344 replies to this topic

[Mother Tereza]

Unfortunately, you have no idea how complicated this objective is. Fortunately, TF2 community has some math geniuses who had succeed in development of this system. Hopefully, we will integrate Jon's system into TF2C in a near-ish future.

 

please for the love of benji... you can't base a matchmaking algorithm on some obscure thing some guy did for himself and nobody else seen what's inside that black box.

 

First of all, Jon has made this for the community benefit. It might look obscure for the most of users, but not for the devs who involved into this project. His system works. Yes, I agree, it is not ideal and has some flaws and I personally find some of them a quite annoying. But we don't have a better alternative, do we? And Jon could improve its algorithm in a future, especially if TF2C would be providing him the stats from advanced lobbies.

 

The goal of advanced lobbies is providing the best of possible competitive experience. If we will succeed in this, why players should care about any algorithms behind this system?

[Kengur]

 

Unfortunately, you have no idea how complicated this objective is. Fortunately, TF2 community has some math geniuses who had succeed in development of this system. Hopefully, we will integrate Jon's system into TF2C in a near-ish future.

 

please for the love of benji... you can't base a matchmaking algorithm on some obscure thing some guy did for himself and nobody else seen what's inside that black box.

 

First of all, Jon has made this for the community benefit. It might look obscure for the most of users, but not for the devs who involved into this project. His system works. Yes, I agree, it is not ideal and has some flaws and I personally find some of them a quite annoying. But we don't have a better alternative, do we? And Jon could improve its algorithm in a future, especially if TF2C would be providing him the stats from advanced lobbies.

 

The goal of advanced lobbies is providing the best of possible competitive experience. If we will succeed in this, why players should care about any algorithms behind this system?

 

If nobody cares about the algorithm then why not set a goblin with a bag of dice? At least everyone would be aware how that one works.

[Mother Tereza]

If nobody cares about the algorithm then why not set a goblin with a bag of dice?

 

Shhh, don't tell anyone about our trade secret!

[MasterNoob]

In general, it works as follows: as long as you win all your matches/lobbies, you will end up at the top... Details will probably distract you anyway...

[siouxsie]

I completely disagree with the "Casual lobbies" name, even calling them "Standard lobbies" doesn't sound good to me. Have "Advanced lobbies", and "Lobbies", elevating the prestige of the dedicated players' games with a fancy name is fine but retitling normal lobbies as "casual", "basic", or "standard" demotes them for no useful purpose and only servers to segregate the playerbase. You're adding a new feature, not taking anything away, there is no need to demote the status of the current system with a new title.

[GentlemanJon]

 

 

How do you rate people?

Needs to be fair, not just dm or assists. Because sometimes it's just about having the right pick at the right moment. How do you rate that?

 

Well our first step would possibly consider how each gains or loses leader board points Unlike my example, Starcraft 2, it will not be as easy for us to find a way to rank people. One thing we should do is even out how many classes gain how many LB (leaderboard) points per points scored ingame. lets say Class 1 scores an average 40 points per game, and Class 2 scores 80 points per game, Class 1 gets one LB point for every 20 points they score and Class 2 gets 1 LB point for every 40 points they score. This makes sure that no class is hoarded to gain points. we then just compare how many points said player has and raise or lower them accordingly, like a leaderboard.

 

Unfortunately, you have no idea how complicated this objective is. Fortunately, TF2 community has some math geniuses who had succeed in development of this system. Hopefully, we will integrate Jon's system into TF2C in a near-ish future.

 

please for the love of benji... you can't base a matchmaking algorithm on some obscure thing some guy did for himself and nobody else seen what's inside that black box.

 

 

Masternoob has seen exactly what is inside the black box, as can anybody who wants any input I can give to help their lobby/pickup system. It's up to them whether they want to make use of it or not. I'm sorry it bothers you that you aren't in a position to asses it's efficacy but that's not really relevant here.

[MasterNoob]

I'm not sure I would be wanting to expose the internals of such system, just to avoid people try to play the system, rather than play the game properly. Eg. if i could bump my medic ranking by not dying, I would just stay in spawn to avoid that, right?

[GentlemanJon]

No, it's obvious why it shouldn't be public, not just because people might try to game the system but also because the details would be misinterpreted by many players and people would play what they think are system gaming strategies but are in fact just screwing their team as well as themselves. The best way to get a good ranking is to play your best and win games. No big secret.

[Kaelan Frey]

I'm not sure I would be wanting to expose the internals of such system, just to avoid people try to play the system, rather than play the game properly. Eg. if i could bump my medic ranking by not dying, I would just stay in spawn to avoid that, right?

Medic performance shouldn't be judged by KAD anyway. You can still do 1000 HPM and die more than 20 times in a 30 minute game - it's a bit hard but entirely possible.

