Thursday 23 January 2014

Low Latency Trading

In hindsight, perhaps I shouldn’t have taken a day off from blogging. I don’t believe I have ever had so many serious comments on a single topic. Possibly my thoughts on the great Adam Heathcote which generated a few more, although they were generally of much poorer quality, but very amusing, especially when my concerns over his 'trading service' turned out to have merit.
"Can I post a full and frank apology. Got this all very wrong. The service being provided is truly truly, utterly horrific."
All comments are published in the comments sections of the posts they responded to, and many copied at the end here, but since there is some replication, I’ll be addressing the points raised in general.

The topic is that of the rights or wrongs of the practice of “data-harvesting”, as one article I read yesterday referred to it. Let’s first be clear that court-siding is not illegal. It may often be a breach of the terms and conditions of entry to a sporting event, but it is not illegal, nor does it have anything to do with match-fixing. On those points, I think we can all agree.

We can also all agree that people will always receive their information with varying delays. On-site is faster than radio, is faster than analog TV, is faster than cable TV is faster than satellite TV is faster than internet scoreboards and so on, and yes, if someone chooses to trade off the internet scoreboard or from the close of play Test Match report in their morning newspaper, then that is their decision.

And yes, nothing that Sporting Data or any similar companies are doing is breaking any rules.

The heart of the issue for me is that an in-play market should be as fair as possible to the great majority of people. The disclaimer that "The extent of any such delay may vary, depending on the set-up through which they are receiving pictures or data." satisfied the Gambling Commission in their 2009 study, and they saw no evidence that suggested there was an issue with the differences in data speed. They recognised technological advantages, but in general they were content, in no small part guided by the answers provided by Betfair themselves, that no abuses were taking place.

Here are some responses from Betfair in 2008 to the Commission:
Given the facts emerging from this case, the answer if the question were asked now would have to be different. There is at least one sophisticated operation, run by former employees of Betfair, attempting to do this, and probably more.
However, the details revealed in the past week show that for some time now, at least three years, there has been an organised effort to gain an advantage in markets over, not just a handful of people who willingly trade behind the protection of the countdown, but over a majority of people active in the market. Now again, there is nothing illegal or that contravenes any rules here. Any of us could make the investment to send out ‘analysts’ to various events around the world and give ourselves this edge, so technically one could argue, as several have, that it is fair. All is fair in love and betting after all.

Except for the complete novice, who should soon learn, and one hopes it’s a cheap lesson, anyone trading in-play markets will soon be able to determine their delay behind the first mover. Emp mentioned a cricket match trading at 2.24 and suddenly the price drops to 1.7. A key event has clearly occurred, and it’s the same, to varying degrees, in other sports.

To those who point out that it is possible to develop trading strategies despite this disadvantage, you are absolutely correct, but they will be strategies that you would implement during a pause in play rather than "in-play". There is after all "in-play", and "in-play", by which I mean technically, trading at half-time is in-play betting, but it’s rather different to trading while a game is literally in-play.

When I mentioned that betting during a break in play was akin to betting pre-match, this is what I meant. Perhaps the terms in-running and in-play, should, rather than be synonymous, be used more specifically in this regard? – a market is in running, but betting in-play means ‘during action’. Just a thought, but I think it would be a useful distinction.

Even Emp, who self-reportedly finds it astonishingly easy to develop a profitable strategy despite having obsolete data, must see that trading in-play with a delay greater than the countdown safety net cannot be profitable long-term, assuming the first mover knows what they are doing of course, or is extremely risk averse and leaving some value on the table. Otherwise any bets you are matched at will be at poor value. And if you are making a long-term profit,your profits are reduced from what they would be. As a respected poster on the Betfair Forum puts it:
I'll keep it simple.

You are vulnerable to operators with quicker information than you if
i. you leave up unmatched bets in play
ii. you are unable to cancel these unmatched bets within 5 seconds of a point being considered to be won.
i.e. you are attempting to be a market maker with significantly out of date information, i.e. you are an idiot.
So while the in-running markets may still be lucrative to those with an edge based on skill and knowledge of the sport, the in-play component of these markets is unlikely to be.

Anyway, back to the core of the matter. Below is a silly chart showing, with completely made up numbers, how an in-running market’s traders might be getting their data, and with what delay.
The numbers aren’t important, and are for illustrative purposes only, and what they illustrate is that in the above scenario, in an in-running market with the standard five second delay, someone court-side (not necessarily literally, but with access to court-side data) will be able to match bets left in the market by all but other court-siders and radio traders. In my example that would be 96%.

