Transparency in the ranked system

It's not "transparent" because they're still experimenting with it, so it's pointless to announce how they have it because it'll just be changed again anyways.

yeah this is true. If the devs know that the report system is bottom tier why even discuss it? There is nothing positive to say about it so just ignore it and bleed players. If they confess that they cant do an anti cheat or good report system they will bleed even more players.
 
why would it be pointless? If we dont know how the system works its impossible for us to give feedback on it.

I dont see how transparency could ever be bad. Knowing how the system works could address alot of peoples perceived issues and complaints with the system. Or it could help with proper feedback that could lead to fixing them.
1730557162910.png

A good return of information is the win rate within each rank. When they are similar it means that the pairing is working well. If they are discrepant then it is not working and people with very different performance are in the same classification.

This may take a while. It may even never happen. A good quick and easy meter to do would be per character. In victory or defeat each character could easily be compared.
It would help to understand how everyone is using each doll.
Something like this site is doing: https://deadlocktracker.gg/heroes/abrams
However, it doesn't work with the entire server in order to only view the matches that we can watch within the game.

I can compare my numbers on souls, KD, player damage. This way I can really know if I am above average but it is not classified in rankings.
It's the best I can get back at the moment.
 
why would it be pointless? If we dont know how the system works its impossible for us to give feedback on it.

I dont see how transparency could ever be bad. Knowing how the system works could address alot of peoples perceived issues and complaints with the system. Or it could help with proper feedback that could lead to fixing them.
there are different methodologies for testing stuff, some of those are blind methods which means that the testers knowing too much or too much information on how it works could mess with the data, ex: knowing the variables people could start heavy gaming the system and skewing the data.
then again, i'm not a valve dev so this is all conjecture 🤷
 
why would it be pointless? If we dont know how the system works its impossible for us to give feedback on it.

I dont see how transparency could ever be bad. Knowing how the system works could address alot of peoples perceived issues and complaints with the system. Or it could help with proper feedback that could lead to fixing them.

It's "bad" for the same reason why Valve doesn't want hero feedback to be public. Because people will immediately start trying to min-max it and skew the results.

It's the same reason why we have blind experiments in science, where you have a control group that is given a placebo, and neither the experimental group or the control group know what they were given.

You don't need to know how the technical details of matchmaking works for you to give feedback on your quality of matches, which, surprise surprise, what the pinned thread in this forum is exactly for.
 
It's "bad" for the same reason why Valve doesn't want hero feedback to be public. Because people will immediately start trying to min-max it and skew the results.

It's the same reason why we have blind experiments in science, where you have a control group that is given a placebo, and neither the experimental group or the control group know what they were given.

You don't need to know how the technical details of matchmaking works for you to give feedback on your quality of matches, which, surprise surprise, what the pinned thread in this forum is exactly for.
if the system is gameable its already flawed.
 
Still funny how I uprank when I lost 5 games and win 2 in ranked last week, but I always win the laning stage, maybe the laning stage stats is also important in determining the rank.

Who knows, I don't know, maybe even valve don't know how the rank system works.

But what I know why there's no transparency because valve don't want people to meta gaming the rank while this is still experimenting phase of the game, we are their lab rats still.
 
Still funny how I uprank when I lost 5 games and win 2 in ranked last week, but I always win the laning stage, maybe the laning stage stats is also important in determining the rank.

Who knows, I don't know, maybe even valve don't know how the rank system works.

But what I know why there's no transparency because valve don't want people to meta gaming the rank while this is still experimenting phase of the game, we are their lab rats still.
If stats are a part of the calculation then it will be eventually found out from data (which many sites are collecting) and reverse engineered and people will start optimizing for it and the goal of the game will shift from winning to stat optimization.

if this is the case the system is fundamentally flawed, and the devs really need to rethink what theyre doing.

Although we pretty much already know for certain that stats are a part of the calculation, but it seems after the first week of ranked they realized they fucked up big and decreased the weight stats have drastically, but as long as they have any weight they will be gamed until the point where the weight is so close to 0 as to not even be there.

good stats are not winning the game, they literally do not matter. if you consistently get good stats but cant win you are not a good player, and individual games do not need to be adjusted for since all luck will even out over large samples.
 
