When Sports Betting Is Legal, the Value of Game Data Soars
nytimes.com
nytimes.com
Others dismiss that view. Marc Edelman, a professor of law at Baruch College, said he believed that only "pre-scripted" events were subject to copyright — meaning that while professional wrestling performances might qualify, football, basketball and other true competitions would not."
Remember when Spotify sought a blank cheque to do what they wanted with your personal data?
this seems close to sports data, in that its not "pre-scripted" and the company that runs the markets owns that data and sells it, in alot of cases its one of the larger sources of revenue.
I can't see how the stock exchange can own that data but the sports leagues can't own the data about their leagues. Though as someone who used to do sports betting I wish this was the case:)
In the sports analogy the market participants are the same as sports fans watching the game. They get to see the plays unfold (market data) but have to pay for that "privilege".
In that case even if you are a participant of the stock exchange you still need to pay for the market data and historical data. Sports leagues could piggy back off this I guess. Seems like a stretch tome but I'll bet some league tries this.
If you were to be able to access and compile the data independently you could publish it and claim copyright.
This is similar to how the guys in the article are being ejected from the stadia ... because they're controlling access to the source data.
* I am not a lawyer.
* I am not a lawyer.
* Not a judge on the supreme court
I'm not a lawyer, so if you're going to court based on what I say please consult with somebody who is.
For example, if I use convolutional neural networks to parse broadcast video and derive statistics from that data, how could I not own that information?
Somebody else could independently gather that data, produce alternative, even identical results and their work would be copyright.
I am not a lawyer but that's how I understand it.
It's not a copyright violation to collect and share the same information. It _is_ a copyright violation to simply copy and republish the data from a given source.
So how can a map-maker tell that someone copied them instead of collecting their own information? They insert watermarks in the form of inaccurate data (such as inserting a small city that doesn't actually exist). See: https://www.google.com/amp/s/gizmodo.com/the-fake-places-tha... and https://en.m.wikipedia.org/wiki/Fictitious_entry.
I wonder what kind of fictious entry you could inject into sports data that wouldn't compromise the quality of the data too severely?
Now if the game is truly rigged (ie pro wrestling) then the "score" is part of a prewritten script. That's different. But even there, fair use is a thing. One cannot stop someone from paraphrasing the plot of a script, such as in a review. So while experts like Rodenberg may be correct, they are correct in such a narrow way that it doesn't matter.
That is pretty fascinating. In horse racing (where paying for past performance information has been standard operating procedure for more than a century), there is (at least in the US) exactly one company that holds and brokers that information, and it is quite a profitable business.
It's interesting that they theoretically can't claim copyright to protect that business.
The fact of what physical geography exists cannot be copyrighted. Anyone is free to produce their own map by measuring the physical world. In theory if everyone used accurate measurements, then all maps would be identical. However a map itself can be copyrighted, and producing your own map by copying another for which you don't have permission is illegal.
But how would anyone ever know whether you went out and measured the real world, or just copied someone's map because the underlying data is the same.
Historical horse racing data is the same. You could start gathering all the data by recording the outcome of races, and in 10 years you would have 10 years of historical data. Or you could just pretend you have already done that and actually just copy the data from the other company. The result would be indistinguishable.
'Trap steets' is one way the mapping industry deals with this problem - https://en.wikipedia.org/wiki/Trap_street - So I would not be surprised to find some deliberate mistakes in the past performance data provided by this company, which should they be found in your data set, would be pretty conclusive evidence of the source of your data.
"What colour are your bits" is also relevant and interesting article - http://ansuz.sooke.bc.ca/entry/23 - Despite what techies would like to believe, the source of your data does matter, even if multiple different sources would actually result in an identical block of data.
> The district court held that Motorola and STATS did not infringe NBA's copyright because only facts from the broadcasts, not the broadcasts themselves were transmitted. The Second Circuit Court agreed with the district court's argument that the "[d]efendants provide purely factual information which any patron of an NBA game could acquire from the arena without any involvement from the director, cameramen, or others who contribute to the originality of the broadcast" [939 F. Supp. at 1094].
