• Welcome to Smashboards, the world's largest Super Smash Brothers community! Over 250,000 Smash Bros. fans from around the world have come to discuss these great games in over 19 million posts!

    You are currently viewing our boards as a visitor. Click here to sign up right now and start on your path in the Smash community!

Can we talk about how bad the low GSPs are

Status
Not open for further replies.

Luigifan18

Smash Master
Joined
Feb 19, 2015
Messages
3,134
Switch FC
SW-5577-0969-0868
Wow, people who have just started playing the game tend to not be very good at it! What a shocking revelation! :4pacman:
 

meleebrawler

Smash Hero
Joined
Sep 8, 2014
Messages
8,156
Location
Canada, Quebec
NNID
meleebrawler
3DS FC
2535-3888-1548
Elite isn't some hidden For Glory paradise, you can still get matched in FFAs

And Nintendo is basing their ****ing balancing on them, jesus **** this online is garbage
It's almost like there isn't actually a top brass of players online when a game has just launched!
 

meleebrawler

Smash Hero
Joined
Sep 8, 2014
Messages
8,156
Location
Canada, Quebec
NNID
meleebrawler
3DS FC
2535-3888-1548
Laggy players can also make elite by virtue of being laggy.
Unless you're talking about lag-switchers, players suffer equally from one of them having a bad connection. Even then, those who abuse that cheat will likely get caught out before long with how much discrepancy can be seen between the two player's responsiveness.

Besides, the whole point of it is to track how characters at high level are used, not how well players are doing in general. If it's clear one is preying on inexperience and/or lag to get ahead, they can just disregard those matches and deal with the offending players as necessary.
 
Last edited:

1FC0

Smash Lord
Joined
Aug 21, 2013
Messages
1,801
To be honest I consistently get Stock 1v1's in Elite Smash and often even in Omega stages. For me Elite Smash is a FG paradise.
 

StoicPhantom

Smash Ace
Joined
Dec 11, 2018
Messages
618
Unless you're talking about lag-switchers, players suffer equally from one of them having a bad connection. Even then, those who abuse that cheat will likely get caught out before long with how much discrepancy can be seen between the two player's responsiveness.

Besides, the whole point of it is to track how characters at high level are used, not how well players are doing in general. If it's clear one is preying on inexperience and/or lag to get ahead, they can just disregard those matches and deal with the offending players as necessary.
I'm talking more wi-fi players. It's a stuttering mess for me but not for them. I find elite is actually where these types are more common.
Lag is going to affect the match differently then offline play. I don't know how useful that will be to track.
 

Andinus

Smash Journeyman
Joined
Apr 10, 2014
Messages
462
Location
USA
NNID
kyonyuudaisuki
I'm talking more wi-fi players. It's a stuttering mess for me but not for them. I find elite is actually where these types are more common.
Lag is going to affect the match differently then offline play. I don't know how useful that will be to track.
Quick play just sucks all around elite or not, arena battles is the way to go, I find the connections overall tend to be better in arenas, and you always get to play how you want. Screw Nintendo’s match making garbage, I’m making my own matches just fine.
 

Andinus

Smash Journeyman
Joined
Apr 10, 2014
Messages
462
Location
USA
NNID
kyonyuudaisuki
try joining us down in the 0-70,000 gsp range

its only 1v1s down here
How can you possibly stay that low? I played for one day, and woke up the the next to like double GSP and elite smash, and haven’t touched quickplay since Monday and I’m still in elite smash. I say arena battles all the way. spread the word, join, make arenas and turn quick play into a ghost town in hopes they change it.
 

OF 'til I OD

More vibes, please.
Joined
Oct 21, 2005
Messages
2,603
Location
Wisconsin
3DS FC
3797-8618-3772
Switch FC
SW-4939-9118-0296
How can you possibly stay that low? I played for one day, and woke up the the next to like double GSP and elite smash, and haven’t touched quickplay since Monday and I’m still in elite smash. I say arena battles all the way. spread the word, join, make arenas and turn quick play into a ghost town in hopes they change it.
It seems to be random, though. I'm in the 1.5 mil range with my two characters I use, yet no Elite Smash.
 