Yeah
If stats are a part of the calculation then it will be eventually found out from data (which many sites are collecting) and reverse engineered and people will start optimizing for it and the goal of the game will shift from winning to stat optimization.

if this is the case the system is fundamentally flawed, and the devs really need to rethink what theyre doing.

Although we pretty much already know for certain that stats are a part of the calculation, but it seems after the first week of ranked they realized they fucked up big and decreased the weight stats have drastically, but as long as they have any weight they will be gamed until the point where the weight is so close to 0 as to not even be there.

good stats are not winning the game, they literally do not matter. if you consistently get good stats but cant win you are not a good player, and individual games do not need to be adjusted for since all luck will even out over large samples.
Yeah the system still flawed, I don't mind if they're not transparent with their rank system, but at least make it make sense, it's just not right when my friend, a player who have better win rate in rank calibration match yet he's placed in initiate while I am with similar win rate before ranked introduced, and have more lose than winning in rank calibration got oracle.
 
If stats are a part of the calculation then it will be eventually found out from data (which many sites are collecting) and reverse engineered and people will start optimizing for it and the goal of the game will shift from winning to stat optimization.

if this is the case the system is fundamentally flawed, and the devs really need to rethink what theyre doing.

Although we pretty much already know for certain that stats are a part of the calculation, but it seems after the first week of ranked they realized they fucked up big and decreased the weight stats have drastically, but as long as they have any weight they will be gamed until the point where the weight is so close to 0 as to not even be there.

good stats are not winning the game, they literally do not matter. if you consistently get good stats but cant win you are not a good player, and individual games do not need to be adjusted for since all luck will even out over large samples.
i agree 100%
the stats will be found out and there will be people playing with keeping them in mind, instead of doing their best for the victory.
 
Individual statistics drive me to be a better player. They will always exist even if the main stat is win or loss because I can't improve a random team so I must improve myself.
However, we do not have server statistics to compare ourselves with. We have some information from websites that collect it, but it is not enough.

Big game statistics work for professional chess players if they are going to spend their whole life playing, I want to play between 5 and 10 games a week. It's enough to want to know if I'm making any progress, but it's a quantity that individually I won't have access to knowing what I'm doing right or wrong to evolve.

I'm playing with 11 other people. Any chance that changes destiny to defeat or victory can make me think that I did something better or worse in this match.

I need individual stats because I'm not a one-figure player (and clearly the game doesn't want that since it forces us to choose more than one). A victory could just be a synergy on my team between the chosen characters. It could be a great match between me and great synergy between the opponents. I could be changing what I do based on the wrong premise. But if I have individual statistics I would know if my performance was below or beyond my average.

Without individual statistics (per character) we will just be a pigeon in a box that contains a button and a small door. The small door opens and we can eat. And maybe this has something to do with the button. Maybe not. We will press the button, then we will start speculations, superstitions, and we will create conversations like this but there is only one certainty: we don't know if we can or even if it is possible to control the door where food comes from.
 
Win\lose elo system was designed to express a skill of individual in a 1 vs 1 game - chess.
But the task at hand is to evaluate a skill of different individuals who are randomly assigned together into a team.

Elo system is misapplied to this task. It was not designed for it. It treats the team as individual which is fine in premades, but it is not fine in solo queue matchmaking. There is petabytes of copium and or rage discussions over this, people who come out on top of the system say its fair and it "averages out over time", people who come out on bottom weep, whatever. Simple fact is that if I'm on your team and I play bad and we lose, you yourself did not become a worse player. But your rank would drop. Same as if you carry me repeatedly, I did not become a better player. But my rank would go up.

Bottom line a functioning "True skill" system as you called it is needed, and just because it has not yet been invented does not mean that it is impossible or that it should not be attempted. Best case scenario it should be resistant to gross stat optimization abuse, but even if its not, stat optimization will still force players to perform better in games. And if it's not 100% resistant then I'd still rather it was there but not transparent.
 
if the system is gameable its already flawed.
There's a difference between a system that is gameable and a system that people think is gameable.

