While player skill varies a lot between classes it doesn't help the balancing capabilities of the algorithm or make much difference individually to separate players by game type (6v6 or HL). The correlation between the ratings of players who play a lot of both game types is very high.

I was surprised by this, I thought it would make a big difference to delineate along those lines but apparently soldier is soldier and scout is scout whatever game type you play.

How about Australian Seagull?

For obvious reasons the precise details are going to remain private, and will be determined by statistical study. Winning and losing are also stats, but how it's all blended together are the trade secrets.

This depends on the goals, the highest level mix groups have players trolling in them so it's a problem at any level, but only if you're after an idealistically pure rating. I wouldn't pretend that this is what will be produced here. It will definitely be a warts and all rating for any players who don't play seriously, but most players in that position probably aren't that bothered.

TF2 Center and TF2 Pickup won't be mixed though, so there won't be any leakage between differing behaviours. The ratings for players who play the same classes a lot on the two services actually correlate pretty well, about 75% similarity, but that's not close enough to combine them in my judgement.

It'll be logs.tf based at the start at least so as International as TF2 Center gets. Anything coming later depends on being able to get the right information out of those services. The main difficulty is identifying populations that are good for it, just throwing everything in doesn't work well.

It varies depending on the service being measured. TF2 Pickup has about 800 rated soldier entries as it's linked to ETF2L and it's Highlander pickups haven't taken off, and TF2Center has around 7300 with a much broader player base.

My instinct is to be incredibly tight with the top few ranks, just fixed numbers so they are a very small group regardless of population size, but more evenly distributed after that.

There's some great stuff so far :D

Now this is good thinking

I'm begging for help with a new website. Some of you may know that I've worked on creating a rating system for TF2 balancing and ranking as detailed here, and despite my titanic efforts implementation in lobby services is basically nil and not coming any time soon, so because zoob is so godlike and makes all the data necessary available to everyone I'm going to do it myself, and maybe if it's available as a web service it might get used.

But I don't really want to publish precise ranking figures because they expose the system more to being gamed, like CS I'd rather hide these behind rank names like these:

http://www.pro-hl.com/images/prohl/csgo_ranks.jpg

And because I have no imagination I'm throwing it open to you. What rank names do you want to see? Do we just need a tier number instead of a name?

This should be ready in a matter of weeks so if you want to be at the top of the ranks you need to get your try-hard hats on and start crushing faces at TF2Center or TF2Pickup.

+1
Unlock abuse for the win, not just the shitfest

It was started, then no one did anything on it for a month or two, then it was finished. Then the match making leaks came out and there was panic, then it was finished again. About a week actual effort maybe? Lange is being a bit hyperbolic cos he's not happy. It happens, he's still The Man.

I have a degree in software engineering, lots of experience and a job, so you can take it from me that it's possible if they want to do it and they have architected the system with making it possible in mind. Only an in depth understanding of the differences between the two engines could give genuine insight into how hard it is or what the result would be, and that's something nobody here has. The benefits are obvious - the game stays up to date, latest shaders, better performance, longevity, etc. People forget TF2 is the first shooter (that's not an MMORPG) the publisher wants to live forever because of the hats.

On topic: whether it's a good idea for Valve to do something along these lines is only for them to decide, but the fact is they're doing something towards some kind of match making and the comp community would be remiss to not put ourselves in position to collaborate/offer our cooperation with that if it's possible. Whether it's a new game mode, something dumb or something we wouldn't recognise there's something appearing in the source code and it is not MvM.

However if you want a TF2 MMR without Valve Time I'll probably have something up in a few weeks ;)

They got you mixed up with the Wireplay guy

If not a backlash (which I think is a valid concern) then also the additional risks of it being pushed off course, dilution of the message and loss of control of the agenda were concerns. Once it's public property then it's also fair game for everyone which would cause significant problems maintaining the focus of the project.

I don't think publishing is out of the question forever (particularly if met with apathy by Valve), I would say it's just a decision taken at a certain time for reasons of expediency to try to ensure a consistent message and focus.

The reaction from various people who signed or didn't want to sign was presumably enough to convince the decision makers otherwise.

Fear of a community backlash.

Yeah, it's actions that count. It's probably better if they just get on with it - either they think it has value or not and if it happens all the better. Although I've just had an evil idea, maybe they wanted to sell bans to monetise the format :D