Register    Login    Forum    Search    FAQ Awesomenauts



Post new topic Reply to topic Go to page Previous  1, 2, 3, 4, 5  Next

Author Message
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Thu Aug 23, 2018 4:20 am 
Offline
User avatar

Joined: Thu Oct 02, 2014 3:16 pm
Posts: 9443
Location: Washington, USA
The Lord Protector wrote:
Xelrog wrote:
I find that--just knowing that I'm moving, that the matches I'm playing matter--to be very rewarding.

Would you still feel the same way if your progress through the leaderboard were primarily down it?

Yes, I would still prefer it to one in which I feel like nothing I do matters. If I'm losing every match and moving down because of it, it comes with the knowledge that if I start doing better, I'll start moving up. There's hope. The current system lacks said hope.

_________________
Burningdillo wrote:


On Steam | On Twitch | On YouTube
Who's Your Awesomenaut?

Mains: :chew: :fist: :ayla:


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Thu Aug 23, 2018 4:27 pm 
Offline
User avatar

Joined: Sun Feb 25, 2018 9:37 pm
Posts: 26
This will be a long post, so for those who want to see the cool bits scroll down to the number crunching and see TL;DR for a quick overview without details.

Introduction

My personal take on this seemingly easy but complicated problem, would be to use metric-based statistic analysis for individual performance. This has been mentioned before and is my most optimistic method.

There are a whole lot of ways to contribute to a match in Awesomenauts, but a big problem is how the game is not strictly class-based. We have damage dealers who can push effortlessly, tanks that can instagib squishies, healers that usually deal more damage than healing, etc. This leads to a more complicated formula for measuring individual performance. We can't simply just calculate damage from damage dealers, damage absorbed by tanks, or kills by assassins.

Therefore, a system that measures skill deviation from a wide variety of metrics statistics can determine how much better / worse a player is performing compared to the average player per 'Naut.

The matchmaking algorithm will mostly stay the same, with the match making rating being used internally, while the new points you earn will be displayed in the leaderboards for progression.

Calculation

We have a legendary Lonestar player who always rushes AA and melts structures. While they die a lot, their structure damage usually out-values their teammates. Here are the stats simplified:

- Structure damage: 12 000
- Kills: 3
- Deaths: 7
- Assists: 4
- ...

By looking at these stats, we have nothing to compare to, but that's where the metrics come in:
- Structure damage: 12 000 (+ 3 365 from average)
- Kills: 3 (-1.4 from average)
- Deaths: 8 (+2.6 from average)
- Assists: 4 (-0.5 from average)
- ...

According to these stats, we can apply a simple percentage to how much better this player played a certain role compared to the average:
- Structure damage: 12 000 (+ 3 365 from average) [+43%]
- Kills: 3 (-1.4 from average) [-32%]
- Deaths: 8 (+2.6 from average) [+43%]
- Assists: 4 (-0.5 from average) [-11%]
- ...

Of course, a lot more data has to be stored. For instance: knockback applied, damage dealt, droids killed, xp gained, etc.

When all the data has been compared, the players rating gain will be slightly modified based on data, (not entirely, since data always has inaccuracy due to human unpredictability, the primary weight in rating gain will always be a win/loss) and could be calculated in three to four steps (notice how I only use three of the stats above for simplicity):

Code:
 - Step 1: Metric deviance = 1 * ( (player.kills / metrics.kills) + (metrics.deaths / player.deaths) + (player.turret_damage / metrics.turret_damage) / 3 )
 - Step 2: Rating disparity = (player4.rating / player1.rating) + (...) + (player6.rating / player3.rating) / 3 [IMPORTANT: Player1 is always the player in question]
 - Step 3: Rating multiplier = 1 * sqrt( (Metric deviance) ) * (Rating disparity)
 - Step 4: Outcome = Rating multiplier ^ result*
 * Where result is equal to 1 for a win, and -1 for a loss.


Basically we first calculate the difference of performance of the player in comparison to the metrics in the first step. Secondly, we use a formula that allegedly is used right now which compares ratings between each individual and sees which team is favored. The final calculation simply multiplies each factor but uses a square root for reduced impact of the metric-based performance. This way it's harder to exploit statistics in order to gain more rating and it mostly comes down to if you win or not.

Speaking of winning or losing, simply multiplying the entire final step by -1 will invert the results and thus a calculation for a loss. I didn't include it for simplicity.