 

Not exactly hard not to die as Medic either way, since that in the majority of your matches as Medic you're the class who dies the least. If you aren't, then you either have terrible flanks or a terrible pocket soldier/scout or terrible positioning from your part.

 

Either way, yeah the last thing you want to do is expose the internals of the ranking system.

[elliott]

if your ranking system is actually accurate then no one would be able to game it without raping the other team anyway :\

 

like how do you farm a system where the only way to be a positive player involves killing the enemy team or healing your own? Like unless some unfair weighting goes to healing or something, there'd really be no way to game it out.

[GentlemanJon]

if your ranking system is actually accurate then no one would be able to game it without raping the other team anyway :\

 

like how do you farm a system where the only way to be a positive player involves killing the enemy team or healing your own? Like unless some unfair weighting goes to healing or something, there'd really be no way to game it out.

 

Accuracy and susceptibility to gaming are very different things. I've tested player reaction to the various parts of the ranking model and it is almost always misinterpreted, ironically particularly by people who care about it. I have no doubt it would cause sub-optimal behaviour and that this would at least partly be due to misinterpretation or a lack of real interest in studying the numbers carefully.

 

I'm confident that they key parts of the ranking match solid gameplay whether you are a strong or a weak player, objectively encouraging serious and appropriate play. I'm sure Valve's various ranking systems do a good of that too, but they don't publish theirs either. The problems both with Valve's and with mine come when players on the same team try to get advantage over each other and damage everybody's chances while they do it.

 

I would like to be able to publish it because I think it's interesting and the potential to get new ideas would be valuable, but as I've learned over time people won't actually read the boring bits explaining it, they'll read the juicy bits and decide "it's just this" (the most common phrase amongst the various TF2 brains I know) and proceed along a totally incorrect and grossly simplified premise. This myth will immediately become the truth, particularly when an influential player starts to espouse it.

 

This may sound elitist but really I have to consider the worst case and my anecdotal experience is that it would be bad news, I'd have to invest a huge amount of time in fire fighting questionable opinions and most of that wouldn't be taking place in the context of academic debate, a large percentage would be flaming and concerted efforts to discredit it/me. Not something I really want to do.

[Foxy]

 

This may sound elitist but really I have to consider the worst case and my anecdotal experience is that it would be bad news, 

 

 

Out of interest...why? Gaming the system may happen, yes. But it will happen by a handful of players, and result in those few players having either a deceptively high or deceptively low rank. There are no prizes for top or bottom of the ranking system. If players are throwing games by just gaming the system, let them be found out in different ways (reporting, noticing quick switches of rankings).

[GentlemanJon]

Out of interest...why? Gaming the system may happen, yes. But it will happen by a handful of players, and result in those few players having either a deceptively high or deceptively low rank. There are no prizes for top or bottom of the ranking system. If players are throwing games by just gaming the system, let them be found out in different ways (reporting, noticing quick switches of rankings).

 

 

Apart from everything I said in the post above I just don't think it would be limited to a few players. As soon as you give people something to play for a considerable number will try to discover shortcuts, optimisations and ways to get what they want at other player's expense - that's just human nature (not everyone, but a significant slice). Making it an offence you have to police would lead to a nightmarish grey area for admins as well with a huge workload. Every time a player did something another player didn't like that could be remotely interpreted as being done because it's in their favour rather than the appropriate gameplay decision you'd have cases made against them.

[lafc]

 I'm sorry it bothers you that you aren't in a position to asses it's efficacy but that's not really relevant here.

 

actually everybody is in the postion, since a ranking with a secret formular is always a useless ranking.

[sad-panda]

I have a few issues.

 

Just pairing the higher skilled players together based on stats / ranking / div and other such things isn't magically going to make for more balanced lobbies. The skill difference between the tiers is huge. It also comes down to players attitudes as well. I am worried by having advanced lobbies your gonna split the community and player base in 2.

 

I play pick up now and again and its very rare on there you get any sort of close game even though most people who play pick ups have decent comp experience

If you get 2 or 3 top players playing their best classes and trying hard  on one team they can easily destroy or  roll teams. And if you get 2 or 3 players on your team who aren't trying hard or off classing or running silly load outs it can totally fuck your team up.

 

http://logs.tf/82980...561197976362532

There is a great example.

 

I actually like lobbies more then pickups. They fill quicker... you can get games most times of the day... you can play and have fun without having to try as hard and you can play with friends without hassle. Yes you do get some rolls and one sided games but you also get some crazy, fun and close games too.

 

Its a casual version of comp which is why a lot of people like it. I always seen lobbies as the next step from public servers.