OF 'til I OD

More vibes, please.
Joined
Oct 21, 2005
Messages
2,603
Location
Wisconsin
3DS FC
3797-8618-3772
Switch FC
SW-4939-9118-0296
It seems to be going up everyday by 100,000ish. I remember when it was 800,000ish.
Ah, thanks.

I suppose it does make a bit of sense that the more players that play, the higher the threshold. I suppose I am okay with not being in. I still find competetive matches with others ranked near my GSP. I feel I'm an alright player, but if I'm still occasionally losing now, I'd probably be weeded out quickly in Elite anyway.

Can't really devote more than a couple hours per day of playing, either, and those few hours are not optimal training time.

Ah well.
 

StoicPhantom

Smash Ace
Joined
Dec 11, 2018
Messages
618
Ah, thanks.

I suppose it does make a bit of sense that the more players that play, the higher the threshold. I suppose I am okay with not being in. I still find competetive matches with others ranked near my GSP. I feel I'm an alright player, but if I'm still occasionally losing now, I'd probably be weeded out quickly in Elite anyway.

Can't really devote more than a couple hours per day of playing, either, and those few hours are not optimal training time.

Ah well.
No problem. If it makes you feel any better it's not too different from where you're at. I'm pretty entry level so I don't know really know how high it goes.
 
Last edited:

Insistent

Smash Rookie
Joined
Jan 2, 2019
Messages
21
Location
Quebec, Canada
Ive found 1-1.5M gsp battle to be the easiest on all 3 character i play, hardest is under, even lower than 100k, search me why
 

Xquirtle

Smash Journeyman
Joined
Jan 30, 2015
Messages
232
Location
Columbus, Ohio
Ive found 1-1.5M gsp battle to be the easiest on all 3 character i play, hardest is under, even lower than 100k, search me why
Yeah the system makes no sense. I have characters ranging from 200k to 2.8 mil, and its maybe a mild correlation in skill. I will say that i run into the absolute best players on my high GSP mains, but i also run into opponents that I would describe as pretty damn good at like 300k gsp. and I mean players that understand the mental aspect of smash bros that have to be out played and mixed up constantly. Ofc your side B spamming turd balls are also more prevalent at low GSP, and the generally come with a giant lag advantage to handicap precise play.

Anyway, my Chrom is buried at like 300k, and i can't get him out! Literally play every single other FE sword fighter (even roy) and cloud over 2 mil. I lost my first 3 or 4 sets with Chrom and he was just done after that. Not sure how this works but its hard to understand how inconsistent it is across the same players roster.
 

leafgreen386

Dirty camper
Joined
Mar 20, 2006
Messages
3,577
Location
Playing melee and smash ultimate
Yeah the system makes no sense. I have characters ranging from 200k to 2.8 mil, and its maybe a mild correlation in skill. I will say that i run into the absolute best players on my high GSP mains, but i also run into opponents that I would describe as pretty damn good at like 300k gsp. and I mean players that understand the mental aspect of smash bros that have to be out played and mixed up constantly. Ofc your side B spamming turd balls are also more prevalent at low GSP, and the generally come with a giant lag advantage to handicap precise play.

Anyway, my Chrom is buried at like 300k, and i can't get him out! Literally play every single other FE sword fighter (even roy) and cloud over 2 mil. I lost my first 3 or 4 sets with Chrom and he was just done after that. Not sure how this works but its hard to understand how inconsistent it is across the same players roster.
It'll just take a lot of games. You can always climb out, but at low GSPs, it may feel like you're making very little progress, even when you're meaningfully increasing your rating. Once you get a few hundred thousand GSP higher, GSP will start to swing much more violently with wins and losses, as there's more players in that range.
 