Formula examples

Godslayer420 has won a match against higher rating (more skilled) players, and had 4 crowned stats. Since the player won, the fourth calculation is unnecessary:
- Metric deviance is higher since Godslayer420 has higher performance than average
- Rating disparity is higher since the enemies were of higher skill
- Rating multiplier = 1 * sqrt( (1.5) ) * (1.3) ≈ 1.59
(- Outcome = 1.59 ^ 1 = 1.59)

VirginYoolip has lost a match to lower rating (less skilled) players, and had the most droid kills:
- Metric deviance is lower since overall VirginYoolip had underwhelming performance than average, despite having one better stat
- Rating disparity is lower since the enemies were of lower skill
- Rating multiplier = 1 * sqrt( (0.75) ) * (0.8) ≈ 0.69
- Since this was a loss, the result will be: 1 / 0.69 ≈ 1.45 (increase to rating lost)

Here we can see a distinct difference in how the rating modifier would play out.

Inflation

A very important aspect in the leaderboards would be to control inflation. There will always be those that can play over 1 000+ matches and rack up a load of points. Simply a few regulations and mechanics in the system will address this:
- Rating decay for inactive players
Suggested by many, this could for instance activate after a week of not playing more than 5 ranked matches.
- Reduced rating gain for higher-rating players
A very efficient way of making the leaderboards' top spots have a lower difference in points. This leads to a problem where players at a certain point loses twice as much rating as winning, but with the new formulas this will be mitigated by playing well. Numbers can always be tweaked as well.

With these methods in effect, hopefully inactive accounts with few games will have less of an advantage by playing early season (which also needs changes; placement matches would work but only if queues were split for casual players), while also players who rack up a lot of games don't get free top spots due to a thousand games played.

Inaccuracy

No system is perfect, and most likely there will pop up different ways to exploit the system. By measuring skill by 'Naut performance, there might be some characters that a majority of the player-base struggles with (Yuri for example), and thus playing him well results in more points gained than let's say, Ayla. This leads to misleading results and need to be tweaked accordingly. A data-gathering period when the system launches and utilizing a "test leaderboard" would come a long way for seeing if exploits arrive. One-tricks, premading, soloq, everything needs to be evaluated.

Another problem could be that after splitting Match Making Rating (MMR) with the new point system, it might be better using the rating disparity formula for a different rating system. This can also be changed accordingly.

When a player grows in skill, they will most likely always perform better than the average player, and thus gain more points. If it will be enough to simply reduce rating gain the more rating you have, then this is solved. If not, then there needs to be reduced weight in this calculation for higher-ranked games. Inflation is a serious issue in leaderboards.

Summary - TL;DR

- A new way of measuring skill per 'Naut basis by comparing an individual's performance compared statistics of how the average player performs with a certain 'Naut. Rating gain is then no longer based on rating disparity or whether you win or lose solely, but also on how well you perform.
- These points are the main form of progression, and MMR (Match-making Rating) still remains invisible. Rating calculations will be based on MMR.
- Rating decay will be implemented to penalize inactive accounts. If you play less than 5 games a week you'll start losing rating. This is because you realistically "get rusty" when not involved in the game as actively. While it does sound penalizing, it's mostly not a big deal for casual players.
- A testing phase will be present during the implementation to live of the system, where Ronimo can adjust different 'Naut specific formulas if they seem imbalanced for instance. Think of it like the beta matchmaking period.

Thanks for reading, this was a very improvised post, hope you can understand a tiny bit of the point I'm trying my best to get across!

_________________
Firr/Firren's 2nd acc.
Rankings.
Steam.


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Thu Aug 23, 2018 4:56 pm 
Offline
User avatar

Joined: Tue Jul 28, 2015 10:24 am
Posts: 2554
Location: In my palace, mixing my martinis with paint thinner and a healthy squirt of epoxy adhesive.
This is one of the few examples where I support word for word copying the way Dota 2 does something.

The phrase "copy Dota 2's system" is more than 99.963% more concise, though.

_________________
Say no to SAM wrote:
Give this guy a duck for making such a good job with this thread

Nekomian wrote:
Give this guy a duck lololol

DeezNauts wrote:
Nobody can see it. Maybe instead of asking for ducks, you should put it in your signature.


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Thu Aug 23, 2018 5:09 pm 
Offline
User avatar

Joined: Tue Dec 09, 2014 7:35 pm
Posts: 966
Location: Grenoble, France
SequenceBreaker wrote:
This will be a long post, so for those who want to see the cool bits scroll down to the number crunching and see TL;DR for a quick overview without details.