As many readers will know, the delay is often more than five seconds. Some Spanish football matches are at 8 seconds, and I believe I have seen as much as 10 seconds on some markets. The NHL delay was 5 seconds in the early days, and I believe is now 8 seconds, so the 5 second ‘rule’ is not a rule. Betfair change it at will – it is not consistent across sports, and in baseball I have seen different countdowns on different matches - and while I am not sure what factors they consider before making changes, it’s reasonable to assume that getting picked off will have generated complaints and result in markets dying if individuals feel their protection is gone. Five seconds is presumably the default, because that is the amount of delay that ensures most traders can cancel bets if there is an adverse turn of events. Perhaps it came from the worlds of horse-racing and football, and was just applied to other sports without much thought.

So one might reasonably ask why has there been no change in the tennis markets regarding the countdown. Betfair will know that the same (for example) 20 accounts are frequently winners, but have chosen to ignore the data showing them that an increase to 8 seconds would protect all but the satellite and internet traders (and newspaper readers). Again, I need to emphasise that these numbers are completely nonsense – please focus on the idea.

My assertion here is that Betfair, as a publicly traded company beholden to its shareholders to maximise profits, should take steps to stop the same accounts from continuously profiting at the expense of other traders, and ultimately at the cost of reduced profits for Betfair.

I suggest that by not putting a stop to this practice, they are also jeopardising the long term regulatory approval for in-running betting. As mentioned before, the Gambling Commission has already previously expressed concerns, and evidence showing that Betfair have continued to allow a preventable “unfairness” to persist won’t look good. The fact that at least three beneficiaries are former Betfair employees with inside knowledge of the company’s products, application and workings is another wrinkle.

The Gambling Commission may have been ok with a handful of people having an edge from fast data at one event, and perhaps a different handful of people having the same edge at a second event, and so on, but I doubt that the GC would think too highly of a small group of  people giving up jobs, setting up limited companies and paying for flights to foreign lands, hotels and all the other expenses that go along with that, to make money from a timing loophole. Such a sophisticated operation exploiting a (currently legitimate) loophole would surely be of interest and probably concern to them. What’s their mission again? “Keeping gambling fair and safe for all”.

One final question on fairness – Emp asks why it is fair for someone of higher intelligence to use that advantage while betting, but gaining an advantage from court-siding isn’t.?

I think we’ve already covered that ideally, trading should be a contest between individuals with the outcome based on “skill and knowledge”. While some individuals will lack the dedication to put in the effort and hours to master trading, and others, through no fault of their own, will not have the basic tools to ever learn in the first place, imagine a naturally gifted athlete who trains hard for an event, only to be beaten by a less hard working athlete who decides to use steroids. The authorities do what they can to keep sport clean – perhaps they may seem ineffective, but drug testing and results can be complicated and tainted, but Betfair know which accounts are benefiting from an advantage that could be closed – and yes, again, syndicates using fast-data are doing nothing illegal, nor breaking any rules.

For the numerous commenters who picked up on my reference to the financial markets being regulated, yes it was slightly tongue in cheek. As many readers will know, I have long stated that day-traders sitting at home in their underpants are at a disadvantage. In 2010 I quoted the CEO of Charles Schwab on the subject of day-trading saying “It's a tough gig. You're competing against mega-institutions that are trading in hundredths of a second."  Plenty of others have information faster than you, and some of it is inevitably inside information. And I have written before that it is the same with horse-racing, and with perfect timing based on my Twitter feed, there was something a little dodgy that went down at Newbury just yesterday.

Here are most of the comments on my last post, with the occasional additional response.

Thatwasthepension writes:
This courtsiding isn't fair to the casual Betfair traders but it's within all legal rules. If you think the financial markets are better, forget it. The series of cheating spotted last year by the Nanex team should open the eyes of all financial traders. A good entry point starts in this post: http://www.nanex.net/aqck2/4436.htmlIt's only the tip of the iceberg I believe.
All addressed above I believe.

tyttetrading:
Great post, and great topic. Not sure I understand this part:
"It is illegal for Australian operators to offer micro-betting. In-play betting is allowed on the outcome of a match, but not via online platforms. Punters have to ­telephone a wagering company instead. Many international jurisdictions allow online in-play betting"
If that's the case, where does the in-play liquidity on Betfair during the AO come from?
Anonymous had an answer: 
@tyttetrading,
The statement is almost right. If I correct it I imagine that would answer your question. It should really be:
"It is illegal for Australian operators to offer micro-betting. In-play betting is allowed on the outcome of a match, but not via online platforms FOR AUSTRALIAN PUNTERS.
AUSTRALIAN Punters have to ­telephone a wagering company instead. Many international jurisdictions allow online in-play betting"
So the answer to your question is "from punters not in Australia"
John Walsh had this to say:
I'm sure that my television feed is different being in Canada.

