left-shadow

Technology

Articles related to some of the technologies we’ve created or that we’re into.

31Oct

Could Prozone’s Data Analysis methods be applied into the world of eSports?

Data analysis is becoming fairly prominent in modern football. You only have to look at Manchester City’s recent Opta Initiative as proof to how much of an integral role data analysis is playing in terms of how football is played today.

Prozone is one of the most commonly used resources for data analysis in football, with 15 out of 20 current Barclays Premier League squads using it and over 100 sporting teams using the analytical service worldwide. The service primarily focuses on the following coaching progress: Live Analysis, Post-Match Analysis, Trend Analysis and Opposition Analysis.

Now while the service that Prozone provides is traditionally aimed at the sporting industry as most people know it (Football, Rugby etc.), the above coaching progress could very easily be applied to a very different type of “sport”: eSports.

“eSports”, which concerns the competitive play of video games, is more popular than it’s ever been, with dozens of different leagues and tournaments all over the world, concerning all different types of genre of video game from FPS to MMORPGs, even FIFA has been a part of the World Cyber Games since its beginning in the year 2000. While there’s many different genre of games that are played in eSports, for the purposes of this blog, let’s examine how Prozones methods could play out in an FPS like Halo.

First, we look towards how the Live Analysis function would play out. Prozone’s Live Analysis provides an insight into the performance on the players on the pitch. It allows its users to review key events as they play out in real time and monitor the performance of individual players and the team’s units as the match progresses. As to how this service could play out in eSports with a game like Halo, it could just as easily be converted to giving an insight as to how each individual virtual player on a team is performing, which player has the strongest kill-death ratio? who’s the most accurate shooter? how long has the team held the lead for,? or if it’s a game mode like Capture the Flag, who has held the flag the longest? who has attempted to get the flag the most? who has protected it the most?.

Secondly, we examine how the Post Match Analysis function would be used for an eSports team. This analyses every aspect of the performance once the match is over, it also assess both the players and the team from either single game or multiple game. To place this in the eSports context, it could serve as means of providing an overall report of how the team as a whole performed against their opponents, how consistent they have been with their wins in terms of using it over the course of a few games? who is consistently playing the game well and who the weaker players are?. It can also help the teams understand which game mode suits their playing style better if the modes are alternated between matches; is the eSports team more effective in Slayer than they do in other modes?. This service can be easily used to help an eSports team to get a better understanding of what is working and what isn’t working in their current strategies.

Thirdly, we will look at how the Trend Analysis feature would work in an eSports setting. This is basically a much more specific examination of what works and what doesn’t  whilst also looking at what style of play is being utilized by the team the most and what each individual player does more than anything else. With eSports this could be used in determining what weapons are selected the most frequently? what perks are picked the most frequently?; which player is using the armor lock perk the most? who is using the jetpack perk the most? and more importantly, are these choices of perks working for that individual player? Is there something better they could be using to improve the team overall?.

Finally we look at how the Opposition Analysis could be used in eSports. Prozone uses its Opposition Analysis feature to identify upcoming opponents strengths and weaknesses, therefore informing management and coaches of their tactical decisions as well as improving the preparation of their team. This would be directly applicable to its possible use in eSports, helping teams figure out what areas of the game their future opponents are strongest and weakest at, so for example, do they hold their leads longer than you do? What types of weapons are they using? Are they better with them than you are? Which game modes do they play better in? Do they kill spawn? Do they camp?. In many sense, Opposition Analysis covers the first three areas just applied to opponents instead of your team.

While there are options available for Data Analysis such as eSports Scout, they are currently very basic and are limited in the amount which are available to use. They are nowhere near as refined and as evolved as services like Prozone.  If such precise level of service was made available, Data Analysis could become as much of an integral part in eSports as it is in football today.

So as for asking the question of could Prozone’s methods of data analysis be applied into eSports, the above statements suggest that the data that would be collected wouldn’t be too far away from the player in the real world and the player in the virtual world.

This guest blog by Ryan Edwards (PR intern) – Ryan is a former International Journalism student from Liverpool John Moores University, with a working background in PR and Marketing. *All views are his own.

https://twitter.com/RyEdwards91

17Apr

Channel 4 hails project with “forward thinking Setgo” a success

A belated entry for some work we did for Channel 4 back in late 2011. We may cover the projects themselves in more detail in a future entry. In the meantime, here’s a description from Channel 4′s Winter Briefing which sums them up better than we can right now:

“This year Convergent Formats has been developing a range of simulations and prototypes that illustrate programming in a converged world. One particular success was a news project with forward thinking Setgo. They proposed a content curation concept around Channel 4 News that could be experienced in different formats. Each news item had a short, medium or long version and the viewer controlled the length of the item they wanted to watch, allowing them to view headlines, skip through stories, watch an item of interest in full and so forth. This was particularly successful as it considered how people behaved around digesting this content now such as scanning newspapers or surfing online and brought this experience to the television.  Setgo also built on this concept with
developing a clever use around archive and past stories where viewers could access related media, find out more about a subject that they were watching within the viewing experience.”

