Forum

TF2 Player Rankings website - e-peens at the ready

Created 10th January 2015 @ 23:38

Add A Reply Pages: « Previous 1 ... 8 9 10 ... 14 Next »

Selek

Dr. med.

In my opinion it’s a very interesting project, I have waited for something like this since the use of logs was established. Let’s just hope it doesn’t go the way of metacritic, where a single number has way too much weight on the assessment of quality/performance.

For instance, I noticed the Prem/Div 1 Spies are ranked all over the place. Spy is arguably one of the hardest classes to asses using numbers, though. And it might be for other reasons like recent inactivity.

RTC

Quoted from Kaneco

[…]
Which is completely awful since you can stack any lobby u want as long as you’re the lobby leader and/or bring some friends.

Which is exactly why I don’t think it’s best suited for TF2Center at this current moment in time. Advanced Lobbies have little set in stone at this moment in time, so there’s a chance a randomized system similar to that used on TF2Pickup may be used for advanced lobbies in order to ensure more balanced games.

Gentleman Jon

Quoted from Selek

In my opinion it’s a very interesting project, I have waited for something like this since the use of logs was established. Let’s just hope it doesn’t go the way of metacritic, where a single number has way too much weight on the assessment of quality/performance.

For instance, I noticed the Prem/Div 1 Spies are ranked all over the place. Spy is arguably one of the hardest classes to asses using numbers, though. And it might be for other reasons like recent inactivity.

The current rating system uses several metrics (which vary by class) and is probably going to head into the direction of using more rather than less as long as it doesn’t compromise accuracy, as this makes it far harder to game.

Variation in player ranking is down to a lot of factors, not playing particularly seriously being a very large one. There are a handful of players I’m in contact with who (having seen their lowly spot) have been keening it up to maximise their rankings over the last few days with considerable success.

Anyone who’s not happy with where they are should try stringing a handful of lobbies together where they determinedly kick ass and see where they end up. One thing I might increase is the strength of the inactivity decay effect.

Also bear in mind that it’s only in beta and making it public is an important part of the development process, I’ve received several bits of grumpy feedback that have already proved useful in planning new improvements.

Kaneco

hey Jon, just wanna clear up some stuff…

Does this system take in accounts win/lose essentially? What I mean is, is it like in dota, you would never lose points when winning a game, and you would never win points when losing a game? Or at least very based in that is very difficult to lose points if you win or vice-versa.

For example when you have the roamer who is the designated “sacrificial lamb” you are sacrificing your own stats for the team, even if you dont get a pick, sometimes you get a force which afaik is not detectable through the logs.

What I mean is that it would be very easy for some classes or team roles to be excessively penalized when they work towards a team objective, and if it doesn’t work that way, the ranking (if made more popular) would only encourage players to play for the stats, and not with a team objective in mind.

So, my point is, ultimately I think win/lose should be the most relevant stat to have in consideration for any evaluation of performance of a player. Of course if you have a player with 50 kills or a demo with 15k dmg you should have that in account and that should be reflected positively in the performance evaluation but it still shouldn’t take the merit of team cooperation of a team based game.

Gentleman Jon

Quoted from Kaneco

hey Jon, just wanna clear up some stuff…

I don’t want to get too far into specifics, there are problems with applying win/loss in TF2 without looking at individual performance, and at the same time there are ways of winning that don’t always produce great stats.

I’ve tried, and continue to refine attempts, to ensure players who win ugly aren’t unduly punished but at the same time I’m restricted by the original goals which are to identify and compensate for blatant imbalance in lobbies, so anything that makes those attempts less effective I can’t use, even if it seems fairer.

Sorry it’s a bit wishy washy, I used a very similar system to look at player performances in Season 19 and Rising and T-Mac routinely outscored several pocket soldiers from other teams, although I never made the direct comparison between the roles, so things aren’t stacked entirely against the roaming soldiers.

Kaneco

Quoted from Gentleman Jon

[…]
I don’t want to get too far into specifics, there are problems with applying win/loss in TF2 without looking at individual performance, and at the same time there are ways of winning that don’t always produce great stats.

I’ve tried, and continue to refine attempts, to ensure players who win ugly aren’t unduly punished but at the same time I’m restricted by the original goals which are to identify and compensate for blatant imbalance in lobbies, so anything that makes those attempts less effective I can’t use, even if it seems fairer.

Sorry it’s a bit wishy washy, I used a very similar system to look at player performances in Season 19 and Rising and T-Mac routinely outscored several pocket soldiers from other teams, although I never made the direct comparison between the roles, so things aren’t stacked entirely against the roaming soldiers.