You bet on the NBA. Are you implying that your feed is more than 5 seconds delayed but place bets during the action anyway?
"In-running" certainly, not necessarily "in play", but occasionally depending on my delay.

Matt from the dormant punt.com submitted a small book:
Good articles Cass..
I believe SD are probably set up as a company so that they can employ people, pay them properly and the relevant tax etc. The data is then sold to themselves basically, acting as a betting syndicate, buying the data from SD the company. Betting syndicates not needing to pay tax.. or depending on how they have it set up, they might be betting on their own individual accounts, risking own cash.
The liquidity in the Aus Open is still being provided by remaining courtsiders (imo not many, maybe one or two) and internet feed users (bet365 etc).. net feeds are fast enough to allow you to cancel before you get 'sniped' by courtsiders. Net feeds are still going to be preferable to enetpulse anyway as their info, although it is probably quicker, will have mistakes and also delays around hawkeye challenges.. Though I guess if folks like SD have an automated strategy, they could put safety measures in place. The human eye is superior.
Is it possible for the ordinary punter to win on tennis in play ? Yes. In play more so than pre off.. there's plenty of value, most of the bigger market makers and value bettors use staking calculations which end up playing around the current sentiment... sentiment is emotional, and so there are plenty of emotional cul de sacs the market goes down you can find to exploit if you are patient. At other times, the market is very 'mathematically correct' with tight spreads...maths lacks reality and judgement, if you are good at reading body language or you know particular player attributes then a bet at market price can still have an edge, or at least, a temporary one.
Liquidity is really good in this respect.
Regarding courtsiding in general.. It has cost me a very very large amount of money over the years - I have been sniped 100s (maybe more) times by these guys... It's my mistake, and for the most part, the platform's fault for making it possible, and not the guys exploiting it (moral / ethics aside).. just my opinion, however I accept, I am well informed of the situation, ordinary punters will not know the effect this has to any edge you thought you had over time. I think many moons ago I wrote a blog post saying that when you run a platform, you have platform running responsibilities that bookmakers do not have. That is the ensure there is a level playing field for all your users, and total transparency and best execution in the markets (cross matching is horrific). Betfair have flunked these responsibilities, instead, choosing to play their bookmakers license card when it suits them. Platforms have specific regulation to prevent conflicts of interest (stock exchange?)...Betfair does not need to play by those rules.
Personally I find point to point fluctuation sniping to have some minor shades of grey, I'd rather them do 'something' to make sure it didn't happen, it would be lovely to be able to relax trading these markets, solutions are not readily forthcoming though. The problem with point to point stuff is, there is still room for opinion on price, did you intend to leave you money up? did you fat finger it? did you get carried away or suddenly form a strong opinion on who would win..? You could legitimately place some of these snipey looking bets.
A very large issue for me, was the period of time when courtsiders could grab the entire depth of market after the match point had ended and before even Betfair could suspend the market - that is totally wrong on every level. They have become much better on this in recent times. Betfair response to this was to again suggest that they must allow for people guessing, for chance and opinion.. why could this not be someone guessing that player B was going to win the point? ....The issue is, previous point odds can easily be 1.1 or higher (Wawrinka last night was 1.67 on match point).. they are never going to bet at 1.01 unless they 'know'. Saying that there was an example last week with MP odds of 1.2, courtsiders swiped everything down to 1.01 only for there to be a late call from the umpire, they ended up with 20k at 1.01 and 02 on their books. I'm sure this wasn't enough discouragement though.
Could they use the incoming bet feed to suspend? Could they display it for users perhaps?
Their issue with this (besides whether it is technologically possible) is probably that traders might well just cancel everything when unknown quantities are coming in taking their bets.. even on tight spread action. Matched volumes could go through the floor. It would certainly be fair though..? and they are called, BetFair.
It would certainly be technically possible for Betfair to do something with an incoming bet at 1.01 if the current back price is 1.67. Maybe have an additional delay or reject the bet for being 'out of range' perhaps. But then there's the layers at 1.01 who would be pissed if the 'winning' shot turned out to be long...
Emp is completely right in his statement. If you want to trade in the markets (financial, sporting, name your arena) then it is beholden on you to ensure that you know what you are doing. That means knowing what the latest state of play is as much as it means being able to work out what the correct value is given that state of play.