And here’s the bit that explains who commissioned the work at Channel 4 at the time:

“Convergent Formats is a relatively new, future focused division in Channel4.  Its aim is to invent truly exciting, ground-breaking TV formats ensuring C4 remains the industry leader in creative innovation. This  commissioning fund is developing and building creative concepts for the next generation of televisions and connected devices often referred to as connected televisions.”

Click here for the full Winter Briefing.

We’re hiring! Check out our Jobs section for more details.

 

3Oct

MongoUK 2011

Mongo DB

Welcome!

At last, a good excuse to write a technical article: The Mongo DB 2011 UK conference was held in London on September the 19th and we wanted to relay some of our impressions from this (excellent) event. We won’t go into each talk that we attended in detail, but pick out what we thought were the main takeaways and themes.

What is Mongo DB, anyway?

You haven’t heard of Mongo? Ah, that’s OK, it’s still some way from mainstream. It joins Cassandra, CouchDB, HBASE and several other new databases that are based on a “NoSQL” philosophy. Their raison d’être is to support the massive horizontal scalability required from Internet applications with millions of users, and they do this by throwing away the idea of (enforced, database level) referential integrity and fixed schemas. Mongo uses a JSON document as its data model, with very few restrictions on the shape of data that can be written within a single collection (the equivalent to a database table).

The Conference

First off, a word about the conference: It was a well run, unpretentious affair, with senior 10gen executives mucking in and helping with registration, etc. Everyone was very open and honest, often highlighting gotchas and limitations with the product, which we found very reassuring.

If you didn’t know Mongo DB intimately you’d have got on just fine, it was mostly use-cases and performance tips that were applicable to most users and interested parties.

Heavy Lifting

There were some notable success stories from large-scale mainstream users: The Guardian were big fans after feeling trapped on Oracle with no way of scaling up without “handing over their wallets to the Oracle reps”. Their different post types (news, articles, opinion, blog) and links match perfectly to the schemaless nature of Mongo and, having made the switch, they now feel empowered to iterate their site in ways that would have been impossible with a RDBMS and even open up their data to 3rd parties. [more]

The National Archives was another migration from RDBMS hell: Only this time it was many disparate SQL Server databases with a total of 2000 tables. By mapping them into a single namespace they too were able to give people access to their peerless catalogue. It was interesting to hear from an organisation that we would have expected to be technologically conservative – it appears that after their successful trial they will be ramping up as they start to archive 100s of TBs of government data.

Of course, both of these organisations are a near-perfect fit for Mongo: Millions of flexible documents in a read/query heavy environment. What we also wanted to know was what happens in less traditional “web app” workloads…

Gotchas & Limitations

Note: This might read like a “reasons not to use Mongo DB” list, but it really isn’t: Mongo seems a fantastic tool with an almost perfect feature-set. But it is young, and there are some sharp edges in there still.

Write Performance

The primary takeaway for us was several warnings on write speed as a pinch point. Unlike the versioning, master-less Cassandra, Mongo DB has a single primary machine per shard responsible for processing writes, so you should carefully monitor the amount of time the master is spent locked for writes. This is compounded by AWS EBS volumes having unreliable, variable and just down-right poor I/O performance. The workarounds proposed seemed rather desperate at best: Massive striping (16 drives in a RAID 10 configuration) or using the ephemeral local instance storage for volatile stores.

Sharding Fun

The real fix for write performance is to throw more boxes at the problem via sharding. Whilst conceptually this is fine, in reality some talkers had experienced pain and bugs implementing it in earlier Mongo versions. You also have to make sure that your sharding key (the piece of data you will use to determine which shard you want to store each document) is well formed – if you use the date of the document, for example, you will find all your new documents hitting one shard. The other gotcha is to shard early: If your boxes are maxed out already, there won’t be any horsepower available for moving the load (chunks) off.

Tips & Tricks

As well as “avoid this world of hurt” type advice, there were a few cunning ideas to wring more performance out or make life easier. The engineers from uberVU compress multiple attributes (e.g. “Gender=Male, Location=London”) into a single array of integers per document that represent each key/value pair. This allows them to use a single multi-key index for all attributes and makes their filtering operations very fast indeed. They and other teams also roll their own _id field (the default GUID for each document) and embed as much information in as they can into 64 bits while still leaving some room for random salt to avoid conflicts.

Overall Impressions

This was definitely one of the best conferences we’ve been to – anyone and everyone with an interest in building high-performance and high-availability databases (at a large scale) would have found something to spark their grey matter.