Xquirtle

Smash Journeyman
Joined
Jan 30, 2015
Messages
232
Location
Columbus, Ohio
It'll just take a lot of games. You can always climb out, but at low GSPs, it may feel like you're making very little progress, even when you're meaningfully increasing your rating. Once you get a few hundred thousand GSP higher, GSP will start to swing much more violently with wins and losses, as there's more players in that range.
Yeah, i mean i realize that its in my control to increase it. its just disproportionately annoying and difficult to dig something out of low GSP when compared to my multitude of characters that never go below 2 mil, even random crap that I rarely play. Like my Chrom gets matched against 2.5 mil GSP when its at 300k, and if i lose, I lose like 75k GSP lol. Whatever logic is at work here is totally asinine, at least in a day and age were virtually all online match making either has a visible or hidden ELO system backing it. Almost feels like it considers my roster GSP when match making, but punishes / rewards the character individually as if the match were fair.

Anyway, just using my Chrom as an example of how this stuff makes no sense since hes a huge outlier in my roster. If the system were even remotely close to an ELO standard, you would never see such variance in rating and percentile across the same players roster.
 

leafgreen386

Dirty camper
Joined
Mar 20, 2006
Messages
3,577
Location
Playing melee and smash ultimate
Yeah, i mean i realize that its in my control to increase it. its just disproportionately annoying and difficult to dig something out of low GSP when compared to my multitude of characters that never go below 2 mil, even random crap that I rarely play. Like my Chrom gets matched against 2.5 mil GSP when its at 300k, and if i lose, I lose like 75k GSP lol. Whatever logic is at work here is totally asinine, at least in a day and age were virtually all online match making either has a visible or hidden ELO system backing it. Almost feels like it considers my roster GSP when match making, but punishes / rewards the character individually as if the match were fair.

Anyway, just using my Chrom as an example of how this stuff makes no sense since hes a huge outlier in my roster. If the system were even remotely close to an ELO standard, you would never see such variance in rating and percentile across the same players roster.
It's because GSP is reflecting the hidden rating change. The change in GSP is based on how many players are in the range you passed through. Dropping 75k makes sense for 300k, since once you start winning some games, you'll be climbing by hundreds of thousands GSP per win all the way up to 2+ mil.

Let's assume the distribution of players follows a normal curve (a reasonable assumption), and that there are exactly 3M players on the ladder. I know there are more by now, but this is just for approximate figuring anyway. Without knowing the actual rating values assigned to players, we can't calculate what a standard deviation *is*, but it doesn't really matter. What's important for change in GSP is how close those players are on the ladder, since GSP is just a stack ranking of players.



A single standard deviation contains 68% of the data, which is to say, about 2.04M players are within a single standard deviation of the norm. That means everything from around 500k to 2.5M, with the center at 1.5M. That's a *lot* of players all within a relatively tiny region of the rating system. Based on what we know of elite smash, it's slightly less than two standard deviations above the norm (it's approximately top 3.5%, whereas two standard deviations above the norm would be top 2.5%).

What this means is that GSP will swing violently within this 2M gap; you you can climb (or drop) from one extreme to the other in a very short time. The 1M-2M region is especially volatile, and you can probably climb from 1M to 2M in somewhere around 3-4 games, if you win all of them. However, the further you are from the center, the less drastically your GSP will change. GSP will also change more if you lose to someone far below you or win against someone far above you.

This is basically just a really long way of saying "win a few games, then your subsequent wins will cause your GSP to skyrocket."
 

Xquirtle

Smash Journeyman
Joined
Jan 30, 2015
Messages
232
Location
Columbus, Ohio
It's because GSP is reflecting the hidden rating change. The change in GSP is based on how many players are in the range you passed through. Dropping 75k makes sense for 300k, since once you start winning some games, you'll be climbing by hundreds of thousands GSP per win all the way up to 2+ mil.

Let's assume the distribution of players follows a normal curve (a reasonable assumption), and that there are exactly 3M players on the ladder. I know there are more by now, but this is just for approximate figuring anyway. Without knowing the actual rating values assigned to players, we can't calculate what a standard deviation *is*, but it doesn't really matter. What's important for change in GSP is how close those players are on the ladder, since GSP is just a stack ranking of players.