 

Advanced lobbies will only work if the better players take them seriously otherwise whats the point in having them. I thought that was the whole idea behind pickups was to get a higher skill level of players together to play more serious games but 90% of the ones i have played in have been a joke.

Edited by sad-panda, 22 May 2015 - 02:04 PM.

[GentlemanJon]

actually everybody is in the postion, since a ranking with a secret formular is always a useless ranking.

 

 

So the vast majority of online game ranking systems are useless. Good to know. As I pointed out the people who make decisions have seen everything.

Just pairing the higher skilled players together based on stats / ranking / div and other such things isn't magically going to make for more balanced lobbies.

 

I haven't promised any magic in this area but as Center has no balancing at all then it will bring an improvement, but it won't turn every game (or anything like every game) into a close game. What it should do is ensure that very low skill or low experience players (as well as medium and high skill) are generally evenly spread out even if they don't have any competitive record to show it, and players who don't take things seriously will also be poorly ranked because they will do less well generally and the ranking system records that.

 

You're right that a huge amount depends on individual attitude and approach right in the moment when the game is going on which is why no promises can be made, but if skill level delineation is used then it will at least ensure that the vast majority of players you play with are, if not the same understanding or skill at the game as you, in the same ballpark which is rarely true in open lobbies.

 

It's worth remembering that even at the highest level when teams are very evenly matched over a best of 3, the overall game can be close but individual maps can easily go 5-0 as well as golden cap. There's no way I've found to guarantee close matches and I find it unlikely one exists.

Edited by GentlemanJon, 22 May 2015 - 03:16 PM.

[lafc]

 

So the vast majority of online game ranking systems are useless. Good to know.

yep, although there are not that many that do this.

 

 

 

As I pointed out the people who make decisions have seen everything.

you reason the accuracy of the ranking, because its approved by the tf2c admins?

[GentlemanJon]

 

you reason the accuracy of the ranking, because its approved by the tf2c admins?

 

 

No, you said everybody is in this position and they're not. My early methodology is explained here http://www.vanillatf...ancing-a-lobby/, although I've actually introduced some more mathematically rigorous techniques since then, particularly modelling techniques such as linear regression. If you have specific criticisms then feel free to make them, I'm aware of several shortcomings in the method outlined in that article that have since been corrected.

 

Although the assertions made about Valve obviously don't hold much credence now

Edited by GentlemanJon, 22 May 2015 - 08:22 PM.

[lafc]

 

No, you said everybody is in this position and they're not. 

if someone doesnt know how the ranking works, you might as well roll a dice for the ranks. the person doesn't get less information from that.

 

My early methodology is explained here http://www.vanillatf...ancing-a-lobby/, although I've actually introduced some more mathematically rigorous techniques since then, particularly modelling techniques such as linear regression.

mathematic techniques doesnt mean anything, if you use don't the right factors from the game.

 

If you have specific criticisms then feel free to make them, I'm aware of several shortcomings in the method outlined in that article that have since been corrected.

how can i (someone) know whats wrong with it, if i don't  even know how it works?

[GentlemanJon]

 

 

No, you said everybody is in this position and they're not. 

if someone doesnt know how the ranking works, you might as well roll a dice for the ranks. the person doesn't get less information from that.

 

My early methodology is explained here http://www.vanillatf...ancing-a-lobby/, although I've actually introduced some more mathematically rigorous techniques since then, particularly modelling techniques such as linear regression.

mathematic techniques doesnt mean anything, if you use don't the right factors from the game.

 

If you have specific criticisms then feel free to make them, I'm aware of several shortcomings in the method outlined in that article that have since been corrected.

how can i (someone) know whats wrong with it, if i don't  even know how it works?

 

 

Foxy, this is the level of discourse I'm talking about. Not even trying to engage with the methodology (which you don't need any detail to criticise), not engaging with the results which aren't obfuscated in any way, just one line answers sticking to a single point because apparently they're entitled to a full explanation. The only possible use for a ranking is for them to know how it works out what it says about them. It couldn't possibly be used anonymously to alter the environment they play in for the better, it doesn't even cross lafc's mind that there is a potential benefit there - it's just useless (despite there being a several thousand word long article pointing out the benefits). At the risk of speculating to the negative, there's a very good chance he's exactly the kind of player who would ruin his team's chances thinking he's gaming the system when he's just making things worse for everyone including himself, because as far as I can tell he's completely centred on himself.

 

Here's a hint lafc - the ranking contains zero information about how to get better, improve your play or otherwise be classified as a better player. Even if you knew everything about it then for an individual player it's still "useless". Improving your death matching, game sense and team work are what make you a better player. Work on those and win more games and you move up.

 

I could waste any amount of time with these kind of exchanges, and obviously I'm not going to.

Edited by GentlemanJon, 23 May 2015 - 09:16 AM.