|
I want to help the discussion remembering that the changes applied to the system are not perfect and have their negative points, but were necessary from my pov.
I'd also want to point up that we have 7 real ranks; with A- in the middle. As you can see the system automatically regulates middle ranks vs the lowest, red and yellow with +0 and +13 points per win. So don't get shocked if you see players with a matchlist played only vs low ranks hitting A-, because it's a middle rank and the "real ladder" starts from there.
Actually we encourage to play vs anyone, even only vs low players, getting as many green ranked as possible.
|
rank resets every season dont seem to make much sense
|
If there was no resets, over time, every player would have olympic rank. Not a joke.
|
I forget where I mentioned this, and if it was earlier in this thread then I'm sorry, but one change that I'd like to see that I don't think is too controversial is better reset compensation. I don't know a better way to say this in as many words, so I'll explain:
Currently, if a player is B- when the reset comes, they start at C- when the new season begins. I'm not sure if there are any others (A- to B- is not a thing, I think?).
Why not expand this system to all ranks? If you were C, you start at D+. If you were C+, you are now C-. B- > C. A+ > A-.
One issue I foresee is that people have to play fewer games to rank up with this system. As Bonyth astutely pointed out, the current system increases game numbers. I can understand how people would be less motivated to grind if they were only 2 "grades" lower. To what extent this would be an issue, I think no one can say with any certainty.
This also wouldn't prevent people from doing /clearstats before or after the reset, or from smurfing. But, as I am certain I said recently, 90% of the people who whoop on what are probably late-season C- Terrans at D rank in the first few weeks are NOT smurfs and I don't think smurfing is a genuine issue.
However, what this would accomplish is motivate people to reach their max rank in the first place. A lot of the comments I hear from people who are ~C is that it takes them like 100 games just to get there, because they lose to every B and higher player who passes them and don't win against all C players. "Why is this an issue? Doesn't that mean that this person is truly C?" the issue is that as these people are climbing the ranks to GET to C, they are often either crushed by better players or roflstomp worse players (which is the focal point of the whine so far in this thread). If the C player started even at D+, that would be fewer games needed to play to get to C where the competition is more "even" for them. However, this also means that people would reach their max earlier in the season while more people are climbing... Ah, it's complicated.
Another thing that seems to be affecting people is that they start at D and then in just a few losses fall to D-, and for some, all the way down to 1 point. I can see how this is depressing. Why not start the ladder at C so that it takes them longer to get to these depressing levels? The gap between D and D- is really unforgiving.
Just spitballing towards the end there but hopefully I made at least one or two interesting points.
I need to sleep more fml.
|
i just wanted to say that the real ladder starts at A-, but iccup.face beat me to it.
|
On March 02 2017 07:26 iCCup.Face wrote:I want to help the discussion remembering that the changes applied to the system are not perfect and have their negative points, but were necessary from my pov. I'd also want to point up that we have 7 real ranks; with A- in the middle. As you can see the system automatically regulates middle ranks vs the lowest, red and yellow with +0 and +13 points per win. So don't get shocked if you see players with a matchlist played only vs low ranks hitting A-, because it's a middle rank and the "real ladder" starts from there. Actually we encourage to play vs anyone, even only vs low players, getting as many green ranked as possible.
Just by playing with the same person who you are equally skilled with on MOTW, you'll both reach infinite rating.
I really dislike the current Iccup system, because you rank is completely meaningless, since there's endless points inflation. A rating that has much more value is in which percentile are you compared to the rest of the people on ladder.
Saying things like real ladder starts at A- is silly, because the truth is the real ladder never starts, because there's always point inflation, it just gets slightly smaller.
I think all games need to be point neutral, if you want to increase activity, then something like bonus pool that is time bound, not number of games bound is a lot better.
|
|
No idea what u saying FiWiFaKi, anyway you can't play vs the same person infinite times, it's against rules.
6. Playing against the same Player
6.1 It's forbidden to play against the same player more than:
• no limit between E to B+ if no abuses detected • 9 games in 24h between A- to Olympic
Also, actually the points you win vs same rank are "neutral" as you say, so idk
|
On March 02 2017 09:57 iCCup.BwTV wrote:No idea what u saying FiWiFaKi, anyway you can't play vs the same person infinite times, it's against rules. Show nested quote +6. Playing against the same Player
6.1 It's forbidden to play against the same player more than:
• no limit between E to B+ if no abuses detected • 9 games in 24h between A- to Olympic Also, actually the points you win vs same rank are "neutral" as you say, so idk
Yeah, those limits only exist because the ladder system is broken.
