Transparency in the ranked system

[sam.louro.monteiro]

the 2nd and 3rd examples point towards some true skill system with heavy weight on individual performance, which enables people that are essentially smurfing to get jumped to the highest ranks without ever being tested against real opposition.

Imo this is much more similar to a glicko rating system. while the game uses your normals mmr as a base, like in dota2, league and various othergames (even wc3champions addon) when you enter the ranked ladder you enter with a very high uncertainty so variance in rating gained and lost is much higher because it doesnt know yet where you stand.

I have only found this thread now and the links you proposed are not from eternus players and you see how it fluctated. They did try some true skill approach (theysaid that individual character mmr's) but probably stopped as it created chasms in mm when i play my 5th game of haze but have a total 200games already right?

So with each win even if the system doesnt create games that did challenge him as much as possible, him winning created a very high gain in mmr which would then come down as he would be poorly match made and the system would correct itself.

Obviously at the start of the ranked system there will be disparity because there is low sample size. They might try different systems and works around as they are still cooking with everything in the game, but having someone skyrocket in the first week is fine they will go down if they do not bellow to the skill the rank expects with time (again from the links you post you see how everyone has gone down a bit, not much because they probably have decent skill to have gotten up there in the 1st place). Glicko is not perfect and there are odd cases where the system needs to be manually adjusted to create a good laddering experience (as an example what happened to Direct strike in wc3).


As everyone's bar starts at this, the 30% dota asks you is the equivalent in deadlock of the 7games to give you a rank where they think they know where you stand, then with mroe games it corrects itself. Should it need mroe games to tell? i don't think so, your games will be the same if it tells you or not, its just for the other's people's ego that it matters, and if its 10games does it change that much? 20? this is a test build, i don't get 20 games a week so how would i fair in this system? There has to be a low limit where you see result from your wr instead of being obscured (hehe).


TL;DR : don't take low sample size as pure evidence of a system working or not

 

[le-sky]

To be fair, in CSGO Valve didn't use straight Elo but used a modified Glicko-2 rating system which worked fine.

nevertheless in CSGO kills are more important than in DL plus it has less TTK

 

[offsetcyan]

nevertheless in CSGO kills are more important than in DL plus it has less TTK

We have no data on what's more important for ranking in CS:GO or CS2. Just that wins = up and losses = down.

 

[cdas]

They really just need to be more transparent in how they assign their ranks and hidden MMR system for players as transparency will NEVER be bad in a game that is this early in development, especially when its a MOBA. Transparency will actually help developers in making their MMR system and matchmaking even BETTER than most games further optimizing and placing the proper players in their proper respective skill groups instead of the mess that we have now.

I would say the previous matchmaking system was absolutely fine for me and my group, sure we will have the occasional new player or someone who just doesnt know what he was doing, but it is a MOBA if you have played any MOBAs before in unranked, this is bound to happen at some point in time. I also have a feeling that games arent being tracked properly with the new system especially in unranked, there is no way that our group still suffers from wide skill range even after 10 or 11 consecutive game wins.

There is realistically no point in putting stats for ranks over win/loss as shown in the few examples given by OP at the start of this thread. I can be honestly useless the whole match, not helping team and just perma farming having 50 - 70k souls at the end of the game and still lose while the team are permanently in a disadvantage when trying to go for objectives or contesting them in a team fight with the enemy, 80% of the time will result in them losing the engagement with it being a 5v6.

 

[noobdoomguy8658]

I don't really get why transparency matters *now*. We're basically playtesting the game for Valve and are expected to generate data and provide verbal (written) feedback - we're not analysts, and providing even more data to process in form of our opinions on the systems and numbers in place that will not stay the same for as long as a couple of weeks, it just sounds like creating some extra work that will not lead anyone anywhere.

Neither the ranks nor WR nor your performance mean much at this point because it's all going to change drastically. It's there to let the developers test the stuff on real players, in real scenarios, not to provide some definitive experience and showcase anything.

As for measuring "true skill" - the numbers won't do that. If you have horrible stats, taking the most damage in your team, while dealing the least, but you're the one distracting the enemy team with your tankiness and your annoying presence and your non-measured ability to divert the opponent's attention, which ultimately leads to your team's opportunities to take objectives, execute kills, and win games, then you're doing some nice plays; individual stats won't reflect that, but a victory might.

The algorithms lack the means to analyze matches in their full complexity, because the outcome depends on way more than just numbers and stats.

 

[cdas]

Transparency matters now solely because we are playtesting the game, we are giving Valve feedback, valuable or not, whether they take it or consider it is up to them to decide. Having transparency allows playtesters to not only help analyze how the matchmaking works but also helps other external developers in developing tracking software using their api think deadlocktracker and tracklock.

"True skill" will never be properly measured in any game with our current technology especially not in a MOBA, what people are asking for is a way to "justify" receiving the rank they get and the rank that other players get. There is absolutely no way for games to be fully "balanced" with players of the same skill group which is why matchmaking always say players of similar skill group.

Right now there is a problem where people are confused with how the current score is being calculated and some are being locked out of having fun and playing with their friends solely because of the wide skill matchmaking system. People who are shooting way below their skill group getting higher ranks than expected and both teammates and player are suffering from it, the player is getting trashed 24/7 in game and their teammates having a "burden" on their team. Transparency will really help reduce this shit from happening, not saying it will be 100% foolproof, outliers can and will happen, but right now to get people on board and generating more data, matchmaking/score/rank calculations have to be both transparent and not hinder players

 

[ASX3]

Transparency will really help reduce this shit from happening, not saying it will be 100% foolproof, outliers can and will happen, but right now to get people on board and generating more data, matchmaking/score/rank calculations have to be both transparent and not hinder players

And what exactly would people do with this information? Playtesters aren't statisticians. Our jobs are to just play the game, and report whether it was a good or bad match. The overwhelming number of us are simply here to be a part of aggregate data. Take a look at the bug report threads, how many of them are high quality reports? People don't even understand what constitutes a proper bug, or what kind of info to include in the report.

Valve likely has actual trained professionals in charge of this. If they wanted players to be aware of MMR calculations, or give specific feedback on it, they would request it. Just trust the process man.

 

[gimmekeyplskthnx]

Ranked is seeded off of normals performance, and normals used to have zero restriction on 6 stacks fighting against solos. A lot of the low wr players in very high rankings tend to either be too low sample size or stack abusers.