What I said still remains true, that it's still a very experimental calculation. Rank hasn't even been out for a full month.

We don't even know how they're calculating it, and so what's the point in making a bunch of assumptions on how it works based on rumors, and then complaining about these completely theoretical assumptions?
 
Individual statistics drive me to be a better player. They will always exist even if the main stat is win or loss because I can't improve a random team so I must improve myself.
However, we do not have server statistics to compare ourselves with. We have some information from websites that collect it, but it is not enough.

Big game statistics work for professional chess players if they are going to spend their whole life playing, I want to play between 5 and 10 games a week. It's enough to want to know if I'm making any progress, but it's a quantity that individually I won't have access to knowing what I'm doing right or wrong to evolve.

I'm playing with 11 other people. Any chance that changes destiny to defeat or victory can make me think that I did something better or worse in this match.

I need individual stats because I'm not a one-figure player (and clearly the game doesn't want that since it forces us to choose more than one). A victory could just be a synergy on my team between the chosen characters. It could be a great match between me and great synergy between the opponents. I could be changing what I do based on the wrong premise. But if I have individual statistics I would know if my performance was below or beyond my average.

Without individual statistics (per character) we will just be a pigeon in a box that contains a button and a small door. The small door opens and we can eat. And maybe this has something to do with the button. Maybe not. We will press the button, then we will start speculations, superstitions, and we will create conversations like this but there is only one certainty: we don't know if we can or even if it is possible to control the door where food comes from.
O, the mystery and the splendor. O, how majestic are the shadows which dance before me, gliding across these impenetrable walls. Could there be more to life than this...?-- Ah, a breadcrumb! (Waddle waddle; peck, peck, peck.)
 
Win\lose elo system was designed to express a skill of individual in a 1 vs 1 game - chess.
But the task at hand is to evaluate a skill of different individuals who are randomly assigned together into a team.

Elo system is misapplied to this task. It was not designed for it. It treats the team as individual which is fine in premades, but it is not fine in solo queue matchmaking. There is petabytes of copium and or rage discussions over this, people who come out on top of the system say its fair and it "averages out over time", people who come out on bottom weep, whatever. Simple fact is that if I'm on your team and I play bad and we lose, you yourself did not become a worse player. But your rank would drop. Same as if you carry me repeatedly, I did not become a better player. But my rank would go up.

Bottom line a functioning "True skill" system as you called it is needed, and just because it has not yet been invented does not mean that it is impossible or that it should not be attempted. Best case scenario it should be resistant to gross stat optimization abuse, but even if its not, stat optimization will still force players to perform better in games. And if it's not 100% resistant then I'd still rather it was there but not transparent.
but because you are the only constant in your teams, each match you have a random set of opponents and teammates, over a large set of games we see only your impact, because the randomness of teammates and opponents evens out over a large sample.

Its perfectly mathematically sound, there is no issue. it is not misapplied.

no limited set of statistics can tell you whether someone played good or bad.

you can play bad overall, especially when it matters, but still have good stats, and conversely it is also true.
 
Win\lose elo system was designed to express a skill of individual in a 1 vs 1 game - chess.
But the task at hand is to evaluate a skill of different individuals who are randomly assigned together into a team.

Elo system is misapplied to this task. It was not designed for it. It treats the team as individual which is fine in premades, but it is not fine in solo queue matchmaking. There is petabytes of copium and or rage discussions over this, people who come out on top of the system say its fair and it "averages out over time", people who come out on bottom weep, whatever. Simple fact is that if I'm on your team and I play bad and we lose, you yourself did not become a worse player. But your rank would drop. Same as if you carry me repeatedly, I did not become a better player. But my rank would go up.

Bottom line a functioning "True skill" system as you called it is needed, and just because it has not yet been invented does not mean that it is impossible or that it should not be attempted. Best case scenario it should be resistant to gross stat optimization abuse, but even if its not, stat optimization will still force players to perform better in games. And if it's not 100% resistant then I'd still rather it was there but not transparent.
Ehhhh, it's a combo of both. Someone finding a way to game the stats required to be seen as "good" doesn't mean those stats directly contribute to winning.