I think that designing a MMR that takes into consideration your own individual performance in a team-based game, is much more complex than it looks.

In your metric deviance, you assume that each stat is equally important. Getting 20% more kills is as important as getting 20% more assists. Getting 20% more structure damage is as important as getting 20% less deaths. And this does not scale well ; the more you add stats, the less important a particular stat becomes. Say your "..." is actually 20 other stats. Then on your formula, your kill increase would only make up for 1/20 of your metric deviance.

The most important issue i have with this, is that it assumes that 1) one's skill is measurable by simple metrics, and 2) you can write by hand all these metrics and weight them correctly. I think it would be very complex to have a system that measures one's skill accurately enough based on what happens in the match. It's not just writing formulas that look more or less like it ; in order to have an accurate system, you need to do some high level math stats (that are way above my knowledge) and have a proper modeling of the situation.

The win/lose consideration that Awesomenauts does might be simplified, but it is much easier to build an accurate system based on that, since there are already existing algorithms that a lot of people worked on. If you are looking at websites like rankade that provide their own MMR algorithm, the algorithm stays secret for a reason : it is very hard to build one (it requires several very advanced people working on it), and you don't want anyone to steal yours.

And, aside of the fact that the MMR alone is not a good way to rank people ; I think the Awesomenauts MMR does the job very good when it comes to matching! Premades aside, I feel like matches are as balanced as they can be given the low amount of players.

Perhaps something like Hearthstone rank could work ; you would have to defeat enough Rank 12 enemies to go to Rank 11. And if you lose too many matches against Rank 12 folks (once you are Rank 11), you go back to Rank 11. But I think that would require to have two queues : a casual queue where you're matched based on your MMR, and a ranked queue where you are matched based on your rank. And then you have to take premades into account as well.. Very complicated problem + dual queue impossible.

The Lord Protector wrote:
This is one of the few examples where I support word for word copying the way Dota 2 does something.

The phrase "copy Dota 2's system" is more than 99.963% more concise, though.

If Dota 2 system is open source, yeah it makes sense. If not, you can only have a guess at what they actually do.

_________________
Admin of the Rollcage steam group
GRIP - Intense futuristic combat racer


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Fri Aug 24, 2018 1:58 am 
Offline
User avatar

Joined: Tue Jul 28, 2015 10:24 am
Posts: 2554
Location: In my palace, mixing my martinis with paint thinner and a healthy squirt of epoxy adhesive.
Yeah, Valve keep the actual gears of the mechanism under some serious locks to prevent anyone having even the slightest ability to abuse it, but pretty much everyone that I've spoken to thinks it works the way described above. It certainly seems to from testing.

I can't say it's perfect (or even if it'd work for Awesomenauts), but it works pretty well for Dota 2, so there's that. Obviously, it's not going to happen for this Awesomenauts, but a sequel?

Delicious.

_________________
Say no to SAM wrote:
Give this guy a duck for making such a good job with this thread

Nekomian wrote:
Give this guy a duck lololol

DeezNauts wrote:
Nobody can see it. Maybe instead of asking for ducks, you should put it in your signature.


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Fri Aug 24, 2018 7:55 am 
Offline
User avatar

Joined: Sun Mar 24, 2013 2:08 am
Posts: 217
potterman28wxcv wrote:
I don't see how relevant this is. The reason why the 50% got there, is because he won more matches against stronger than him. I guess the Top 250 with 90% winrate are either people who are very good (and like in the top 5), or people who managed to find a time where not many good players are around so they can get a high win-rate, or perhaps they just premade all the way and never actually meet stronger than them as a team ; even though each of them can be weaker individually.

There is just no direct relation between win-rate and MMR.


You don't have to explain to me how they got their rank as if I don't understand. I think that's obvious stuff for anyone that's experienced in the game. My entire point is that the current leaderboards doesn't determine the skill well of the individual player. I'm not interested in some kind of mathematical approach. You are giving multiple examples of how the path of gaining rank can be planned out even when the player is not so skilled. Another method is that 50% win soloq players also spam the hell out of raelynn and get tons of rating for 'important' matches, losing almost nothing when completely alone. In reality being effective playing behind a powerful ally. This is mostly a character issue aswell i won't go into detail. (E.g. Droid spawn still exists)

Quote:
The rating system only makes sense when you actually play the game. The rating is there to match people with or against you. If you don't play the game, you will not change the matchmaking of the players who play the game. If you do play the game ; welcome back ! If your rating turns out to be wrong, you will lose over and over until you get back to a rank more accurate.


