Competitive ranking should be based on win rate



  • I know I am thinking pretty far ahead, but when a competitive rank system will be set up, it should be solely based on win rate.
    What is a matchmaking system really meant to do? Set up games where everyone is at similar skill, right? Well that’s just a simplification. Its goal is to set up games where both teams have the same chance of winning. That’s why most ELO systems are basically just glorified win rate calculators.

    So how do we calculate one’s chance of winning? Well, I would argue that your win rate basically is that. When you queue up for a game, your chance of becoming the victor can be estimated at how many times you’ve won/your total number of games. The total win rate of one team should be roughly equal to the total win rate of the other. That means that each team has an equal probability to win.

    What if you’re consistently unlucky and you always get queued up with horrible teammates? Or what if you’re very lucky and you get carried every game (like me)? Well, good thing luck is already factored into your win rate. A lot of whether you win or lose is determined more by your team than yourself (you only make up 1/6 of it after all). This system takes luck into account just as much as skill. In fact, it automatically takes into account every possible factor that wins or loses games. Something that is practically impossible to program.

    Skill is unevaluable. No game has been able to completely accurately determine how good you are at it. They always only know a large ballpark range. The only way around this is to not rank people based on their indeterminable skill, but their win rates. This means there’s nothing in the game that will tell you how good you are, because there’s no way for the game to know that. All it can tell you is your likelihood of winning.

    Here’s an example: Consider these four players:
    Player 1: Avg KDR 3.1 - Win Rate 60%
    Player 2: Avg KDR 0.2 - Win rate 59%
    Player 3: Avg KDR 2.7 - Win rate 40%
    Player 4: Avg KDR 0.5 - Win rate 43 %

    Which two would you match up? Player two looks really bad based on his KDR, right? But then how is he winning so often? Clearly, something else if going on. Player three is also an outlier. He preforms pretty well with a KDR of 2.7, but for some reason he doesn’t win too often. Which pairs would you pick to have the most equal match?



  • W/L should obviously have an impact on the ELO, but I don’t think this is a good idea over all. Imagine a new player getting thrown into a match, wins it due to luck, loses 1 wins one more. Now he has 66% winrate which means he’s palying against people who are A LOT better than he is. I would rather want a pointbased system for the game, but that obviously takes W/L into consideration… somehow.

    Example:

    • Everyone starts with placement matches, and depending on how all of them go, you’re placed into a rank, perhaps this could be based more on personal skill, taking into account K/D and objective points.

    • As for the regular matches each win could give you 25 points and each loss could remove 25 points. which would mean with a 50% winrate you’d stay ±0 which is fair enough because it means you actually belong in the rank you were placed.

    • To add onto that there could be certain bonuses for winning, such as after two consecutive wins you will gain one etra point for every 1k score (or objective points, though objective points isn’t all, I don’t need to push the payload if I kill everyone trying to stop it before they get there) you earn in the game up to 5 max.
      Thiswould mean that the first match you win is always 25 points, and if you do well in the next map and win you may climb slightly. If you lose, you lose 25 points and the next win will be max 25 points again.

    • Another way would be to have 25points added/reducted per win/loss, and then just giving bonuses or penalties depending on wether or not you win several times in a row, example after 5 wins you might get 50 points per win and only lose 10. Though this rewards streaks more than it does overall W/L.

    As you mentioned KD might not matter as much as winning, but if I go 19-5-12 KDA and most of my team mates have 3-11-2 KDA and I’m actually playing the objective and we get a close loss, then it’s not really my fault.
    But there’s also the opposite, the meatshields, aka the people who are bad at the game but still has to be killed, they can still get a great W/L ratio due to getting lucky with team mates. I notied in quite a few of the games I’ve played which have been pretty one sided there’s still someone who manages to get a KD of 1 or less on my team and I have no clue how.
    And that is why I think something that rewards your score, either by making you lose less if you have a really good score, or giving you some extra points if you win due to your score.



  • @Xylvion The problem about giving people points depending on score is that people might play differently to chase the score that gives more points that may result in playing overly defensively so you wont die and it will get frustrating if the enemy team is for some reason hardcore focusing you (maybe they realise you are the best so they want to get you out of the way fast) and then you will lose more points.



  • I completely agree with placement games. Of course new players shouldn’t be thrown right in with the 60% win raters.