I could play vs my friend who is same skill level as me (no abuse) and we would keep getting more and more points, once we get to A-, okay, we would play 9 games a day because of the rule, but we would still keep getting points.
And the thing you posted doesn't match up with the picture in my post, and it doesn't consider MOTW either.
|
On March 02 2017 07:39 ARREST_HILLARY_NOW wrote: rank resets every season dont seem to make much sense
On March 02 2017 07:44 Bonyth wrote: If there was no resets, over time, every player would have olympic rank. Not a joke. The reset makes exactly that much sense, like there were a reset in chess ÉLŐ`s four times a year. The reset only exists because the Dota section needs it for some unknown reason. As many stated the real ladder starts at A-, too bad it does actually not just starts but stops right there, as the player pool become so much limited (+many other factors). I really would prefer the old point system combined with endless ladder / or with ÉLŐ points. If anyone wishes to reset, he can just do it, but reaching A- to not be able to play is a no go for a lot of people. Maybe a simple solution would be that if you really need to reset the server, keep track of previous ranks, and start the players with the same rank, I bet, It would increase the ladder activity.
|
On March 02 2017 07:44 Bonyth wrote: If there was no resets, over time, every player would have olympic rank. Not a joke. this. theres already big noobs with blue and green icons during end season..
edit: damn I sound so harsh :S I mean the required skill to those ranks seems to have drop significantly
|
I think you over complicate the debate. Players who are better will still rise in ranks faster due to sheer win/loss ratio, also meaning that eventually they will run into other people with similar win/loss-ratio at say around C+/B- which will stagger their rise in rank. Also in return, if we propose that an B/B+ player can maintain the same win/loss-ratio up until a certain point as an A-/A/A+ player then this will differentiate the two once they collide on B-/B-level with equal win ratio. Meaning the player who is actually B-/B will start loosing a lot more than the A-/A player. Thus I don't really see the problem, yes obviously you can heavily inflate your rank by just massgaming/playing lesser opponents but your games played will reflect that. Any ranking system that allows you to play players of a different skill than your own will automatically mean it can be inflated.
Only AMM-systems that will always match you against an opponent of appropriate MMR are immune to such things. Considering the player base at ICCup that would never work because the search time would be infinite.
It comes down to the question whether 100 % accurate ranking is better than actually having someone to play. These days I'd say the latter alternative is the better one.
|
On March 03 2017 01:47 dignitas.merz wrote: I think you over complicate the debate. Players who are better will still rise in ranks faster due to sheer win/loss ratio, also meaning that eventually they will run into other people with similar win/loss-ratio at say around C+/B- which will stagger their rise in rank. Also in return, if we propose that an B/B+ player can maintain the same win/loss-ratio up until a certain point as an A-/A/A+ player then this will differentiate the two once they collide on B-/B-level with equal win ratio. Meaning the player who is actually B-/B will start loosing a lot more than the A-/A player. Thus I don't really see the problem, yes obviously you can heavily inflate your rank by just massgaming/playing lesser opponents but your games played will reflect that. Any ranking system that allows you to play players of a different skill than your own will automatically mean it can be inflated.
Only AMM-systems that will always match you against an opponent of appropriate MMR are immune to such things. Considering the player base at ICCup that would never work because the search time would be infinite.
It comes down to the question whether 100 % accurate ranking is better than actually having someone to play. These days I'd say the latter alternative is the better one.
Chess elo lets you play anyone without having any point inflation (the winner gets as many points as the loser loses)... Let's say you play someone the same level as you, winner gets 50, loser loses 50. You play someone a rank above you, you win you get 75 and they lose 75, you lose, then you lose 25 and they win 25. The numbers are there to give an example, they can obviously be anything, and better yet, defined by an equation like:
Predicted result #1 = 1/(1+x^((rating2-rating1)/d)) Predicted result #2 = 1/(1+x^((rating1-rating2)/d))
Then based on the results of the game, adjust the score
New rating = old rating + K(actual result - predicted result)
For both players, which in essence is the ELO system. The only things left is choosing x,d,k. The x and d essentially define the distribution of points between players. For example the classic x = 10 and d = 400 means that a player with 400 points more will be 10x as likely to win. But you can change x to a higher number to space out the distribution more, and lower the k, so that for example you achieve a distribution where 3000 points will be average, and 1000 points will mean that they are 10x as likely to win. So the difference between 4k and 3k would be consistent, where a 4k has a 90.9% win rate against a 3k player, and a 5k player has a 90.9% win rate against a 4k player.