Players don't drop to their accurate rank as long as they keep up the plan and stick to it. And eventhough it doesn't affect your rating if they never play in the season again, without decay they do take up the rank slots as long as you are below them.

Quote:
The only issue I see is that if someone plays early season, gets lucky in his match-ups and go say to L2, then comes back 4 months after ; if he actually should be in L6, then he will lose a lot of matches and it will be very painful.

You don't gain rank by luck in the long run, it's being aware of the meta, matchups and how LB works and plan to climb. Before new steam UI you could also see your opponents in lobby which has been used a lot to influence matchups.

Quote:
Question : in how many different teams did you play since the beginning of the season? How many times for each team? As soon as you start to queue with player C who isn't player B, you form a different premade, and would get a completely new rating? That's completely unfeasible, the "premade" leaderboard would be a big mess. And how would you compare premade's rating with soloQ's rating ? Say if you have 3 full premades and 97 soloQs queuing. How do you determine who goes where ? If the rating of the premade is a seperate value from the soloQ rating, there is absolutely no possible comparison.

The rating of the premade should be a function of the ratings of those forming the premade ; not a separate value. Which is what the system does currently if i'm not mistaken.


I'm not saying a seperate system, but a separate board for each time you play in a party. See 3on3 matchmaking in Rocket league for reference. For it to really work it should only put up pre VS pre, and I think the playerbase is too low for that now, but in theory I think it helps determining the skill on ranks better.

Quote:
If you remember the previous win% stats that were published, the naut with the most win% was Sentry with 57% (it was in late 2016). The least was Yoolip with 44%. Most others were between 47% and 53%.
Link to the stats

Perhaps there is a short term issue when a naut is just released and he is broken either because of balance, or because people don't know how to counter him yet ; but a month after and it's fixed most of the time.

In my experience you lose mostly because of the enemy comp, rather than the actual potential of each individual naut.


It's inherently character problem but that doesn't mean the board can't be tweaked. This bulk of data says almost nothing and i'm not surprised the % are close together. It's looking at wins only. If we look at the support characters, raelynn, and yoolip, we see that they can be as succesful as unsuccesful because they lean on the team incredibly hard. People also refrain from playing characters they do terrible with, what happens if everyone played Skolldir for a week? Full premades hardly influence this chart eventhough they win almost all games. The difference between leagues? The data is valuable if we could dit deeper. Experience and metrics are both important.

Quote:
OK. How do you detect whether the alt and the main accounts are played by the same person?


You don"t, but you can lower the incentive to make alts if your account progress mattered more (to more people)


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Fri Aug 24, 2018 6:41 pm 
Offline
User avatar

Joined: Tue Dec 09, 2014 7:35 pm
Posts: 966
Location: Grenoble, France
Japiepatatzakie wrote:
potterman28wxcv wrote:
I don't see how relevant this is. The reason why the 50% got there, is because he won more matches against stronger than him. I guess the Top 250 with 90% winrate are either people who are very good (and like in the top 5), or people who managed to find a time where not many good players are around so they can get a high win-rate, or perhaps they just premade all the way and never actually meet stronger than them as a team ; even though each of them can be weaker individually.

There is just no direct relation between win-rate and MMR.


You don't have to explain to me how they got their rank as if I don't understand. I think that's obvious stuff for anyone that's experienced in the game. My entire point is that the current leaderboards doesn't determine the skill well of the individual player. I'm not interested in some kind of mathematical approach. You are giving multiple examples of how the path of gaining rank can be planned out even when the player is not so skilled. Another method is that 50% win soloq players also spam the hell out of raelynn and get tons of rating for 'important' matches, losing almost nothing when completely alone. In reality being effective playing behind a powerful ally. This is mostly a character issue aswell i won't go into detail. (E.g. Droid spawn still exists)

I did not mean to explain the obvious, but I prefer to precise things a lot so that there is no misunderstanding of how I think, and what allows me to draw what conclusion. It is my way of writing things

The thing is that when you look at the leaderboards, and you see #2 has 62 matches played and 100% win-rate.. How can you design a system where he wouldn't be the "most skilled player" ? Can you play 60 matches and win 100% of them? I am sure that even tournament winners cannot.