I am not sure if you have been following the finance markets recently (i.e. over the past decade!) but there has been an exponential growth in High Frequency Trading. This is where the big banks spend ever more money to get over lower latency feed and market access.
Market access used to be measured in hundreds of milliseconds. It has now got to the point where if you are not working to nano second response times you are history. In the light of that then I guess the multi second sporting data latency we are talking about here is an outright joke.
Sporting Data obviously have an edge you don't have. I think the forums need to either work out how to regain that edge or stop complaining.
It seems to me that a lot of the punters who are complaining now are people who have historically done well out of their (value) edge but are now concerned that they are losing it to the low latency players. I don't recall these historically successful people complaining about the money they were taking off the mug punters who didn't have their value edge in the past.
Adapt or die.
Most points already addressed, but just to emphasise again that the "low latency" players value edge comes from knowing what has happened, not from researching, from models or from studying games.

Anonymous (same one) again:
You only need ultra low latency if your edge is really just about getting into the market quicker than everyone else. If you have a different type of edge that is latency tolerant then you are fine.
Most people would say that the most successful "trader" in the financial markets is Warren Buffett. Probably as far away from the high frequency setups as it is possible to get.
As I said before: adapt or die.
Not really. Warren Buffett isn’t what anyone would think of as a "trader". He isn’t in and out of a stock 50 times a day for a start. He does his analysis, he looks at the long-term, and he buys (or sells). More akin to betting ante-post than on an individual match, set or game really.

Anonymous:
I appreciate there's a big intersection on the Venn diagram of Betfair punters and people who wear foil hats, but what would lead someone to believe that just because someone once worked for Betfair that they get any sort of special arrangement?
The people at Sporting Data left Betfair in 2009, since when Betfair's entire management team, every single solitary one of them, has been replaced. Why would Betfair's current management want to give away money to people they've never even worked with?
I don't know that there is any evidence of a special arrangement, but we know they are using their personal accounts, and it would be remarkable if Betfair weren't aware that three of their Premium Charge generating accounts belonged to former employees. Why their management allows this to continue is also unknown at this time.


Hejik left a comment on the Sultan sideshow:
You would appear to have ruffled a few feathers with your cynicism Cassini.

Sultan has got a right paddy on and won't publish my comments (very polite and very fair questions) which is hardly in the interests of fair-play, I'm sure you will agree.
With no right to respond with a contrary opinion on his blog, I figured I'd little option but to raise the question here where readers will hopefully follow his link.
I hope I haven't overstepped.
The Sultan issue isn't a priority for me to discuss right now, but suffice it to say that the argument that you are held back by a lack of money, yet you have an edge, is nonsense. The magic of compounding means that if you have an edge in a sport like tennis, you can soon build a decent bank from any amount. Edges are not derived from bank size as Odwyer (whatever happened to him?) seemed to think.

Emp:
I'd like to address a lot of what you said since you think it's so very ridiculous. I have quite a few observations.
1) It seems somewhat rude and is definitely incorrect to say that I don't think.
2) Betting during a break is most emphatically not the same as betting pre-game, because in sports other than football, prices don't move in that smooth Poisson distribution based manner. Prices that are efficient pre-game can become inefficient during the game (certainly in some sports at least).
3) On the possibility of winning methodologies, a few questions and some observations.It's obviously true that being able to bet after every point is better than only during breaks, but I'm not sure why not being able to do so would cripple all methodologies except those of the people doing this strategy (which is what you imply when you say that only those same accounts will keep winning).
4) If your point is that no one would bet if they knew these sharks were doing this, I assure you that's not true...plenty of people just enjoy gambling and many of them are OK with it being slightly -EV or not as profitable as it would without those tactics. Being first to the info just doesn't seem all that heinous to me, given that even without court-siding, feeds would be assymetric depending on where in the world you live.
5) As some have mentioned, info isn't as a matter of fact symmetrical in financial markets either, and leaving aside stocks, in currencies and commodities,(where insider trading is perfectly legal) there is wild information asymmetry and this doesn't seem to have dried up markets or killed non-shark participation.
6) In the market I really know- cricket, court-siding can be profitable if one did it, but it doesn't hurt market viability at all. To begin with, if one's looking at Betfair, it's pretty easy to tell what a sudden change in odds represents in terms of the action that happened. You'd have to be utterly stupid to see the price change from 2.24 to 1.75 and not realize a wicket has fallen. Recognizing inaccuracies in the adjusted price is more than sufficient to be massively profitable.
I think this is significant because cricket is the best sport for in-play betting and court-siding doesn't even begin to affect the functioning of the market (which isn't even primarily on Betfair).
I would disclaim though that the best approach to cricket-betting is based on hugely in-depth knowledge of the players and not mechanical trading type approaches. In other words, in-play value betting.
I also want to point out additionally something I missed out:
I don't agree with your distinction of this information as being "unfair" but edges created by use of Poisson Distributions, Markov Chains and highly simulated computer models (which most big sports-betters now use and which no recreational punter would ever be able to use or recreate due to financial barriers) as being fair.
Can you please explain a) how automated and algorithmic betting models and the like are fair but this isn't b) how it's fair for me to be watching with binoculars from a high-rise building and placing my bets (definitely possible on most cricket grounds) but not courtsiding and c) assuming courtsiding was banned, why would it be fair if my TV feed was on a 5 sec delay and yours on a 15 sec delay?
Life isn't fair Emp, we all know that. As someone who finds it all so easy to make a fortune, you are clearly one of life's lucky ones.
...develop a method that can withstand being 10 seconds late on info, it really isn't that tough.
Most, if not all, of the key points raised have been clarified above already, although I'll just point out that while all markets are likely to see one party or another have an advantage based on timely information at one time or other, the difference here is that the same small group of people always have that advantage. Quite different.