It’s a really fascinating line that the court drew. The concept that a certain player scores a 3 pointer with around 2 seconds left on the clock is clearly a fact knowable to every patron. But what about the exact location from which they took the shot? What about the number of milliseconds the ball was in the air, or the angle at which the shot was launched? These are facts that mere humans cannot access with accuracy; they require the involvement of cameras, camera operators, and other entities characteristic of a copyrightable expression of the fact. If I were to use these to create a 3D simulation of a game, would that not be a derivative work of the film from which I obtained sufficient facts to make that simulation? Where do we decide that pixels on a video feed are more special than other parts of the content?
It comes from the Copyright Act: 17 USC 102 states that copyright protects "original works of authorship fixed in any tangible medium of expression." It also states that in "no case does copyright protection for an original work of authorship extend to any idea, procedure, process, system, method of operation, concept, principle, or discovery, regardless of the form in which it is described, explained, illustrated, or embodied in such work."
The way this has been interpreted is as a dichotomy between ideas/facts and expression. You might use very expensive equipment to, for example, map the ocean floor. That data cannot be copyrighted. But a video visualization of that data could be copyrighted.
The result is, as you point out, somewhat unintuitive in that case, since generating the data is the hard part, not creating the visualization. But that's where Congress chose to draw the line in order to circumscribe the scope of copyright.
And then the fun part, where data owners intentionally add false data to catch people. As a false item, it's not protected under the fact exclusion, so any inclusion of it is clear infringement.
To put it another way, if I build a stock-picking algorithm, the “facts” about what trades it makes are public knowledge, but can it be argued that the algorithm can be protected from reverse engineering attempts?
Maybe Marc Edelman thinks we've got this copyright thing all backwards.
For years, people getting into sports analytics have done so via baseball (and the Sabremetrics community), and the NBA because the data has not been seen of commercial utility. It's been collected by fans.
That will change dramatically, but it should be resisted. Leagues and players should embrace open data because it will in the long-term lead to analysis that helps them, but more importantly, fosters a deeper interest in their game and therefore makes their own careers more valuable.
It would be an extremely dumb cheater that only began to cheat because they could conveniently bet on a site in a state that has jurisdiction over them.
You don’t even have to match fix to alter behavior in a way that has a financial return for things like fantasy sports that are stat based.
I just don't think there will be a massive increase because Americans sites allow gambling. Asian sites move billions of dollars a year in largely unregulated betting markets. Athletes that wanted to cheat could already do so in relative safety. They would be foolish to start cheating in a situation where both the likelihood of them being caught is increased and the consequences of them being caught is worse.
Depending on which data you need, there are already some good sources of free football data.[1][2][3]
Someone has also conveniently wrapped much of this in an R library.[4]
Football is actually one of the better sports in terms of easily obtainable data at no cost. Rugby is much more difficult to find extensive datasets, although there are some interesting attempts.[5]
Decent cricket data also exists in a few places[6], but generally requires faster and more regular updating. However, there are R libraries for cricket data too.[7] This one scrapes from the ESPN Cricinfo site.
It is possible to obtain horse racing data for the UK and Ireland at a reasonable price, for personal use[8] and Hong Kong does a great job of making a huge volume of horse racing data available at no cost, but not in a particularly machine usable format (extensive scraping required). Sadly, other large racing jurisdictions such as Australia and the US don't have anything free, or even reasonably priced, as far as I'm aware. Ray Paulick has covered this as a general problem for the sport for a few years now.[9]
[1]http://www.football-data.co.uk/data.php
[2]https://github.com/openfootball
[3]https://github.com/jokecamp/FootballData
[4]https://github.com/dashee87/footballR
[7]https://github.com/tvganesh/cricketr
[9] https://www.paulickreport.com/news/the-biz/gardner-horse-rac...
