not giving us the full or true story, how can any of us say what other risks there may be out there?
Well that sure is a nice, comforting viewpoint to take into the new year of gambling.
But better to be aware that unaware of what technical glitches might potentially still be out there .
Just adds more reason to always gamble with these "disaster" downsides built into your risk parameters.
Don't overstake, spread your risks etc.
And perhaps don't get caught with your leg too high in the air when placing pre race wagers you want to trade out of in running.
Best summed as don't trade too doggie style eh ?
But better to be aware that unaware of what technical glitches might potentially still be out there .
Just adds more reason to always gamble with these "disaster" downsides built into your risk parameters.
Don't overstake, spread your risks etc.
And perhaps don't get caught with your leg too high in the air when placing pre race wagers you want to trade out of in running.
Best summed as don't trade too doggie style eh ?
The forumite DStyle was the one who first suggested the 32 bit integer theory. Well done to him.
However, I still have some further questions on this issue.
1. 2^31 equals 2 147 483 648. In pounds this is £21,474.364.48. In signed and unsigned bits, this number represents 0 I believe? I.e 21,474.364.47 (1 less than 2^31) is the maximum number Betfair can store in memory. However surely Betfair's software had some kind of fail-safe mechanism if it recieved a negative value (£21,474.364.49 would equal -1)*28= -0.28. How is it possible to have negative exposure? Surely, a negative value should have triggered some kind of error?
2. The punter in question had £1000 in his account. Therefore am I right in saying that if he submitted a bet anywhere between £21,474,364.47 and £21,474,872.19 (which in signed/unsiged bits would represent -3571 or -£35.71 * 28 = £999.88), the same thing would have happened. So the significane of 2^31 isn't so important, once the number was between 2^31 and 2^31+3571 is the crucial thing?
Am I correct?
However, I still have some further questions on this issue.
1. 2^31 equals 2 147 483 648. In pounds this is £21,474.364.48. In signed and unsigned bits, this number represents 0 I believe? I.e 21,474.364.47 (1 less than 2^31) is the maximum number Betfair can store in memory. However surely Betfair's software had some kind of fail-safe mechanism if it recieved a negative value (£21,474.364.49 would equal -1)*28= -0.28. How is it possible to have negative exposure? Surely, a negative value should have triggered some kind of error?
2. The punter in question had £1000 in his account. Therefore am I right in saying that if he submitted a bet anywhere between £21,474,364.47 and £21,474,872.19 (which in signed/unsiged bits would represent -3571 or -£35.71 * 28 = £999.88), the same thing would have happened. So the significane of 2^31 isn't so important, once the number was between 2^31 and 2^31+3571 is the crucial thing?
Am I correct?
3. And still nobody knows where the 29 came from. Or are we accepting that the bot just multiplied the SP by 10?
4. Also surely someone in the history of Betfair has tried to submit a bet greater than 21,474.364.47? What was so special about this person's bot that it was able to bring BF to it's knees?
4. Also surely someone in the history of Betfair has tried to submit a bet greater than 21,474.364.47? What was so special about this person's bot that it was able to bring BF to it's knees?
This is undisclosed and I think it's a poor show to state in a bulletin that they are giving us all the known facts when clearly they are not.
That said, does it really matter? What this incident proves, if anything, is that even the most tried and tested computer system will still have bugs in. Anyone who says otherwise is either lying or not au fair with the binary world.
What we can say quite categorically is that Betfair has one of the most advanced, tried, tested and robust databases anywhere on the planet.
That said, does it really matter? What this incident proves, if anything, is that even the most tried and tested computer system will still have bugs in. Anyone who says otherwise is either lying or not au fair with the binary world.
What we can say quite categorically is that Betfair has one of the most advanced, tried, tested and robust databases anywhere on the planet.
Surely, a negative value should have triggered some kind of error?
Not unless you were expecting and checking for negative values. Whatever Betfairs logic in checking it failed, probably some boolean logic in their to so it got into a right twizz when adding or multiplying numbers and making assumptions the result is positive.
Be surprised if this didnt happen before in previous 10 years, a big bet over 21m
Not unless you were expecting and checking for negative values. Whatever Betfairs logic in checking it failed, probably some boolean logic in their to so it got into a right twizz when adding or multiplying numbers and making assumptions the result is positive.
Be surprised if this didnt happen before in previous 10 years, a big bet over 21m
What has probably happened is a combination of 2 errors occurring at the same time
1.....a customers bot has multiplied a series of bets that has equaled an overflow figure
in total stake at 32 bit size data.
2.....a fatal flaw was programmed in the part of the api that checks the total amount of stake
of an account that needs matching and has allowed the bet to be matched because of incorrect bit parity testing of overflow numbers.
or to put into laymans terms , a bug existed in the betfair API bet matching bot.
all this goes to prove that if you combine the talents of 2 fools you get a result
far greater than than the total sum of the parts!
