Working on something totally different...

[BattleMode]

Stopping with: I have decided not to spend any more time on COD4 or TF2 development in the foreseeable future. The biggest risk I see there is that I spend a lot of time on something but I have no channels or connections to get people to actually play them. I don't mind spending time and effort on creating things, but I do mind making things no one uses. I will not abandon my UT3 stuff, but I will probably not create anything new for UT3. I already made more stuff for the game than most other MOD developers anyway.

But starting on: I have decided to do some work on websites instead. From 1998-2003 I ran http://www.theclq.com. A 200.000.000 pageview website with 160.000 registered gamers dedicated to monitoring game servers. Nowadays there are sites like game-monitor.com, serverspy.net and gametracker.com. None of them come even close to the scope TheCLQ used to have. I had good reasons to stop with TheCLQ 5 years ago, but several of those reasons are no longer valid. Especially hardware has come a long way, so I think it should be possible to do more. Also I learned a lot about gaming and Internet technology the last 5 years so I think I can do a better job this time.

This is all hobby (I also have a real life), and I will also keep spending time on keeping BBS alive and kicking so I am not sure how far I will get with this.

What are my plans (what actually happens might be completely different):

• #1 do some research on other websites that do server scanning
• #2 rent a few powerful servers
• #3 resurrect my old game scanning technology and get it to work with new games
• #4 start a small site (AllGameServers.com) with a few games/gametypes and a few stats to test everything out
• #5 resurrect my old database technology (used to do 100.000 transaction/second 10 years ago, should be able to do better now) and try to integrate that with the scanner and the database (used to be MS Jet and MS SQL-Server, will probably be MySQL this time)
• #6 add neat functions until people start noticing the site, and then keep adding neat functions

What kinds of stats and functions do I have in mind (again what actually happens might be completely different):

• Popular servers per game
• Details stats per game server
• Who is playing on what server, with some kind of alert system (e.g. a Firefox extension) helping you to decide when and where to play
• Some form of ranking, TheCLQ used to be global I think this time I will probably do separate ranking for each server
• Multiple different ranking methods; TheCLQ's ranking was not liked by everyone, which will apply to any ranking system, so perhaps I simply should have multiple methods and let the visitors decide for themselves
• Some form of awards and achievements (perhaps both global and per server)
• Tools helping gamers to organize fun matches, or perhaps even complete tournaments where the site could automagically keep the scores etc.
• Map statistics (e.g. which maps scare people away, which maps bring in new players to assist server owners in optimizing there map collection)

[rhiridflaidd]

Sounds amazing. Microsoft insists that every game has rewards - and often, like in ut3 for 360 it's implemented half heartedly by developers that don't fully realise the addictive power of stats and profiles.

If you then set targets and rewards for players, this could take off in a really, really big way, for games that don't have a good ingame ranking system. I've made a fair few posts discussing these sorts of general principles already. I can't actually do anything myself though but happy to post suggestions.

Full circle would be to have an online game that would need points generated by real game playing to progress in. But that's way off.

[BattleMode]

Originally Posted by rhiridflaidd

Sounds amazing. But rather than target server owners, targetting the players themselves ould be good. Microsoft insists that every game has rewards - and often, like in ut3 for 360 it's implemented half heartedly by developers that don't fully realise the addictive power of stats and profiles.

If you then set targets and rewards for players, this could take off in a really, really big way, for games that don't have a good ingame ranking system.

Like TheCLQ in "the old days" I want to target the players. I do indeed intend to create some sort of targets and rewards from outside the games.

What I have in mind ATM is to make AllGameServers.com more oriented towards servers and TheCLQ more oriented towards players.

But it is all in a very early stage. Although I did get my old game scanner working again today. Scanning UT2004 servers with it to experiment.

[rhiridflaidd]

Is "Gamespy monopolising access to the servers" still a problem?

[BattleMode]

Originally Posted by rhiridflaidd

Is "Gamespy monopolising access to the servers" still a problem?

Yes it is. But I designed ways to work around that now. I am usually quite open about things, but in this instance I won't give more details.

[Shambler]

Interesting, I do remember TheCLQ when I played Unreal1, a long time ago indeed

Some negative things about stats/ranking which I noticed in the past though, is that overall it promotes a more self-oriented approach to playing games, which can end up taking away from teamwork and a lot of 'fun' elements of a game; also promoting tactics that aren't conducive to fair play (e.g. constant spawnkilling, team stacking etc.).

It also promotes exploiting the actual stat system itself in order to gain increased points etc; you will notice with the UT2004 stats, that the system was abused so that mostly cheaters were at the top.



What is your experience of this in the past? Have you got any way to prevent this kind of stuff? (it's my experience that it can actually do a fair bit of harm to communities playing the game)

[BattleMode]

Originally Posted by Shambler

Interesting, I do remember TheCLQ when I played Unreal1, a long time ago indeed

Some negative things about stats/ranking which I noticed in the past though, is that overall it promotes a more self-oriented approach to playing games and can end up taking away from teamwork, and promotes tactics that aren't conducive to fair play (e.g. constant spawnkilling, team stacking etc.).

It also promotes exploiting the actual stat system itself in order to gain increased points etc; you will notice with the ut2004 stats system, that the system was abused so that mostly cheaters were at the top.

What is your experience of this in the past? Have you got any way to prevent this kind of stuff?

You sum up the biggest problems quite nicely I must say.

Cheaters used to be a big problem, but I think I can avoid that now. I used to do global ranking (all servers combined) which meant that a single bad server could wreck the entire system. My current plans are more single server based statistics (which is way more complicated but I think modern hardware will be able to handle this). This way the quality of the server management will determine the quality of the stats for that server. So the bad apples will no longer spoil the fun for the rest.