    What I want though, if for the only incentive in a match to be winning it. Nothing else should matter. The points system you described in the last bullet kind of accomplishes this. (Also yeah it makes sense that winning multiple times in a row increases the number of points you get)
    I’m totally fine with using a points ELO system instead. It’s basically just your win rate disguised as another number though.

    I just don’t think it’s possible to determine how much of the loss is “your fault”. I’ll use your example again, but put in a different context.

    Let’s say you join a team, and you are the most toxic player out there. Your teammates are usually pretty good, but because of you, they play really badly. You try to go after the objectives and get the highest KDA that you can, and all of your teammates have KDAs of 3-11-2. Of course you lose. Is it really fair now to punish you less, just because you got more kills?

    Here’s an opposing example. You’re having a tough game for whatever reason. Your KDA is like 6-12-8. It’s a CTF map, and you’re about to lose. You are the only one on your team that’s left alive and you’re at the bottom of the leaderboard. But then miraculously, you kill the flag carrier, run it to their capture point, and win. You essentially won the game for your team, but that’s not reflected in any numbers.

    The only way is to have everyone share equal responsibility for every win or loss.

    Think about it like this. Your team has a hero. Someone who will get a massive KDR and who will carry their team. The team will also have some peasants who just help where they can. Everyone is just fulfilling their role. It doesn’t matter what role you take, just if the role you take results in victory.



  • Let’s say you join a team, and you are the most toxic player out there.

    that would mean I had to be toxic for this to be true though, so it doesn’t really make any sense to me.

    But then miraculously, you kill the flag carrier, run it to their capture point, and win. You essentially won the game for your team, but that’s not reflected in any numbers.

    Except score for defending your flag area, score for killing the person with the flag (perhaps even extra because it’s the flag carrier), score for capping their flag area.
    Which in turn would result in higher score.

    IF you take a look at multi objective drydock, 1 kill = 1 ticket removed from your enemies life, 1 pushed ark is 20, 1 capped glyph is 10. you gain like 5 score for each 0.5seconds you push the payload which adds up and becomes a lot.
    Now 5 score might not be enough, but with a score system for extra ELO points more people would obviously focus on the objective if it gives you MORE SCORE than getting kills. Besides if you lose you still lose points, it’s not like you’d gain anything for losing. But if you won the previous round before the loss you might end up only losing 20 points instead of 25 if you did really really well; that includes kills and assists, but also objective points.



  • I just don’t like how that creates incentive towards things that are not directly winning. Sure, going for the objective leads to winning, but it’s not necessarily optimal.

    To imagine what I’m talking about: I remember trying to go for king on the second objective of Stoneshill. I remember thinking, how can I let them win while also getting as high score as possible. This is a terrible psychology that I’m trying to avoid. What I did was sit next to the objective, not helping my team at all but still racking up the objective points. I would go for easy pickoffs that I knew wouldn’t hinder their push too badly. And I got king. My KDA was great, and I had a bunch of objective points. We still lost.

    Unless we make the only way to gain ELO winning directly, people will find ways to get around it. They won’t try to lose like I did, but they’ll be going for the route that gives them optimal score, not optimal chance of victory.

    As for my example about the toxic player, I didn’t mean you specifically. I just meant that someone could be the reason their team loses while still having the best numbers.



  • But that’s what I’m saying fam, you wouldn’t gain ELO from losing. Losing default -25 ELO, winning default +25 ELO. If you won the previous game and you win this next one and get 5k+ score you gain +30 ELO. If you won the last game but lost this game but with 5k+ score you would lose -20 ELO instead of -25. YOu would still LOSE, but you’d lose less. As I mentioned 1ELO per 1k score you manage as well and a max of +5.

    You wouldn’t sit around and gain score for nothing, because then you’d still be losing ELO points, but you’d be losing less points if you get a better overall score than if you get a worse.



  • Okay well one or two points out of 25 isn’t going to make a big difference anyways.

    I guess your system still incentivises winning, since you have to win the previous game to get your bonus this game.
    Alright. I agree. This could work.
    As long as it’s only a small fraction of your score (Like 5 out of 25) and if it’s only if you won your previous game.

    Edit: The point is, TB, don’t make it like “Oh however many kills you got in your previous game is your new rank lolz”



  • I’d suggest just using rank. For those of you who played Quakelive It had a stats system with every kind of weapon plus
    a total score which transferred to stats and your ratings elo… As I ran Quakelive servers for over 1000 hours ( only just over a year ago)
    I watched
    players constantly joining just to play the stats, meaning players would sit out in spec to join the better stat’ed side
    hence stacking game even more, players would get upto all sorts of other stuff just to get good stat/elo rankings etc.