A single standard deviation contains 68% of the data, which is to say, about 2.04M players are within a single standard deviation of the norm. That means everything from around 500k to 2.5M, with the center at 1.5M. That's a *lot* of players all within a relatively tiny region of the rating system. Based on what we know of elite smash, it's slightly less than two standard deviations above the norm (it's approximately top 3.5%, whereas two standard deviations above the norm would be top 2.5%).

What this means is that GSP will swing violently within this 2M gap; you you can climb (or drop) from one extreme to the other in a very short time. The 1M-2M region is especially volatile, and you can probably climb from 1M to 2M in somewhere around 3-4 games, if you win all of them. However, the further you are from the center, the less drastically your GSP will change. GSP will also change more if you lose to someone far below you or win against someone far above you.

This is basically just a really long way of saying "win a few games, then your subsequent wins will cause your GSP to skyrocket."
Thank you for the lesson on high school level normal curves. The GSP system itself is a straight rank with precisely one person at each rank. This is not a normal distribution as any range of the rankings contains a number of players equal to that range. There are 3 million players between 1 and 3,000,000, there are 1 million players between 1 mil and 2 mil and so on. No curve to be had. It is a flat line with an X axis of 0-3 mil and 1 player at each point. This does not discredit your math. With a mean of 1.5 mil and 3 mil total players, it is indeed true that 68% of the players will be between 500k and 2.5m. What you are missing is that 500k to 2.5m is precisely 68% of the GSP values as well. Your math shows that 68% of the rating system contains exactly 68% of the players. My point is that the GSP system is merely a display for what is going on in the background and doesn't give us any sort of distribution to look at. What is actually of interest is the ranking system that drives GSP, which we cannot see...

...and at best, there is an ELO based system that determines the actual ranks, which is then converted into GSP on the read out. That would be described by a normal curve, however, the matchmaking system does not suggest that this is the case. A 2.5 mil player, which is in the top 80% of the rankings, should never be matched against a 300k player in the bottom 10%. Furthermore, any sort of rational ELO driven system would not punish the 300k player with a significant amount of rating loss unless they lost to a similar 300k ranked player. Losing to a player on the other end of the rating system should result in a rating loss of 0, and a massive rating gain if you win. Vice versa for the higher rated player.

In order for your theory to be true, we would need a GSP ranking represented by a normal curve, which we know it is not, or the system that feeds the GSP ranking to be represented by a normal curve. The latter is possible, but I cannot say that the current match making system is suggestive of it.

I find it most likely that there is some Nintendo jank going on in the background. My theory is that it is just a win/ loss system that does not differentiate the quality of your wins and losses, thus punishing low ranked players for losing to much better players. Feels like a fair claim given their track record on absolutely ridiculous design decisions around rating.
 

leafgreen386

Dirty camper
Joined
Mar 20, 2006
Messages
3,577
Location
Playing melee and smash ultimate
It cannot be purely winrate-based, otherwise it wouldn't be possible to be in ES with a 50% WR. The extremely large swings in GSP in the middle regions and extremely small swings in the far regions cannot be explained by "GSP" being the actual rating system; it's a front for the actual behind-the-scenes rating system. GSP is itself just a stack ranking of players, so of course it doesn't follow a normal distribution, but the system it's based on probably does.

Suppose that whatever rating system it's using behind the scenes has the median player at 1600 rating points, with a standard deviation of 400 rating points. This would place 68% of players (500k-2.5M GSP) between 1200 and 2000, with the top and bottom 2.5% of players below 800 or above 2400. Pros like ZeRo or Leffen would be over 2800. The total range of the rating system is very large, but there's a concentration of players in a relatively small part of it. In the region surrounding 1600, there's a lot of players there, so when you win or lose a match in that region, there's a huge amount of volatility in how much you move.

I got knocked down to 30k GSP when I started playing online, and climbed up from there to elite smash. I've seen the spectrum of this game, and movement of GSP definitely seems to correspond with what you would expect from a normal distribution.
 