Ranking having a bad influence on game play is another one. I cannot really solve that I think. All ranking systems seem to have problems there. Actually all scoring systems seem to have problems there. The best way to solve that is to make team play result in people getting more points, to solve it from inside the game itself.

[Shambler]

Originally Posted by BattleMode

Ranking having a bad influence on game play is another one. I cannot really solve that I think. All ranking systems seem to have problems there. Actually all scoring systems seem to have problems there. The best way to solve that is to make team play result in people getting more points, to solve it from inside the game itself.

Yes that's something I tried with ONSPlus in UT2004, added a few scoring enhancements.

Perhaps one solution would be to create an optional serverside component for your stats system, which handles statistics directly from the game server, and also introduces scoring adjustments/enhancements for particular gametypes, to promote fair play?
There would probably be lots of bonuses with regards to bandwidth/accuracy etc., as you could cache data and only send it at the end of a game too; then (on your website) you could mark servers as 'officially ranked' or something like that, which would signify improved quality of ranking.

It would also make it far easier to prevent people from exploiting the system, and has the potential to make it easier to weed out cheaters as well; you could even go so far as to give each server some sort of unique ID (or require a user/pass for registration), then when displaying the quality of ranking on your website, you could give an extra quality 'point' for every month that the server has been officially ranked? (stops people registering a server, just to abuse the system)

That would allow a whole load of new ways of ranking servers, and of ensuring quality of play at the same time; you could even determine the quality of players stats based upon the servers they've been playing on.

[BattleMode]

Originally Posted by Shambler

Perhaps one solution would be to create an optional serverside component for your stats system, which handles statistics directly from the game server, and also introduces scoring adjustments/enhancements for particular gametypes, to promote fair play? ...

That would be a very interesting aproach. It would be tough to achieve (e.g. building such an component for every online game there is, convincing a significant number of server owners to install it) but it would also create very interesting possibilities.

Will keep this in mind (first have to get something working at all).

[rhiridflaidd]

If you get points for vehichle healing, then for Tank crossing, for example, any vehichle healing should be doubly rewarded since that's the point of the map. The more detail the better really.

I think though it is important that you keep any discussion of how exactly the scoring works fairly close to your chest. I also think that there should be the law of diminishing returns for any strategy - so once you reach a certain threshold for a certain mechanism of scoring, from there in you don't get full credit, encouraging play that is flexible and dynamic.

But that's UT3 - for other games it won't be as straightforward.

Also, don't forget the consoles.

[BattleMode]

Some design stuff.

Ordered a server for this stuff now. Will try 1TB+ storage, Quad Core, 3 GB RAM, Linux, 100mbit.

The raw collected data will be HUGE. With TheCLQ I used to delete the raw data, but perhaps I can do different now.

There will be about 100.000 live game servers in the database of which about 10.000 will be active (have 2 or more players).

I intend to use differential scanning (TheCLQ scanned all servers once per minute, that's overkill):
Class 1: once per minute, top 1000 (?) servers - 1000 servers per minute
Class 2: once per 3 minutes, top 9000 (?) servers - 1000 servers per minute
Class 3: once per 10 minutes, all 100.000+ servers - 10.000 servers per minute
Class 4: once per hour, servers that have been down for over am hour
Class 5: once per day, servers that have been down for over a day

This all adds up to about 10.000 servers per minute (170 per second). I used to do much bigger amounts so that should be possible. Once I have somethign working I can tune these numbers.

Stats to be kept per server (besides stuff like name etc.):

• Last successful scan timestamp
• Downtime (missed scans)
• Statistics for number of players (last hour, day, week, month, etc.)
• Normalized to minutes, PPM per player per week in slots (float)
• Number of played minutes (detected) per player per week
• Number of collected points per player per week
• End game rankings (can only be approached)
• Some form of ELO (not sure yet how to achieve that TBH, since it is not incremental)

One significant difference I have in mind compared to what I did 10 years ago is live data. The AllGameServers.com site should get live data (e.g. who is playing where NOW), TheCLQ.com site should get the rankings etc. Preferably you should be able to enter a list of your favorite players, store that and get an overview of where they are playing on a special personal web page. Where technically possible you should then be able to launch the game directly from the browser.

TheCLQ used to differentiate between gametypes. Not sure yet how to handle that this time. For the player stats I will probably focus on per server statistics this time (global stats are crap anyway, although I might do them anyway) so gametype might be less relevant.

[BattleMode]

Map related statistics per server:

• Minutes played
• Average number of players on the map
• Player increase / decrease statistics

E.g. these stats should allow a server owner to determine which maps to dump because they scare away players.

[BattleMode]

Did some experimenting with memcached, very interesting stuff. http://www.danga.com/memcached

[BattleMode]

Empty site is up now: http://allgameservers.com
CLQ site is also up, but I will not replace that until I have something to show: http://theclq.com (200.673.123 pageviews so far)

I have ordered a server and will use the BBS servers to build things while waiting for that. I will probably not start large scale scanning until the server arrives.

One thing I am not certain about yet is how I am going to do data processing. In the past I use a C application for that since C is very very fast. But in 10 years CPUs have become much faster so I am considering to use PHP now. Disadvantage being that it is slower, advantage being that it takes less time to build things. I might just start with PHP, see where it breaks and then replace the broken parts with C applications.

[BattleMode]

Moved TheCLQ site to a different server, changed the DNS settings and replaced "May the fragging go on forever." with "We WILL be back... ". DNS changes are slow so it will be a while before these changes are visible.