The type of data that people in this thread are talking about would be more in-line with detailed positional information about each of the players on a football pitch over 90 minutes. In a cricket context, it would be more along the lines of the exact release angle and speed for each of the bowlers.
This type of information is clearly available, as Michael Caley is able to quickly generate xG maps for an entire game[1], but I do not believe it's public.
Your [9] link points out that much more information is available to baseball betters, but even baseball has a significant walled garden in terms of data. For example, the raw data used to generate the stats in [2] is not open to the public.
My links were all to post-event data, not live in-play data sources. I still wouldn't call, for example in a cricket match, the number of wickets taken by a bowler a stat. It's just data. A stat is derived from the data, for example bowling stike rate or economy. Or that a trainer had a winner at a certain race track. That's just the post-event data. If you want to derive further statistics, you have to calculate it yourself.[1]
The links above just have, for the most part, raw event data.
[1] https://blog.betwise.net/2018/06/19/loops-with-r-creating-a-...
I'm not being needlessly pedantic, it's an important distinction when considering the level of analysis that one is able to perform. If you are doing major cricket analytics, you need ball-by-ball information, including as much information about the bowler's position, movement and arm motion, batter's position, movement and stroke information, how the field is set up, conditions of the pitch, situation in the match, etc.
For example, consider a situation where we're attempting to compare two bowlers. Bowler A may have got a wicket off a shot that 95% of batters would not play, whereas Bowler B did not get a wicket despite bowling a ball that achieves a wicket 10% of the time. The stats suggest that bowler A is in better form, but a data-driven view of the game suggests that bowler B is actually in better form.
As it stands, stats are available in abundance for every major sport, but detailed data is not. If a better had access to the latter, and they were were able to parse it with an in-depth understanding of the sport, they'd be at a huge advantage versus betters that did not, and they would reap the benefits.
Take the NBA for example. Let's look at this: http://toddwschneider.com/posts/ballr-interactive-nba-shot-c... - this is able to give you super detailed analysis thanks to the NBA's stats API.
The equivalent from Opta is thousands a year per competition. I was fortunate enough to get to play with detailed Opta data and ChyronHego data as part of a Man City hack day a couple of years ago. The latter data simply isn't commercially available.
For cricket, you can do something interesting with ball by ball data, but ideally you want ball tracking data. You want to know speed of release, length, speed and movement after the ball has pitched, and speed after interaction with the batsman along with angles, etc. - and that's just to get started. Ideally you want positional data on fielders, etc. too.
Don't get me wrong, this is a great starting set to get people interested, but there's a way to go for high-quality data being accessible to the hobbyist or academic researcher (although I believe Opta gives academics discounts to help make them "the" standard for clubs, etc.)
RunningBall is all the real time data - you pretty much can't run an in-game book without it. It practically runs the in-game betting world.
For example, consider a game with evenly matched opponents. At 0-0, the server might have odds at ~1.5, with the non-server having odds at ~2.7[1]. When the score moves to 0-15, the odds might move closer to even, eg 1.9/1.9. If you're able to get information about the first point ahead of the crowd and place a bet on the non-server at 2.7 when the true "predicted outcome" is closer to 1.9, you obviously have a massive advantage.
This can also be used to bet on sets or matches, but the advantage is much smaller. Still, a better with any sort of advantage will always win over the long term.
[1] Yes, those don't add up to 100%, welcome to the vig :)
It is amusing to see integrity and sports betting in the same sentence. You created a game where people are incentivized to make bets on the next action. What do you expect? Of course people will want to gain an edge anyway possible.
Oh this is gambling. The House has to win everytime. Sorry I forgot.
Will there be high-frequency sports betting, like HFT on financial markets? Why not?
Most successful long-time sports betters are not playing on a level playing field. In order to beat the vig over the long term, a successful better must be privy to information that is unavailable to other betters, or at least the majority of betters. This either means a novel analytical technique, which is rare, or inside information.
The number of people who simply watch games closely and are able to discern information that allows them to bet successfully on future games over the long term is exceptionally rare in the real world.