I understand, but just taking in account a pickup I played recently (http://logs.tf/624659?highlight=76561197991320838), the enemy team did 3K more damage globally, they had 20 more kills, they had the top 2 fraggers and damage dealers, and they still lost 5-3, because our last pushes were much cleaner (as in we would cap fairly fast without many deaths, and our overal pushes as well, where as their pushes where much sloppier and envolved much more teamfight/deaths, still in the end what I’m saying is that rounds is what counts most, after all thats what would win you a season in etf2l, not topfragging every match.

I kinda get the gist of what you’re saying, this is originally a concept for player ranking not team ranking, and you dont wanna deviate much from that, but I still think in the end, the game result should be a big factor to consider, no matter what statistical representation your performance has on the logs, as it can be deceiving in some cases such as the one stated above.

Gentleman Jon

Quoted from Kaneco

I kinda get the gist of what you’re saying, this is originally a concept for player ranking not team ranking, and you dont wanna deviate much from that, but I still think in the end, the game result should be a big factor to consider, no matter what statistical representation your performance has on the logs, as it can be deceiving in some cases such as the one stated above.

I’m not telling you it isn’t taken into account, I’m also not telling you it is taken into account, I’m not going to give away anything that let’s people game things more easily. My point about the quality of the rating doesn’t mean it has to be biased excessively towards individuals, but changes have to adhere to specific measured goals so team measurements are fine as long as they don’t cause a problem with that. I personally am neutral within the bounds of those measures.

I don’t think I’m punishing players who play non-pocketed roles unfairly, but drawing the line at when a roamer is being sacrificial and when they are being shit is not easily done, ask 10 players and you’ll get 10 different answers, so I think the mathematical guides I’m using are the most neutral measures I’ll get. If the question is whether I experiment with measures that should help players who don’t chase stats to win, the answer is yes.

I’m aware that in an ideal world I’d be able to publish it and no one would be able to game it anyway, but that’s not the current situation.


Last edited by Gentleman Jon,

fraac

JOHN
CENATION

You can value wins higher if you’re more confident that the teams aren’t stacked. I would measure clumpyness of players playing together historically, but I think Jon would consider that too much of an ugly kludge.

I really like these ratings though. It says I’m the best 150dpm lobby scout in the world, and I’m most proud of games like this, so that fits.

Selek

Dr. med.

Quoted from Gentleman Jon

[reply to my post]

All fair points, if you are aware of this all is good.
I’m not particularly surprised about my low ranking either, I haven’t played many lobbies in the last months anyway. I hope you didn’t count me to the grumpy feedback people, that was not my intention :D

Looking forward to how it turns out, great job so far!

Gentleman Jon

Quoted from Selek

All fair points, if you are aware of this all is good.
I’m not particularly surprised about my low ranking either, I haven’t played many lobbies in the last months anyway. I hope you didn’t count me to the grumpy feedback people, that was not my intention :D

Looking forward to how it turns out, great job so far!

Not at all, I just don’t want people to think that I discount negative feedback just because it’s negative. Getting specific feedback from lots of pairs of eyes is important to making improvements because it let’s me focus on particular cases that might have problems. I’ve already had some important bugs and weaknesses pointed out, I’m a bit snowed under with work at the moment so I can’t respond as quickly as I’d like to it all though.

Gentleman Jon

UPDATE:
You can now login via steam and update your name! It’s a bit of a rush job but it should help anyone who’s not happy with what they’re currently called on the site. There will probably be a bit of lag between the update and your name changing in the ranking lists. It seems to be working for people but if you get any problems let me know.

http://beta.tf2playerrankings.com/

Gentleman Jon

I’ve addressed the Heals/PM and DPM problem to some extent. It can still appear a bit odd in some cases where the sample size is small and a player was involved in particularly aggressive or passive logs.

Kaneco

I’m really waiting for the feature where you can track your rank progress over time, that would be neat

Gentleman Jon

Quoted from Kaneco

I’m really waiting for the feature where you can track your rank progress over time, that would be neat

A more detailed profile page is in the pipeline that will fill out some of that.

I’m working on some friends list integrations at the moment.

Gentleman Jon

A few minor but useful site updates have been made. If you’re logged in and your profile isn’t private then your friends list will be displayed on the right to make it easier for you to see how your steam friends are doing.

Under the player search recent popular searches are now displayed.

Finally in the player profile window a player’s position in the list can be clicked to show the context of their ranking, and who is currently around them. You can also use urls generated from these links to link straight to a specific results page. For example

http://beta.tf2playerrankings.com/#/site/1/class/3/page/3

Links to page 4 of the scouts on TF2 Pickup without having to navigate to it first.

Add A Reply Pages: « Previous 1 ... 8 9 10 ... 14 Next »