1.....a customers bot has multiplied a series of bets that has equaled an overflow figure
in total stake at 32 bit size data.
2.....a fatal flaw was programmed in the part of the api that checks the total amount of stake
of an account that needs matching and has allowed the bet to be matched because of incorrect bit parity testing of overflow numbers.
or to put into laymans terms , a bug existed in the betfair API bet matching bot.
all this goes to prove that if you combine the talents of 2 fools you get a result
far greater than than the total sum of the parts!
Brendan. Surely the people who programmed it would have put in some kind of function that rejected negative bets. Even if it was never expected that it would be needed, it would still seem obvious to put it there.
Also I agree. Surely this software oversight has happened in the past. But then how come it was never noticed (the figures involved need to be colassal) and if it was noticed, why was it never fixed?
Also I agree. Surely this software oversight has happened in the past. But then how come it was never noticed (the figures involved need to be colassal) and if it was noticed, why was it never fixed?
bf_fanatic, is the API programmed differently to the normal Betfair system that us mortals use? I would have thought if the API was designed after, most of it's code would be a matter of copy and paste, and then add in the extra functionality afterewards.
one must assume that betfairs API stake matching procedures are built for speed because of ther amount
of bets that have to be processed in real-time, code streamlining is a matter of course in all good programmers practices but it shuld never have priority over stress testing and bullet proofing of the code and this has happened with the result shown to all and sundry!
of bets that have to be processed in real-time, code streamlining is a matter of course in all good programmers practices but it shuld never have priority over stress testing and bullet proofing of the code and this has happened with the result shown to all and sundry!
what usually happens is that once it gets passed a certain stage, then certain assumptions are made about it otherwise your checking and double checking the same number over and over
Would imagine the first check validation has changed subtly and allowed a number in the it shouldnt of and the rest of software has just assumed it is correct/valid. Some recent upgrade or sumfink
Would imagine the first check validation has changed subtly and allowed a number in the it shouldnt of and the rest of software has just assumed it is correct/valid. Some recent upgrade or sumfink
I see. Makes sense I guess.
The fact that it looks like it was an error specific to the API code would probably explain why nobody has ever tried to input a figure this large either.
The fact that it looks like it was an error specific to the API code would probably explain why nobody has ever tried to input a figure this large either.
you would never expect any sane punter to place such large total stake value on any event item outcome but you still have to make your code idiot proof
can anyone remember the bug that was built in the first pentium proccesors, I think it was something to do with the MPU maths processing unit, it took a bright spark to spot it and trigger the event in that case
The API would be accepting all sorts of sh!te from everywhere would make me think its some recent code change. Thats usually the reason bug get added. Classic regression
Could still be a unique set of events that have never happened before but that would be really freaky!
Could still be a unique set of events that have never happened before but that would be really freaky!
No, no sane punter would try that. But a kid pressing random buttons might fluke it, or someone leaning on the numbers section of their keypad might do it. I'm no expert in the world of computer coding and programming but I would have a small appreciation for it, and this error would appear to me to be bread and better to any programmer.
The most frightning scenario of automated madness is the amount of nuclear warheads that can be sent to there targets by a series of atomated detection ssytems around the world but thankfully the process is intervened with a phone call to a nation leader whos has the sense to stop and say"this doesnt make sense"
this has happened a number of times in the past and thankfully a silicon logical device has
been halted by the fuzzy logic of a world leader, mind over matter!
been halted by the fuzzy logic of a world leader, mind over matter!
the automated forex software that runs everyday has the power to bring the world to its knees in the event of market freefall, one hopes humans are allowed to intervine these processes also.
machines can be very dangerous indeed.
machines can be very dangerous indeed.
Do betfair have humans monitoring large stakes players or any activity that is pushing the top of the betfair API processes envelopes? clearly not and at the holiday period even less so!
Slightly off topic but the markets in Europe and the euro are one thing that are massively affected by computer software with pre-programmed commands. Something triggers a drop in the euro's value vs the dollar. Someone's programme is designed to sell once the euro drops below a certain value, this pushes the euro's worth down further, so somebody else sells, and it drops again, and another person sells. Bf_Fananatic you are right. Machines are very dangerous. Could ultimately bring down the euro !!!
If a punter went in ladbrokes and said he wanted to place 21 million on a horse that was leading a furlong out it would trigger mass hysterical laughter, but then again bookmakers arent brave enough to engage in such activity that betfair take on every second of every day via automation.
the euro experiment is on-going and of grave concern to all, because generally we abide and live in a capitalist enviroment we have to endure what ever the mass distribution of gloabal wealth provides in terms of poverty and richness.
the human race post the industrial revolution is itself an experiment and the epotential growth of the human race population with its impact on fauna diversty, food chains, polution levels and the very air we breathe will be more properly known by other advanced beings on other planets, no galatic help desk yet though
"but then again bookmakers arent brave enough to engage in such activity that betfair take on every second of every day via automation. "
Betfair brave??? I think you'll find it's the punters on here that carry the can every second on here, bf.