Anonymous:
Basically unless you can afford a data collection infrastructure delivering light speed latency and you can ignore/tolerate any copyright breach fines on that data, then you are going to lose.
Matt is back:
Not sure if that 'fair' question was directed at me Emp, but to be clear, during a match I'm fine with folks playing with the fastest data - the platform rules allow it, people will do whatever it takes to take out inefficiencies and do this as quickly as possible. Adapt or die as someone else says.. it's just a fact of life (or platform).. Personally I find these guys quite useful at times as I can get a nice price when I spot that they have been able to get some action.
Any comments I made on fairness relate solely to the same speed used once the match has finished in live time, and before Betfair can suspend. That is wrong and unfair imo, there's no prospect of the other side ever winning.
On a lighter note, some BIG money coming in on the Queen's abdication, sent to me by Scott.

And finally, I mentioned Matt's dormant punt.com blog earlier. While retrieving the link for you all, I happened to start reading it and not far down was this excellent post which included these words:
What will eventually topple Betfair does not necessarily need to be completely different (just in the right areas). As their website grew, they moved further away from their original idea. Their decision making process became messy, resulting in poor choices in usability, cost, design, customer service and regulation - the number of areas upon which improvements can be offered by a rival increases, as do the grievances of the current userbase (see Premium Charge, Cross Matching etc. as Cassini pointed out).
What is required IMO is a return to, and improvement upon, what the exchange concept was always about, betting person to person. Of course Betfair is still doing this, but they increasingly push their alternative products, where they take a much greater share of the profits. This aside, there are innovations to be made in the way people bet with one another, the way markets are set up and the social aspect of the site. I have a few ideas, but I won't be posting them here.
One thing I am more than happy to post is that any rival would do well to push for much better regulation. A site that actively seeks the same restrictions (or a variation) placed upon FSA regulated exchanges would be a leap in the right direction. Best practices must surely involve not placing bets on your own exchange (even if it is hedging multiples), not compromising on best execution (cross matching) - exploiting your own platform's short-comings for profit.
Conflicts of interest (or potential conflicts of interest) need to be banished, as does this ability to switch between Bookmaker and Exchange at will, exchanges are very different to the former and should be treated as such.
The length of this post might rival one of Graeme's at The Football Analyst! Thanks for visiting.

2 comments:

Emp said...

Emp asks why it is fair for someone of higher intelligence to use that advantage while betting, but gaining an advantage from court-siding isn’t.?

To clarify, this wasn't what I was asking. I was challenging the assumption that creating a sports-betting model by hiring 15 quants and throwing 15 seasons worth of data through custom-made simulations (modus operandi of most big whales) is the same thing as intelligence. No matter how intelligent an average person is, they can't do this, it's just like paying 'analysts'a matter of resource constraint.

Nick said...

Ex-Betfair employees engaging in this seems to me like someone leaving the vice squad to become a pimp.

The sharks lost trust with Betfair years ago but once the mainstream cotton on it could spell big problems for them.

They need to work out a solution to this else the GC will step in and the only in-play betting will be top flight football.