Addressing the Potato problem via MM restrictions

[Taichunger]

People have been suggesting that the MM -- which is doing great lately, praise to WG for the recent adjustments -- be set to recognize win rates, because they don't want to be on potato teams. I don't either, but I like playing in T4-8, it's fun.

 

The objection to this is obvious: pretty soon everyone would have a 50% win rate. The other obvious objection is that many of us have low win rates pulled down by their potato days at 42% (like me).

 

What if we stratified the MM not by win rate but by experience, during peak hours? For example, when the server population is over 8000, the MM is stratified into three groups: players with 0-1500 (or 2000) games of experience, 1500-3000 (or 4000) games, and over 3000 (4000) games.

 

That would go a long way to reduce potato teams that collapse immediately, and experienced players could be assured they were only playing with experienced players. That wouldn't entirely solve the problem, since potatoes are fertile and grow at all levels. But it would help.

 

Collapses can't be eliminated, because they are built into the map/spawn dynamic, particularly on the Strait, Shatter, Two Brothers, and Hot Spot maps, as well as in the way accumulations of small differences in RNG soon turn into massive differences in ship kill performance. But perhaps they can be reduced....

 

[Aduial]

No, because i like reking potatoes and making delicious mashed potatoes out of them.

 

[Quaffer]

No.

 

It is currently random. That means that inexperienced players have an opportunity to learn from experienced players. If you want to play with better players make an effort to teach them. This is usually most effective if you don't insult them.

[UnseenSpectacle]

Nope. A potato can be a potato at 10 battles or 10,000 battles. I would be very dubious of any statistical correlation between potatoes and matches played once you get outside the learning curve. You will always have this problem short of implementing player skill ratings for the MM to try to use and create balanced teams. I don't get my hopes up for player skill ratings either because frankly this game has too small a population logged in at any particular time to effectively and efficiently handle tiering, divisions, other MM restrictions, and player skill matching. The best you could hope for with the game's online server populations would be MM having teams with similar numbers of potatoes vs non-potatoes. However, validation of rating systems is an exhausting effort and is not something you just slap in overnight.

[SPACEDUDE71]

what u seem to want is team battles.

soon.

[Taichunger]

I hope so....

[Capt_Mexib]

Au, No Boyo, we need the potato harvest! 

[iEaTDaGrudge]

People have been suggesting that the MM -- which is doing great lately, praise to WG for the recent adjustments -- be set to recognize win rates, because they don't want to be on potato teams. I don't either, but I like playing in T4-8, it's fun.

 

The objection to this is obvious: pretty soon everyone would have a 50% win rate. The other obvious objection is that many of us have low win rates pulled down by their potato days at 42% (like me).

 

What if we stratified the MM not by win rate but by experience, during peak hours? For example, when the server population is over 8000, the MM is stratified into three groups: players with 0-1500 (or 2000) games of experience, 1500-3000 (or 4000) games, and over 3000 (4000) games.

 

That would go a long way to reduce potato teams that collapse immediately, and experienced players could be assured they were only playing with experienced players. That wouldn't entirely solve the problem, since potatoes are fertile and grow at all levels. But it would help.

 

Collapses can't be eliminated, because they are built into the map/spawn dynamic, particularly on the Strait, Shatter, Two Brothers, and Hot Spot maps, as well as in the way accumulations of small differences in RNG soon turn into massive differences in ship kill performance. But perhaps they can be reduced....

 

 

Not a bad idea and it will make experience player less salty ...Imagine u ending up to a potato team that cant cap or even a BB chasing a dd..awsome right hahahah welcome to reality.

[Taichunger]

No.

 

It is currently random. That means that inexperienced players have an opportunity to learn from experienced players. If you want to play with better players make an effort to teach them. This is usually most effective if you don't insult them.

 

Thats an excellent point. 

[Takamachi_Nanoha_sg]

There is one critical issue with perfornance based MM...

 

No matter which indicator or indicators are chosen for MM balancing, it will bring everyone's figure towards a balanced number. When this is achieved, this MM stops working as everyone is about the same to the MM logic. Then the teams will again comprise random combination of potatos and tomatoes....

 

However this could be used to competitive modes like ranked maybe?

[UnseenSpectacle]

There is one critical issue with perfornance based MM...

 

No matter which indicator or indicators are chosen for MM balancing, it will bring everyone's figure towards a balanced number. When this is achieved, this MM stops working as everyone is about the same to the MM logic. Then the teams will again comprise random combination of potatos and tomatoes....