What allowed him to have 100% winrate in 60 matches in the first place? 1) Alt account creation (you can create alts infinitely until you achieve to win 60 consecutive matches), and 2) premading with his team in a low playerbase environment.

If we had more players, he would have faced other premades ; and could have lost. But here, when three competitive-level L1 team up and take their best comp, they have to win ; unless they fight in the queue another try hard L1 team.

You cannot solve 1) problem because Awesomenauts is F2P ; and before that, you could use Steam family sharing anyway. You cannot solve 2) neither unless you have a playerbase at least 5 times bigger.

Would decay solve it? I don't believe so. They would just keep playing every once in a while to counter the decay, and would win.

The only solution I see would be to force players to SoloQ. There, you would have a very accurate skill measure system for sure. I don't think people would want that though. And as long as you allow premades, there will always be a way to rig the system within such a small playerbase environment.

In LoL or Dota2 there isn't this problem ; because there are enough try-hard players to have an actual competition at the top, no matter which time of the day you play in.

_________________
Admin of the Rollcage steam group
GRIP - Intense futuristic combat racer


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Fri Aug 24, 2018 9:29 pm 
Offline
User avatar

Joined: Sun Feb 25, 2018 9:37 pm
Posts: 26
potterman28wxcv wrote:
I think that designing a MMR that takes into consideration your own individual performance in a team-based game, is much more complex than it looks.

I'm well aware of that and wasn't the main reasoning behind my post. Data gathering from a game is only numerical and mostly non-logical. It does make it hard to perceive how a player actually performs, but the more stats and algorithms in place (i.e. check if +dmg on bull is active and how much damage was dealt) the more accurate the end result may be. As long as it's including the important stats.

potterman28wxcv wrote:
In your metric deviance, you assume that each stat is equally important. Getting 20% more kills is as important as getting 20% more assists. Getting 20% more structure damage is as important as getting 20% less deaths. [...]

Not really an assumption, more so a simplified formula for how the skeleton of the complex algorithm could look like. Again, there are a lot of holes in my provided formula and that is mostly because I don't have the urge to go into too much detail.

Fair point, however.

potterman28wxcv wrote:
The most important issue i have with this... 1) one's skill is measurable by simple metrics, and 2) you can write by hand all these metrics and weight them correctly

I personally think it's an improvement, but it will need a lot of time to fully function and this simply isn't going to happen but it's just a speculation.

By "simply metrics", I think you misunderstood slightly my intention in the previous post, where there would be a lot of 'Naut-specific data gathering during a game and then check where the player is above the average. Kills, deaths, turret damage, and creep kills would have a higher impact, whereas CCs applied, damage dealt, health restored, etc. would have lower weight. If these would be fine-tuned there would be a way for calculation of skill to be much more precise. Simply win or losses matter the most, but if you would for instance go 20/3 in a soloq match and lose, it would be unfair to lose as much as you do currently.

potterman28wxcv wrote:
Perhaps something like Hearthstone rank could work ; you would have to defeat enough Rank 12 enemies to go to Rank 11. And if you lose too many matches against Rank 12 folks (once you are Rank 11), you go back to Rank 11. But I think that would require to have two queues : a casual queue where you're matched based on your MMR, and a ranked queue where you are matched based on your rank. And then you have to take premades into account as well.. Very complicated problem + dual queue impossible.

I can see this implemented in 'Nauts, but it simply will be too similar to the current system I think. Rating disparity calculation is already present and this seems just like an over-complicated system for a limited playerbase. Just my opinion though. I can't predict how it would play out.

potterman28wxcv wrote:
The Lord Protector wrote:
This is one of the few examples where I support word for word copying the way Dota 2 does something.


The phrase "copy Dota 2's system" is more than 99.963% more concise, though.
If Dota 2 system is open source, yeah it makes sense. If not, you can only have a guess at what they actually do.

There's always a starting point, and you can see where it leads!

EDIT: Formatting.

_________________
Firr/Firren's 2nd acc.
Rankings.
Steam.


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Tue Aug 28, 2018 4:27 pm 
Offline
User avatar

Joined: Sun Jul 01, 2018 5:47 pm
Posts: 24
I have played Dota 2 for a very loooong time & Smite also on & off for quite some game hours among other popular MOBA-s. Both these games have a much larger playerbase than Awesomenauts, & their MMR system also gives you a very profound feeling of promotion & demotion based on your results. While in Awesomenauts, its just totally luck for me when it comes to climbing Leagues. Sometimes I gain points even after I go like 5-10 in a day & other days I lose points after going like 10-5 in that day.