Last edited:

Xquirtle

Smash Journeyman
Joined
Jan 30, 2015
Messages
232
Location
Columbus, Ohio
It cannot be purely winrate-based, otherwise it wouldn't be possible to be in ES with a 50% WR. The extremely large swings in GSP in the middle regions and extremely small swings in the far regions cannot be explained by "GSP" being the actual rating system; it's a front for the actual behind-the-scenes rating system. GSP is itself just a stack ranking of players, so of course it doesn't follow a normal distribution, but the system it's based on probably does.

Suppose that whatever rating system it's using behind the scenes has the median player at 1600 rating points, with a standard deviation of 400 rating points. This would place 68% of players (500k-2.5M GSP) between 1200 and 2000, with the top and bottom 2.5% of players below 800 or above 2400. Pros like ZeRo or Leffen would be over 2800. The total range of the rating system is very large, but there's a concentration of players in a relatively small part of it. In the region surrounding 1600, there's a lot of players there, so when you win or lose a match in that region, there's a huge amount of volatility in how much you move.

I got knocked down to 30k GSP when I started playing online, and climbed up from there to elite smash. I've seen the spectrum of this game, and movement of GSP definitely seems to correspond with what you would expect from a normal distribution.
Yes, if the system even modestly correlates player skill (true normal dist) to its ranking algorithm, then we definitely have some sort of a bell curve. Not trying to specifically argue semantics on the distribution. I'm saying that the behind the scenes system is Nintendo jank, as evidenced by a 300k losing 25% of their ranking to somebody 75% further up the ladder than they are. Movement definitely has to be slower on either end of the ranking system, as it becomes more demanding to get better or worse than the absolute best and worst. That has nothing to do with terrible match making + rating loss in mismatched fights. Like what sort of algorithm can be confident that I am now the 200k best player instead of 300k after losing to the 2.5 millionth best player? It would be one thing if i lost to a player between 200k and 300k, but we're talking about the other side of the hypothetical bell curve here. Its silly and nonsensical.

I don't think its directly ranked off of win rate (characters with 0 games played would be more volatile). They key in that statement is that it doesn't seem to care about quality of your wins and losses, which suggests that a win is a win and a loss is a loss in whatever algorithm is in place. When you combine that with bad matchmaking, which we can see is clearly an issue, it produces a more frustrating system. I would be interested to see how the system would behave if a 200k went 50/50 with say 50 1 mils. Does the 200k stay flat or approach 1 mil? I'm not so sure that it would actually move anywhere.
 

Billy Maize

Smash Rookie
Joined
May 30, 2015
Messages
7
Location
Colorado
Switch FC
0610-7203-0641
I got knocked down to 30k GSP when I started playing online, and climbed up from there to elite smash. I've seen the spectrum of this game, and movement of GSP definitely seems to correspond with what you would expect from a normal distribution.
I agree with this observation. I had a similar experience getting knocked down to around 100k with captain falcon, whenever I won a match at that point it would bring me up about 10-20k. But when I finally got to about 400-500k, my wins started to skyrocket my GSP by hundreds of thousands until I made it back to 1.5 million, where I started.
 
D

Deleted member

Guest
Xquirtle Xquirtle Is that you're favorite word? "Anyway"? Sorry, but a pet peeve has been touched.

Title says it all, I hate item ffa matches
The heck is wrong with item FFAs? I find them much less personal than teams or 1v1.
 
Last edited by a moderator:

Xquirtle

Smash Journeyman
Joined
Jan 30, 2015
Messages
232
Location
Columbus, Ohio
Xquirtle Xquirtle Is that you're favorite word? "Anyway"? Sorry, but a pet peeve has been touched.

The heck is wrong with item FFAs? I find them much less personal than teams or 1v1.
So you have a problem with people saying anyway to transition out of a digression, and you like item FFAs? Am i reading that correctly?
 