 

However this could be used to competitive modes like ranked maybe?

 

This is why when you design such systems that you have to have a way of rating player skill that is not dependent on WR. This is precisely why ELO and other such systems exist and are used by other competitive online games. The only problem with putting them into WoWs is the active player base is currently not large enough to cope well with the further fragmentation that would occur by using such system along with MM's existing rule set. Ranked would be an interesting experiment but they have to develop a rating system first that can robustly cope with a smaller player base. A well designed system would allow you to decouple ranked battles MM from the star progression system. In comparison, Team Battles ratings are easy because they rate the whole team and not just the player. Player ratings become more complicated because not only are factoring in how you expect the player to fair against the competition but how the skill of the player's team and its composition should affect them as well. Like I said before, developing a well thought out and validated rating system is not something that happens overnight.

[Crucis]

There is one critical issue with perfornance based MM...

 

No matter which indicator or indicators are chosen for MM balancing, it will bring everyone's figure towards a balanced number. When this is achieved, this MM stops working as everyone is about the same to the MM logic. Then the teams will again comprise random combination of potatos and tomatoes....

 

However this could be used to competitive modes like ranked maybe?

 

 

I'm not entirely sure that this would be true if MM tried to balance teams by average XP.  I haven't given this much thought, so let talk this through.

 

It seems to me that even if MM balanced teams with average base XP, better players would still earn more on average and weaker players would still earn less on average.  It's possible, maybe likely, that this might cause players' win rates to start moving towards 50%, even though WR wasn't used for balancing teams.  Why?  Well, if you had two teams that were reasonably well balanced in terms of ships and average base XP, arguably neither team has any initial advantage in overall skill, assuming that average base XP is a reasonable measure of skill.  And if this is the case, it would seem logical that over the long haul, players' WRs would gravitate towards 50%.  Am I making any flawed assumption or logical error here?  I don't know.

 

I don't know if players' average base XP's would improve or not.  Oh, I suppose that weaker players' ABXP would improve a little on the assumption that their WR's would be improving as they headed towards 50%, while better players' ABXP would decline some on the assumption that their WR's would be falling towards 50%.  And since winning and losing does affect one's base XP, the higher one's WR is, one should expect that their base XP is somewhat better if only due to winning more often.  And vice versa.  So if WR's were heading towards 50%, ABXP's would rise or fall a bit based on winning more or less often.  But that still wouldn't mean that everyone would be heading towards a single ABXP server average number because better players would still perform better in battle than weaker players.  You just might end up with a very good player's ABXP not being as high as it is now, and vice versa on the lower end of the scale.  And while all players might be have roughly 50% WR's, the ABXP's would be reflective of their talent and production.

 

 

[UnseenSpectacle]

It seems to me that even if MM balanced teams with average base XP, better players would still earn more on average and weaker players would still earn less on average.  It's possible, maybe likely, that this might cause players' win rates to start moving towards 50%, even though WR wasn't used for balancing teams.  Why?  Well, if you had two teams that were reasonably well balanced in terms of ships and average base XP, arguably neither team has any initial advantage in overall skill, assuming that average base XP is a reasonable measure of skill.  And if this is the case, it would seem logical that over the long haul, players' WRs would gravitate towards 50%.  Am I making any flawed assumption or logical error here?  I don't know.

 

Overall, it is a reasonable assumption to start with and I would posit that there would be a correlation between base XP and player skill. However, like all things, the devil is in the details. First, you will need to normalize it against the tier the player has played. Additionally, are you looking at a player's base XP in a particular ship/class or only their overall account? The easiest way to implement would be using the player's overall skill. However, ignoring a player's ability in a class or the tiers they have played will just introduce different types of potatoes into the game. For instance, if you went by my overall average, I would probably get an easier MM than I deserve because only now are most of my lines hitting Tier 7 or 8 and I play relatively few games compared to most players in the game. Likewise, if I wanted to play CV (a class I never play) and you used my overall average while ignored the classes I've played, I would likely get extremely unfavorable matchmaking that would pit me against a much more skilled CV player. So, once again, developing a system sounds easy until you start thinking through the details of how the systems would work and how to make them perform better in reality than the existing system. While we may not like the current MM it is important to remember that complex systems can lead to poor results in reality while on paper they are superior to the "simple" system being replaced if they are hastily implemented and not thoroughly vetted.

