Heh, I really don't mind tbh - it's part of my job. I just would say you guys will get more use out of me if you don't treat me as the enemy. The forum will get more value if it's not about mudslinging but actually understanding the issues of exchange market micro structure, of which I have unique experience to comment on.SeaHorseRacing wrote: ↑Wed Oct 10, 2018 7:59 pmThink we should give Jason a break. It’s like he’s cornered with sucker punches.
Smarkets's skullduggery
Right or wrong that's more info than Betfair have provided following all their outages combinedvide0star wrote: ↑Wed Oct 10, 2018 8:10 pmMy guess, and I want to emphasise it's just a guess, is two reasons.
1) They have an old-ish technical stack. This presents all sorts of problems and gets worse as time passes. An example problem might be you have a database that has 100gb of space. As it fills up you need to find a way to make the database bigger. You can't just increase the database size without downtime. There will be 10s of these issues that get harder if you don't pay what's called "technical debt".
2) Exchanges are hard to split up. That means putting trading on multiple servers so that if one group goes down, the rest still operate as normal. It's hard to split because betting exchanges need to keep track of real time exposure across all contracts and sports. An example would be, say you had a football exchange server and a horse racing exchange server, it would be nice if horse racing crashed, that you could still trade football. It's very difficult though because your exposure in horse racing is linked to football exposure.
As Euler and others have said already it deserves respect and acknowledgement that you have taken the time to try and answer all these questions - something Betfair can certainly learn from
For people interested in the Architecture at Betfair this is an old presentation when they sat on their old infrastructure: -
http://www.hpts.ws/papers/2007/HPTS%20W ... ywheel.pdf
I checked to see if the presentation was still publically available.
http://www.hpts.ws/papers/2007/HPTS%20W ... ywheel.pdf
I checked to see if the presentation was still publically available.
They need a slight lag I think, if the price just changes as you click then it still places the bet at the old price.JTEDL wrote: ↑Wed Oct 10, 2018 7:38 pmSaw this earlier, skullduggery or user error?
https://mobile.twitter.com/SmSportstrad ... 2836814849
- Kafkaesque
- Posts: 886
- Joined: Fri Oct 06, 2017 10:20 am
This is going to be vague, but someone - in and amongst one the many debates on BF downtime - posted an article on the data structure (might be a different definition, as I'm not tech savy to that level). The comparison was made to why social media and other major sites with lots of traffic do not suffer from the same issues, I seem to recall. At the time I read it, but not that carefully as it was just too much of a standalone and without context, again, for someone with little tech background. I do however feel it may well be more interesting in the context of the above, if someone remembers where it was posted? Even without remembering specifics, it rings true with the above.vide0star wrote: ↑Wed Oct 10, 2018 8:10 pmMy guess, and I want to emphasise it's just a guess, is two reasons.
1) They have an old-ish technical stack. This presents all sorts of problems and gets worse as time passes. An example problem might be you have a database that has 100gb of space. As it fills up you need to find a way to make the database bigger. You can't just increase the database size without downtime. There will be 10s of these issues that get harder if you don't pay what's called "technical debt".
2) Exchanges are hard to split up. That means putting trading on multiple servers so that if one group goes down, the rest still operate as normal. It's hard to split because betting exchanges need to keep track of real time exposure across all contracts and sports. An example would be, say you had a football exchange server and a horse racing exchange server, it would be nice if horse racing crashed, that you could still trade football. It's very difficult though because your exposure in horse racing is linked to football exposure.
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
But to be fair you could assume these are their likely weaknesses. They're fairly standard issues with systems that have to remain live, there's no easy opportunity for housekeeping. If anyone has had experience designing exchange systems then they're massively impressed with just how reliable betfair are, especially in the implementation of the ESA. The moaners really ought to look at what's involved. I wish the Smarkets guys all the best because bullet proof bespoke systems don't come cheap, that 3.5mil will get burnt pretty quickly.
Interesting to see that they link their horse and football exposure. Exposure? On an exchange?
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
Kafka social media tech is Lego compared to exchange stuff. They won't have anything like the same problems staying live.
Skullduggery is a great word.
On a side note Betfair have added a few tech articles to the forum:
https://forum.developer.betfair.com/articles
And here (was surprised to see Betfair know what SRE stands for...)
https://ppb.technology/
On a side note Betfair have added a few tech articles to the forum:
https://forum.developer.betfair.com/articles
And here (was surprised to see Betfair know what SRE stands for...)
https://ppb.technology/
A several years ago I had a lot of sympathy for Betfair when they had outages given the complexity of the system they're running, but the sympathy well has now run dry.ShaunWhite wrote: ↑Wed Oct 10, 2018 9:00 pmBut to be fair you could assume these are their likely weaknesses. They're fairly standard issues with systems that have to remain live, there's no easy opportunity for housekeeping. If anyone has had experience designing exchange systems then they're massively impressed with just how reliable betfair are, especially in the implementation of the ESA. The moaners really ought to look at what's involved. I wish the Smarkets guys all the best because bullet proof bespoke systems don't come cheap, that 3.5mil will get burnt pretty quickly.