MG_3989

Smash Lord
Joined
Jan 1, 2019
Messages
1,130
Location
New Jersey
Switch FC
SW-8397-3391-6411
What I don’t understand is why when I pick up a character I have say around 2.5 mil and it gets matched with somebody at 200k. That’s just not fair and terrible matchmaking. Maybe it’s because this person has other characters with higher GSPs but it still doesn’t seem like a fair system
 

Xquirtle

Smash Journeyman
Joined
Jan 30, 2015
Messages
232
Location
Columbus, Ohio
What I don’t understand is why when I pick up a character I have say around 2.5 mil and it gets matched with somebody at 200k. That’s just not fair and terrible matchmaking. Maybe it’s because this person has other characters with higher GSPs but it still doesn’t seem like a fair system
I see the same thing as well. Its my main argument for how we can be pretty dang sure that GSP has really stupid logic behind it. What really sticks out here is how Sakurai has claimed that the system is designed not to alienate bad players. Like it doesn't show peoples names in the ranking, or really add any context to it. But they turn around and put potato's up against experienced players just to get humiliated in 1v1s lol.
 

Sucumbio

Smash Giant
Moderator
Writing Team
Joined
Oct 7, 2008
Messages
8,133
Location
Icerim Mountains
Inexperienced definitely dunno about bad hut God almighty if you're in a team battle and you get 3'd and I still have all my stocks please use one of my stocks instead of expecting me to win 2v1 especially if they are both on their last stock!
 

soviet prince

I am the terror that flaps in the night
Premium
Joined
Nov 10, 2007
Messages
3,142
Location
Kentucky
NNID
7066-9708-9591
I have been playing with my main (ridley) online for hundred of hours trying to practice and get better with him, but no matter how hard or long I try I seem to lose a lot more then I win. It's really frustrating losing after losing esp when you spent hours to bring your gsp up. early today I almost had it at 2 million but end of day at 500,000 back wear I started and it going to take forever to get back to wear I was.

Gsp is not fair at all you win a game it goes up a little but if you lose your drop drastically which is bs. I was a 9500 and lost to a guy I already beat twice and it dropes down to 500,000 which is nuts considering it took multiple wins from 500,000 to get 950,000.
 

Sucumbio

Smash Giant
Moderator
Writing Team
Joined
Oct 7, 2008
Messages
8,133
Location
Icerim Mountains
I'm almost positive the match making system takes more than GSP into account. It also looks at your win loss rate over last 10 and maybe even 50 matches (seeing as you can technically view these demographics yourself). Thus one may indeed streak from 500k to 2mil and then end up back at 500k. For me my character's GSP will swing as high as 3.3 for instance and as low as 2.7 but the net change is usually only plus or minus 50k or so once I decide to stop playing.
 

Coolboy

Smash Journeyman
Joined
Aug 27, 2018
Messages
382
Location
Netherlands
i discovered how GSP kinda works..since i play all modes and also play 4 players matches not just 1v1 matches.. if you play 1v1 matches you get really punished if you lost cause the system sees that as ''you lost and your opponent won'' ^^''

but if you look at a 4 players match.. the winner gets GSP the runner up gets some GSP as well..the 3rd place player loses a bit of GSP and the 4th place player loses much more GSP

i can really say that if you are struggling with 1v1 atm and you really want to gather some GSP first..you should play 4 players matches and make sure you win or get 2nd place, it might make you get less frustrated if you aren't so struggling anymore GSP wise! :D
 

Sucumbio

Smash Giant
Moderator
Writing Team
Joined
Oct 7, 2008
Messages
8,133
Location
Icerim Mountains
i discovered how GSP kinda works..since i play all modes and also play 4 players matches not just 1v1 matches.. if you play 1v1 matches you get really punished if you lost cause the system sees that as ''you lost and your opponent won'' ^^''

but if you look at a 4 players match.. the winner gets GSP the runner up gets some GSP as well..the 3rd place player loses a bit of GSP and the 4th place player loses much more GSP

i can really say that if you are struggling with 1v1 atm and you really want to gather some GSP first..you should play 4 players matches and make sure you win or get 2nd place, it might make you get less frustrated if you aren't so struggling anymore GSP wise! :D
I think that's true too. I think actually your strong finish rating plays into how much your GSP swings beyond the default amount. Total speculation but the results seem to match up.
 
Status
Not open for further replies.
Top Bottom