[Skpstr]

 

This is why when you design such systems that you have to have a way of rating player skill that is not dependent on WR. This is precisely why ELO and other such systems exist and are used by other competitive online games.

 

I'm curious though, would ELO be any more effective in a game with (relatively large) random teams? Which games with a similar dynamic use ELO?

[UnseenSpectacle]

 

I'm curious though, would ELO be any more effective in a game with (relatively large) random teams? Which games with a similar dynamic use ELO?

 

ELO is just one example of a rating system. Closest analog that uses an ELO-type system effectively is League of Legends. Please note that I mean closest in the sense that it uses a rating system for random queue and the player is stuck on a random team with up to 4 other people. It's been awhile since I've played LoL so I can't comment on what their system is today. However, they have so many people playing that game that the player base does not really notice the fragmentation that an ELO rating system produces because they always have enough other potential competitors in the acceptable ELO range to reliably get a match. I don't think WoWs could use the LoL ELO system unless they used larger ELO bands or other compensating mechanisms because the player base is too small and you have other game mechanics (such as tier restrictions) that fragment the player base. Just bolting a straight unmodified ELO system or a system designed for another game onto WoWs random queue would be the wrong way to go. The system could derive ideas from other rating systems but has to be tailored to the unique environment that is WoWs. Overall, having MM be able to pair up relatively skilled teams would be the easy part. The hard part is developing a system that can reliably determine a player's skill in the game and produce a number for MM to use when trying to create a match.

 

 

Overwatch's skill rating system for competitive play also comes to mind. However, like any system, that system has its flaws but some of the concepts used in the rating system would apply readily to WoWs.

[IronWolfV]

Oh goodie, this dead horse makes a come back. SBMM in an already low player population is a BADDD idea. For multiple reasons I am not going to dive into. I am with Quaffer.

 

Player education is the better route.  A potato, will always be a potato unless someone shows him the way.

[Ajax_the_Great1]

In all walks of life, steps are taken to ensure fun and competitive games. From red rover, to pick up basketball, to high school sports, to the NFL, people try to make the games fun and competitive. 1 teams picks a guy, and then the other team and so forth to try and make things even. It's how sports work.

 

But for some reason, every time someone suggests something even remotely similar to that for a video game, the forums freak out and act like that would ruin everything.

[Skpstr]

 

Please note that I mean closest in the sense that it uses a rating system for random queue....

 

And how is the rating arrived at? What makes it go up or down?

[IronWolfV]

In all walks of life, steps are taken to ensure fun and competitive games. From red rover, to pick up basketball, to high school sports, to the NFL, people try to make the games fun and competitive. 1 teams picks a guy, and then the other team and so forth to try and make things even. It's how sports work.

 

 

But for some reason, every time someone suggests something even remotely similar to that for a video game, the forums freak out and act like that would ruin everything.

 

It's already like that, MM takes a look at the pool, and merely divides it up to equal out the tiers inside of the battle weight.  Love how people cry "it's not fair" when it already is.  Everyone gets treated equally under MM.

[UnseenSpectacle]

 

And how is the rating arrived at? What makes it go up or down?

 

I do not have the time to detail it at the moment but here are some links that go into how the LoL system operated when it used ELO and also the current League system. Keep in mind though that their system works well for them because they have the population that enables their rating system and matchmaking to be effective.

 

http://leagueoflegends.wikia.com/wiki/Elo_rating_system

http://leagueoflegends.wikia.com/wiki/League_system

https://support.riotgames.com/hc/en-us/articles/204010760-Ranked-Play-FAQ

[VGLance]

My MM algorithm solves every issue from not affecting wait times to not causing everyone's stats to become the same. Win rate will move closer to 50% but that's a good thing because it means teams are more balanced. The cognitively impaired will eventually let go of win rate as a measure of skill and adopt individual ship WTR (tweaked if necessary) as the standard for expected output. 

 

The concept is simple, think of a pickup game at the park, the 24 players have already been selected like MM does now, but afterwards, it assesses the total WTR of both teams and if it's outside a threshold of 45%-55%, it swaps even tier/class ships from the many non-divisioned players until it hits that threshold. This calculation happens so fast, the human eye doesn't even see it. 

 

People against this fail to comprehend how it doesn't affect queue times and / or fail to let go of the idea that balance can only be achieved with exact class/tier ships mirroring each other or requiring a flawless 50-50 WTR, when that is not necessary. The point of MM balance is to eliminate the 90-10% chance matches that are over before they started. 