The last thing is just set a reasonable k, which is the converging parameter, though a PID controller opposed to a P might me more appropriate. But for example, I'm a 5k player winning against a 4k player, that means my actual result was a 1 (win, and my predicted was 0.909. Make k = 100, and now I'm gaining 9.1 points for that win (though against an equally matched opponent I would win 50 points).
|
I just dropped in to suggest this
|
we dont want elo on iccup. why is there 1 guy every year advocating elo like its the only way of life.
|
On March 03 2017 08:14 Bakuryu wrote: we dont want elo on iccup. why is there 1 guy every year advocating elo like its the only way of life.
I'm not saying it's the only one, I just explained how the current system is very bad (because I could become #1 on the server by repeatedly playing 1 person who is my skill level because of points inflation).
Then I provided an alternative (just so it doesn't come off as I'm only complaining). Most games that have decently large studio has an ELO derived system for their points. Whether that's SC2, Dota, League, CS:GO... Uhh, well I don't know too many games, but I'm never seen a system as rudimentary is this one.
It's not like everyone is exposed to the math, before you start your game, it'll tell you how many points you will get if you win and lose, the math is in the backend, that's how chess does it.
Then assign point ranges for all the letter ranks, and voila. The main thing that makes me not play on iccup is the ladder system (but you prob don't want me there anyways). Anyway, I heard you loud and clear, you don't like ELO type systems, but why? The complexity of this point system is small and easy to grasp.
|
On March 03 2017 08:35 FiWiFaKi wrote:Show nested quote +On March 03 2017 08:14 Bakuryu wrote: we dont want elo on iccup. why is there 1 guy every year advocating elo like its the only way of life. I'm not saying it's the only one, I just explained how the current system is very bad (because I could become #1 on the server by repeatedly playing 1 person who is my skill level because of points inflation). Then I provided an alternative (just so it doesn't come off as I'm only complaining). Most games that have decently large studio has an ELO derived system for their points. Whether that's SC2, Dota, League, CS:GO... Uhh, well I don't know too many games, but I'm never seen a system as rudimentary is this one. It's not like everyone is exposed to the math, before you start your game, it'll tell you how many points you will get if you win and lose, the math is in the backend, that's how chess does it. Then assign point ranges for all the letter ranks, and voila. The main thing that makes me not play on iccup is the ladder system (but you prob don't want me there anyways). Anyway, I heard you loud and clear, you don't like ELO type systems, but why? The complexity of this point system is small and easy to grasp.
Any system is abusable. You can do exactly the same thing with a single friend in an ELO based ladder unless you're also advocating for automated matchmaking? All it needs, is a few additional accounts.
|
On March 03 2017 09:05 B-royal wrote:Show nested quote +On March 03 2017 08:35 FiWiFaKi wrote:On March 03 2017 08:14 Bakuryu wrote: we dont want elo on iccup. why is there 1 guy every year advocating elo like its the only way of life. I'm not saying it's the only one, I just explained how the current system is very bad (because I could become #1 on the server by repeatedly playing 1 person who is my skill level because of points inflation). Then I provided an alternative (just so it doesn't come off as I'm only complaining). Most games that have decently large studio has an ELO derived system for their points. Whether that's SC2, Dota, League, CS:GO... Uhh, well I don't know too many games, but I'm never seen a system as rudimentary is this one. It's not like everyone is exposed to the math, before you start your game, it'll tell you how many points you will get if you win and lose, the math is in the backend, that's how chess does it. Then assign point ranges for all the letter ranks, and voila. The main thing that makes me not play on iccup is the ladder system (but you prob don't want me there anyways). Anyway, I heard you loud and clear, you don't like ELO type systems, but why? The complexity of this point system is small and easy to grasp. Any system is abusable. You can do exactly the same thing with a single friend in an ELO based ladder unless you're also advocating for automated matchmaking? All it needs, is a few additional accounts.
No, it's different.
In my example that I explained, I am not abusing anything. I just have a friend who I like mass gaming with (heck, on SB I often have 15-20 games straight on one day against the same person, like practice partners), and on iccup we generate points out of thin air, with my proposal it'd be a zero sum game.
Of course any system can be abused if you let people make infinite accounts and play vs each other with them, but that's a clear abuse over normal gameplay. I do think that enforcing account limits in some way is a good idea, simply because less smurfs, people act better to maintain and raise their reputation (such as on SB).... But anyway, I recognize it's hard-impossible to do effectively without being very intrusive.