A true skill system based on individual performance is arguably more flawed than the one that relies strictly on a W/L record. Bad players can be carried by good teams to wins they didn't deserve, but overall that should be an outlier that is flattened over time. Someone focusing on say a KD:R could just take no risks and play safe just to ensure they don't die a lot despite, at best, doing nothing to contribute to a win, or at worst actively contributing to a loss.

Sometimes a death is absolutely worth the outcome, rush into mid, steal the rejuve and die? That is unequivocally a net positive for the team and their chance to win. Die trying to back cap the patron, or last hit a walker/shrine to get a flex spot, in most cases that is definitely worth the death.

Someone split pushes most of the game consistently taking objectives, but then gets rolled by the enemy team defending, still a net positive. Definitely better if they didn't die, but overall it's usually more worth it than the death.

The game is WAY too contextual and nuanced to be based entirely on a "True skill" system. Those can be abused, wins and losses cannot. A "True skill" system can contribute to wins if abused, but not always. But a W/L system always contributes to wins, if the only thing being rewarded is a binary win = true/false, then you give players the freedom to do anything necessary to win the game, which is precisely what you want to be incentivized in a ranked/competitive environment.

Humans are incentive based creatures, if you incentivize them to do anything else that doesn't directly contribute to winning, then the entire system fails. I think we have all seen more than enough Haze (no offence Haze mains lol) players farm 70k souls get a bunch of kills and then still die at the absolute worst times leading to a loss late game. Poor habits and behaviours centred around stat stuffing should be rewarded with losses, as any rational person would notice they keep losing rank because they keep losing games, and maybe put 2+2 together.

This is a team based game, you have to reward teamwork, otherwise, what's the point?
 
but because you are the only constant in your teams, each match you have a random set of opponents and teammates, over a large set of games we see only your impact, because the randomness of teammates and opponents evens out over a large sample.

Its perfectly mathematically sound, there is no issue. it is not misapplied.

no limited set of statistics can tell you whether someone played good or bad.

you can play bad overall, especially when it matters, but still have good stats, and conversely it is also true.
I am not saying win rate should not be a factor in evaluation. I am saying that individual contributions of each player can be numerically expressed, compared and factored in for more accurate bracket adjustment. Therefore they should be applied to improve accuracy. Reducing all the complexity of a specific player, game and team because "it will average out" is not "mathematically sound". It is gross oversimplification which, unfortunately, also has infinite margin for error, which is why not enough effort has been put into solving the actual problem, which is why we do not currently have a better ranking system than w/l elo.

PUBG actually has a great ranking system, which is also very simple, but it would not be sufficient for a MOBA game.
 
Since there are already stats in the game and performance can be evaluated based on them, I'd try the following:

We take the best player from the loser-team and the worst player from the winner team and give them 0 points for their win/loss.
The rest of the players' stats are calculated and each gets the next best/worst score with the idea that the closer you are to the best player in the losing team, the less points you lose and if you are the best player in the winning team -- you get the most amount of points.

And it must be a big number of stats that is taken into account, and they may have different weight in the equation.
 
Win\lose elo system was designed to express a skill of individual in a 1 vs 1 game - chess.
But the task at hand is to evaluate a skill of different individuals who are randomly assigned together into a team.

Elo system is misapplied to this task. It was not designed for it. It treats the team as individual which is fine in premades, but it is not fine in solo queue matchmaking. There is petabytes of copium and or rage discussions over this, people who come out on top of the system say its fair and it "averages out over time", people who come out on bottom weep, whatever. Simple fact is that if I'm on your team and I play bad and we lose, you yourself did not become a worse player. But your rank would drop. Same as if you carry me repeatedly, I did not become a better player. But my rank would go up.

Bottom line a functioning "True skill" system as you called it is needed, and just because it has not yet been invented does not mean that it is impossible or that it should not be attempted. Best case scenario it should be resistant to gross stat optimization abuse, but even if its not, stat optimization will still force players to perform better in games. And if it's not 100% resistant then I'd still rather it was there but not transparent.
To be fair, in CSGO Valve didn't use straight Elo but used a modified Glicko-2 rating system which worked fine.
 
Back
Top