[IronWolfV]

My MM algorithm solves every issue from not affecting wait times to not causing everyone's stats to become the same. Win rate will move closer to 50% but that's a good thing because it means teams are more balanced. The cognitively impaired will eventually let go of win rate as a measure of skill and adopt individual ship WTR (tweaked if necessary) as the standard for expected output. 

 

The concept is simple, think of a pickup game at the park, the 24 players have already been selected like MM does now, but afterwards, it assesses the total WTR of both teams and if it's outside a threshold of 45%-55%, it swaps even tier/class ships from the many non-divisioned players until it hits that threshold. This calculation happens so fast, the human eye doesn't even see it. 

 

People against this fail to comprehend how it doesn't affect queue times and / or fail to let go of the idea that balance can only be achieved with exact class/tier ships mirroring each other or requiring a flawless 50-50 WTR, when that is not necessary. The point of MM balance is to eliminate the 90-10% chance matches that are over before they started. 

 

And here is the stumbling block that so many people run into, same as in tanks. Wargamming has not and never will recognize WTR, the WN system from tanks like Wn8 for the simple fact of this. It's a 3rd party algorithm.

 

Sorry but basing SBMM of 3rd party is never EVER going to fly.  Nice try though.

[UnseenSpectacle]

My MM algorithm solves every issue from not affecting wait times to not causing everyone's stats to become the same. Win rate will move closer to 50% but that's a good thing because it means teams are more balanced. The cognitively impaired will eventually let go of win rate as a measure of skill and adopt individual ship WTR (tweaked if necessary) as the standard for expected output.

 

The concept is simple, think of a pickup game at the park, the 24 players have already been selected like MM does now, but afterwards, it assesses the total WTR of both teams and if it's outside a threshold of 45%-55%, it swaps even tier/class ships from the many non-divisioned players until it hits that threshold. This calculation happens so fast, the human eye doesn't even see it.

 

People against this fail to comprehend how it doesn't affect queue times and / or fail to let go of the idea that balance can only be achieved with exact class/tier ships mirroring each other or requiring a flawless 50-50 WTR, when that is not necessary. The point of MM balance is to eliminate the 90-10% chance matches that are over before they started.

 

Your point has a lot of merit. However, WTR itself is a flawed rating system. That is no fault of the creators. As I always I said at work, "If all you have is a hammer, everything looks like a nail." In the instance of WTR, the data coming out of WG for stat sites is so limited that the best thing 3rd parties can come up with is a system based on damage, kills, and win rate even though there is much more to the game than those three metrics. TalonV is also correct. If WG ever wanted to implement an MM that considers skill, they would no doubt develop their own proprietary rating system.

[Skpstr]

In all walks of life, steps are taken to ensure fun and competitive games. From red rover, to pick up basketball, to high school sports, to the NFL, people try to make the games fun and competitive. 1 teams picks a guy, and then the other team and so forth to try and make things even. It's how sports work.

 

 

But for some reason, every time someone suggests something even remotely similar to that for a video game, the forums freak out and act like that would ruin everything.

 

The problem is, IRL, you want to choose the best players. Whatever determines that is typically universally accepted, whether you pick Joe for Red Rover because he's big and fast, or you draft him to an NFL team because of his yards gained in college.

 

You can't get universal acceptance here for any statistic that indicates skill.

 

Also, especially in pro sports, picking the best is fair, because whoever the best player is, the next best is going to be very good too, especially since different teams need different positions.

 

Imagine an NFL draft where everyone had to select a new QB. How good is the first pick going to be compared to the last?

 

In WoWS, the issue is population. Say pick #1 is a BB with 59% WR. How quickly are you going to find a matching player? In WoWS terms, an NFL draft could see the next team picking a 59% DD to equal that. That's not going to work in WoWS, to make it fair, you would have to match players by skill AND ship type. With the population we have, that means a long wait if a really good or bad player has to be matched.

 

You could have the criteria loosen up after x minutes of queue time, but that's going to give you the same results as some of the mismatches we get now, where one team gets 1 or 2 "outlier" players.

 

Also, given the skill distribution we have, most matches are pretty fair anyway, as most of the players in a match will have skill levels near average.

[VGLance]

You guys are missing my point, it's not the WTR (which is why I specifically added "tweaked if necessary").  It's the fact that WG has the data, they don't need a third party.  It's a rating similar in concept to WTR as opposed to using win rate or battles played or other horribly worse metric.  Stop getting caught up in the semantics and focus on the concept as a whole.