And no, I am not advocating automatic matchmaking (it can work with or without it).
|
this is how i see it. it is a ladder system, not a ranking system. the goal is to get as many points as you can in the 3 months while rewarding you with "medals" (=rank) in your profile. our current user base is about 1000 active people, where 500 are in the D+ to C ranks, while the other 500 gets spread over C+ to olympic the goal of the current system (lets say version 3), which you are saying is "bad", is to increase the activity in ladder (= to get more ladder games) if you use the point inflation to play 600~ games against the same person to get to A+/Olympic, then the ladder system is rewarding your activity with the A+/Olympic medal. in the iccup ranking system before that (version 2), you could not do "point inflation 50% win against same skilled player" because on A- you lost 140 point vs A- while only winning 130 (if you play motw). In version 1, A- vs A- you lost 130 points, and back then our player base was still strong so they wanted to thin out the A- ranks by going from -130 to -140 points. all in all, the "real ladder" started around B- back then. because the player base shrunk dramatically after sc2, you could only ladder until B and then you could decide between: 1) playing the same 5 people again and again. 2) waiting between 2 weeks and 1 month for enough people to catch up 3) reset account/start with 2nd account/smurf few seasons later, they changed it to the current system, where the "real ladder" starts at A- and you have a wider range of players you can play. now you can mostly ladder until A and then decide between 1), 2) and 3).
we just have too few players to saturate all the different "skill levels" and i dont see how elo on iccup will give us more players
|
On March 03 2017 09:24 Bakuryu wrote: this is how i see it. it is a ladder system, not a ranking system. the goal is to get as many points as you can in the 3 months while rewarding you with "medals" (=rank) in your profile. our current user base is about 1000 active people, where 500 are in the D+ to C ranks, while the other 500 gets spread over C+ to olympic the goal of the current system (lets say version 3), which you are saying is "bad", is to increase the activity in ladder (= to get more ladder games) if you use the point inflation to play 600~ games against the same person to get to A+/Olympic, then the ladder system is rewarding your activity with the A+/Olympic medal. in the iccup ranking system before that (version 2), you could not do "point inflation 50% win against same skilled player" because on A- you lost 140 point vs A- while only winning 130 (if you play motw). In version 1, A- vs A- you lost 130 points, and back then our player base was still strong so they wanted to thin out the A- ranks by going from -130 to -140 points. all in all, the "real ladder" started around B- back then. because the player base shrunk dramatically after sc2, you could only ladder until B and then you could decide between: 1) playing the same 5 people again and again. 2) waiting between 2 weeks and 1 month for enough people to catch up 3) reset account/start with 2nd account/smurf few seasons later, they changed it to the current system, where the "real ladder" starts at A- and you have a wider range of players you can play. now you can mostly ladder until A and then decide between 1), 2) and 3).
we just have too few players to saturate all the different "skill levels" and i dont see how elo on iccup will give us more players
I feel like it takes too long to get to a steady state rankings because you're getting points until your win rate falls to like 30% (depending your skill level)... So a steady state rank means you're losing most of your games, which is also kind of shitty mentally for the player.
My view is that currently activity is punished, because what happens is that if I play a lot of games I have nobody to play against because my rank shoots up quicker than other people if I'm a high level player, and if I'm a lower level player, well then I'll reach say B- with a low winrate, and any B- player who hasn't played as many games as me will likely destroy me, and so I'll stay at B- by winning 50% of my games against C players, and losing most games against B- and C+ players.
In the current system, the points are playing catch-up (because of the inflation), at the start of the season I reach B-, now I have nobody to play against, now more people played games, and there are more B's now, rinse and repeat.
I understand the idea, but I feel like it's the wrong way of stimulating activity. Improvement in rank feels meaningless, because even if I'm getting worse my rating is going up, it's more difficult to hop on and play a few games for someone more casual (I'm surrounded by a lot of newer people coming back on ShieldBattery)... And importantly like I mentioned, I feel like I'm punished, not rewarded for playing a lot of games, because if I'm playing MOTW and I'm a B player or worse (like most of us)... The I'm losing most of my games ( to maintain my points against same rank players I need 28% win rate)... 50/(50+130) = 0.27778.
Being in this kind of situation just encourages more people to smurf, and makes the genuine regular players care less if they're matched up against smurfs, because heck, your win rate is low already, you don't lose as many points, etc. In my replay folder I have exactly 666 games I've played on SB since October 2016 (most were 1v1's, though some obs games and Racewars, etc)... I don't need points like this to be encouraged to play, maybe short term it's an effective strategy (like within the season), but I feel like long term it's more difficult to track improvement, pushes your hand to play (opposed to purely individual desire), and reduces morale, and thus lowers player retention.
That's my view on the situation anyway.
|
|
|
|