Betfair brave??? I think you'll find it's the punters on here that carry the can every second on here, bf.
what happens to a planet when a species becomes evolved enough to totally change its enviroment
but not advanced enough to understand the long term effects will become the reality of our generations to come, both human and other bio members.
but not advanced enough to understand the long term effects will become the reality of our generations to come, both human and other bio members.
very true ghetto joe, the risk is beared by the customers every second of every day, but this is true in all eco-systems.
maintaing a healthy enviroment for all is best and for gabling look no further
fawwon posted this elsewhere so it may have been lying there for years.
fawwon 30 Dec 11 18:16
There was a bet years ago around this size(maybe bigger), that once popped up on a darts outright market involving Phil Taylor.
Stayed up for a night, I must be the only person who remembers it.
I did take a photo of it but can't find it.
fawwon 30 Dec 11 18:16
There was a bet years ago around this size(maybe bigger), that once popped up on a darts outright market involving Phil Taylor.
Stayed up for a night, I must be the only person who remembers it.
I did take a photo of it but can't find it.
ah yes rem it that big darts backer it was at the right side of a bet and as low as 1.03 it was 200 million or more wow time fly
SoYouThink
31 Dec 11 12:23 Joined: 14 Jun 11 | Topic/replies: 344 | Blogger: SoYouThink's blog
Slightly off topic but the markets in Europe and the euro are one thing that are massively affected by computer software with pre-programmed commands. Something triggers a drop in the euro's value vs the dollar. Someone's programme is designed to sell once the euro drops below a certain value, this pushes the euro's worth down further, so somebody else sells, and it drops again, and another person sells.
This is roughly what provoked the 'flash crash' of May 2010. The Dow dropped about 1000 points and then rebounded within a few minutes. (I managed to nab 300 points in about 15 seconds).
31 Dec 11 12:23 Joined: 14 Jun 11 | Topic/replies: 344 | Blogger: SoYouThink's blog
Slightly off topic but the markets in Europe and the euro are one thing that are massively affected by computer software with pre-programmed commands. Something triggers a drop in the euro's value vs the dollar. Someone's programme is designed to sell once the euro drops below a certain value, this pushes the euro's worth down further, so somebody else sells, and it drops again, and another person sells.
This is roughly what provoked the 'flash crash' of May 2010. The Dow dropped about 1000 points and then rebounded within a few minutes. (I managed to nab 300 points in about 15 seconds).
I've posted this elsewhere.... why didn't the layer or his bot cancel the £21m bet?
Is it possible that BF's system fouled up so badly that it posted the bet as available
and somehow misallocated it thus preventing the layer from cancelling immediately?
There is a bug in the system when you exceed your balance with a lay bet. I've flagged it to BF
previously but they claimed to be unable to replicate it. It's very easy though.
1. Place a lay bet that takes you to the limit of your account balance
2. Using the web interface, increase the price by a tick
3. Say OK to the message about balance being exceeded.
4. Bet is still shown in the list of unmatched bets
5. The bet has been cancelled, but if you click 'cancel' your betting frame freezes up
Is it possible that BF's system fouled up so badly that it posted the bet as available
and somehow misallocated it thus preventing the layer from cancelling immediately?
There is a bug in the system when you exceed your balance with a lay bet. I've flagged it to BF
previously but they claimed to be unable to replicate it. It's very easy though.
1. Place a lay bet that takes you to the limit of your account balance
2. Using the web interface, increase the price by a tick
3. Say OK to the message about balance being exceeded.
4. Bet is still shown in the list of unmatched bets
5. The bet has been cancelled, but if you click 'cancel' your betting frame freezes up
bf_fananatic 31 Dec 11 12:13
The most frightning scenario of automated madness is the amount of nuclear warheads that can be sent to there targets by a series of atomated detection ssytems around the world
one i remember was the scud attacks on israel during the first invasion of iraq. the computers running the anti-missile detection had been left on continuously for longer than had ever been tested, a small error in the internal clocks had accumulated, and they were out of sync with real time to such an extent that they fired off at the scuds too late. so a few scuds got through until they figured out the computers needed a periodic reboot!
(i happen to know as the computers were the same sort that the systems i was maintaining were using )
The most frightning scenario of automated madness is the amount of nuclear warheads that can be sent to there targets by a series of atomated detection ssytems around the world
one i remember was the scud attacks on israel during the first invasion of iraq. the computers running the anti-missile detection had been left on continuously for longer than had ever been tested, a small error in the internal clocks had accumulated, and they were out of sync with real time to such an extent that they fired off at the scuds too late. so a few scuds got through until they figured out the computers needed a periodic reboot!
(i happen to know as the computers were the same sort that the systems i was maintaining were using )
Post Your Reply
Please login to post a reply.
Wonder
Instance ID: 13539