Interesting to see that they link their horse and football exposure. Exposure? On an exchange?
Regarding exposure, I think Jason was referring to the customer's exposure. I always just assumed that their data centres would have a central database of all the customer's transactions and unmatched bets so exposure can always be verified regardless of what sports are traded.
-
- Posts: 3140
- Joined: Sun Jan 31, 2010 8:06 pm
ShaunWhite wrote: ↑Wed Oct 10, 2018 9:00 pm
Interesting to see that they link their horse and football exposure. Exposure? On an exchange?
I think he's talking about customers exposure, Shaun, rather than the exchange's exposure.
I'd basically like to know why we're still waiting for an open API despite smarkets saying they'd be one ages ago. It's no surprise prices aren't available when people try to take them if they're limited to a web interface. I'd imagine the Market makers with API access are more than happy with the current status though as it almost forces players to adopt a mentality of jumping prices in order to get matched.
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
Fair point about the interpretation of exposure but as I've never even contemplated linking those two (and can't see why I would) I made a hasty assumption that he wasn't talking about me aka a customer.
An open to anyone api is vital, and access to it should be free to encourage 3rd party developers plus attractive pass thru bonuses. Yet another gambling web site won't set the world on fire, a plethora of innovative apps, with each publisher pushing them like crazy is the way to rapid growth. The current business model seems really dated.
Or go a stage further, the value in this business seems to be the machinery rather than the brand, so hosting the exchange would seem to much simpler and more lucrative than aping (no pun intended) the Betfair model, perhaps in the same way dozens of poker and casino sites are just skins over handful of game servers. Then allow privileged access to it for the trade customers to benefit from the pooled liquidity. The challenge is making 'Powered by Smarkets' the trusted badge of approval.
At present I can't seem to get exited enough to invest any time, effort or money into what's on offer, "I'm out". And if they've only raised 3.5mil in 7 years then I'm probably not the first to say that.
An open to anyone api is vital, and access to it should be free to encourage 3rd party developers plus attractive pass thru bonuses. Yet another gambling web site won't set the world on fire, a plethora of innovative apps, with each publisher pushing them like crazy is the way to rapid growth. The current business model seems really dated.
Or go a stage further, the value in this business seems to be the machinery rather than the brand, so hosting the exchange would seem to much simpler and more lucrative than aping (no pun intended) the Betfair model, perhaps in the same way dozens of poker and casino sites are just skins over handful of game servers. Then allow privileged access to it for the trade customers to benefit from the pooled liquidity. The challenge is making 'Powered by Smarkets' the trusted badge of approval.
At present I can't seem to get exited enough to invest any time, effort or money into what's on offer, "I'm out". And if they've only raised 3.5mil in 7 years then I'm probably not the first to say that.
I appreciate the fact that Jason takes the time to write here and answers lots of questions.
Still, as the first poster said, the fact that smarkets runs its own in-house market maker and derives a large part of its income from that creates an obvious conflict of interest. This combined with the fact that they don't provide an open API looks like they don't really want smart action.
If you take liquidity and win, then you cost them money directly. And if you provide liquidity and win, then you compete with their own market maker.
Maybe they appreciate increasing activity on their platform so much that they are fine with competition to their market maker, but the fact that they still don't provide an open API looks to me like they prefer the status quo: Their own market maker and then a few others that maybe fill some gaps where their own market maker is not so good at providing liquidity.
Still, as the first poster said, the fact that smarkets runs its own in-house market maker and derives a large part of its income from that creates an obvious conflict of interest. This combined with the fact that they don't provide an open API looks like they don't really want smart action.
If you take liquidity and win, then you cost them money directly. And if you provide liquidity and win, then you compete with their own market maker.
Maybe they appreciate increasing activity on their platform so much that they are fine with competition to their market maker, but the fact that they still don't provide an open API looks to me like they prefer the status quo: Their own market maker and then a few others that maybe fill some gaps where their own market maker is not so good at providing liquidity.
Thanks for the link, Liam! I thought betfair had shut down / outsourced developer.betfair.com.LinusP wrote: ↑Wed Oct 10, 2018 9:12 pmSkullduggery is a great word.
On a side note Betfair have added a few tech articles to the forum:
https://forum.developer.betfair.com/articles
And here (was surprised to see Betfair know what SRE stands for...)
https://ppb.technology/