The interesting part is that both Smite & Dota 2 have an MMR system that is pretty linear (i.e. ~ + / - X points for win & loss) & shouldn't be too difficult to implement in the coding at all. The basic problem in Awesomenauts is that there is no specific standard that players can refer to for estimating their growth (in terms of leaderboard) according to their game results, whereas in the other MOBA-s I mentioned, you can actually calculate how much you will climb or fall based on how many matches you win or lose. So in the latter system I would usually end up blaming my teammates playing horribly or something else if I am on a downward slide & try n see if there is anything I can do (different heroes, different objective priorities, etc) to carry those games rather than just putting it off to 'bad luck' & a dysfunctional leaderboard system like I do for Awesomenauts.

Also, this leaderboard DOES NOT determine skill accurately at all & anyone who has played this game for like even a month would know that. I honestly do not know why Ronimo has not listened to the community regarding the leaderboard in spite of taking suggestions regarding new hero releases & other game contents from the community. As far as I understood after some months in this game, Ronimo does not have too good an idea of how the mindset of true MOBA fans work.

Also, I would like to add, I do not think Ronimo will be doing anything about it coz releasing new nauts is a much better option for short term profit rather than working on things like leaderboard that would have a much more delayed reward, & Imho Ronimo has already given up on this game & is just taking all the profit it possibly can from the last few days of Awesomenauts' lifetime.

This could have been their golden egg though & actually could have earned them billions (n also a lot of recognition) but I guess they achieved the target they had in mind for this game.

_________________
Justice or Just Ice?


Top 
 Profile  
 
 Post subject: Re: Reminder that leaderboards need an overhaul
 Post Posted: Tue Aug 28, 2018 4:52 pm 
Offline
User avatar

Joined: Tue Jul 28, 2015 10:24 am
Posts: 2554
Location: In my palace, mixing my martinis with paint thinner and a healthy squirt of epoxy adhesive.
arghyad1 wrote:
...in the other MOBA-s I mentioned, you can actually calculate how much you will climb or fall based on how many matches you win or lose.

I don't know how Smite's system works, but that's untrue for Dota 2.

Dota 2's matchmaking doesn't attempt to force a 50% winrate on any player, so the amount of matches you've won or lost plays no part in how much MMR you'll gain from a win, or lose for a loss.

(All of this is speculative, but highly likely based on testing.)

The matchmaker attempts (and almost always succeeds) to match players such that each team has approximately a 50% chance to win that particular match, based on the difference in "skill" between the two teams' combined abilities. The matchmaker effectively wagers 50 MMR on the outcome of the match. It's also important to note that there's no indication that the matchmaker attempts to end winning streaks, either.

Usually, the matchmaker does a very good job at forming balanced matches and then predicting the outcome, so both teams are equally as likely to win as they are to lose—hence why most matches net you +25 MMR for a win and -25 MMR for a loss. Sometimes, however, there's an imbalance between the win probabilities of the teams, and if the underdog team were to win the match (against the prediction of the matchmaker), they'd gain a greater amount of MMR. If it were a 40% | 60% split, for example, then the MMR gain/loss would be 30 (-20) | 20 (-30) respectively for the teams.

There have been examples of extremely high MMR players losing to much lower-ranked teams, and suffering a whopping -48 MMR loss as a result.

The new ranked medals are similar, but also take into account player statistics (tower damage, hero damage, healing, etc.). The matchmaker also uses these values, as far as we know, to form the teams themselves. Naturally, this means that, for everyone but said extremely high-ranked individuals (and those at the very bottom of the scale), winrates approach 50%. That, however, is only incidental.

I've always been a fan of player performance playing a role in leaderboard progression. It's a small victory for the support players, whose ability to climb is contingent entirely on their allies' competence. I'm not sure how well it would fit for Awesomenauts, though, as 'Naut roles aren't as clearly defined as they are in more traditional MOBAs.

I suppose it couldn't hurt.

_________________
Say no to SAM wrote:
Give this guy a duck for making such a good job with this thread

Nekomian wrote:
Give this guy a duck lololol

DeezNauts wrote:
Nobody can see it. Maybe instead of asking for ducks, you should put it in your signature.


Top 
 Profile  
 
Display posts from previous:  Sort by  
 
Post new topic Reply to topic Go to page Previous  1, 2, 3, 4, 5  Next