    IN mirage by just going by Rank you have an element of some rank players of the same rank are infact better or worse.
    so its’ not so obvious who is the better player at that rank depending on how some earned that rank, ie one might be
    a ranged weapon expert but nowt good in combat close up etc.

    Quake live died off simply because it went stat mad as did the players thus nobody would play for fear of being
    on a low stats/ elo ranked side etc. Also I don’t think Mirage could have such a complicated stat / elo system
    as each weapon , its accuracy, kills / deaths, speed ranking plus all the other micro stats to keep track of tbh
    where as if you want to check this out http://www.cromctf.com/vsp/pub/themes/bismarck/index.php?config=crom_ctf.php&skin=evilsmurfs

    That infact gives you an idea of how indepth stats/elo has to be to work. The later Quakelive version died off for the reasons
    I mentioned . Thus that link I gave for crom is a much older easier one but I still doubt TB would really want such an indepth
    pointless stats/elo system if it was even possible in mirage tbh.

    Go by Ranks only is my suggestion, it eliminates the over complicated stat keeping plus mirage won’t just end up as Quake live did
    simply dieing off and closing its servers down and removing players stats and friends lists suddenly . Gamers used to join a match
    but after a good kill streak head to spec in case they got killed or even leave server , such is it’s drawback and effect on the game.

    By using just a rank system variance of about 3 up or down per side even , lastly by just a rank system you won’t get pro players
    as you did in Quakelive jumping on lower stat/elo servers to rake up more stats, etc

    I forgot to mention the servers only ran Capture the flag modes or Team modes hence 8 to 15 per side played near every day .
    AS you know in team based games being on mostly every day watching being the admin of my servers, as did a few others
    I trusted, my impression of what stats of such complexity did to the game was first hand. the name stat whores was a very
    common argument in many games, in it’s extreme if a very good high rated player swopped sides to even the match
    then quite simply the former strong sided team would start to complain and leave the server, you see



  • I should have stuck this at the start however players would start a ctf / match in Quakelive but switch to window mode
    while checking the enemy players stats/elo , Thus if they saw higher ratings, accuracy’s etc then would switch sides
    or stack more or even leave servers. With just a simple ranked system only , it wouldn’t encourage such behaviours.



  • @Wilt exactly, though.
    With a system that only (or very heavily as Xylvion wants it) depends on whether you win or lose, it won’t matter as much about how many deaths you have or even how many kills you have.

    Sure, you still have to worry about if you’re winning or not, but that doesn’t sound nearly as complicated as what Quakelive has.

    Of course using a value that can only increase (Like your rank in MIrage or how many wins you have) is another option, but it’s on the complete other end of the spectrum of what we’re thinking of. That would certainly relieve a lot of stress around competitive. I’m not sure that’s a good thing though. Matches still need to feel like they have weight behind them.

    I think the general consensus around competitive games is that there has to be some sort of ELO that can increase and decrease to accurately find even teams.



  • Yes a Rank system would be on the surface as former chivalry players are used to that plus it should suit newer players.

    On the surface it appears as a simple rank system but underneath the rank calculator could take into account a lot more
    ie accuracy , defence assist Kills+ Deaths Suicides Eff % Hits Shots Misses Acc % etc.

    So ranking up takes a lot more into account behind the scenes in Mirage, however only the players Rank is visible
    by other players.

    A file breaking down your stats in depth could be written to on a players machine however.
    A file they can reveal to perspective clans if needed in recruitment .

    So as you suggest the matches would feel to a player at least they had weight behind them, if behind the Rank
    system was a much more complex indepth system working out your points towards ranking up idd. etc

    Imo what ever system you have the trick is keeping a lot of the info private. If mirage went down the path of
    displaying every part of a players stats then you have a system like Quake live had before it died off.
    why bother testing players when you can just pick them up from such an indepth viewable stats/elo system?

    With a Rank only visible system employing another system behind it be that very simple stats or very detailed
    Mirage retains an air of mystery about players skills while keeping the sense of weight as you say private to a player etc.



  • @Wilt said in Competitive ranking should be based on win rate:

    On the surface it appears as a simple rank system but underneath the rank calculator could take into account a lot more
    ie accuracy , defence assist Kills+ Deaths Suicides Eff % Hits Shots Misses Acc % etc.

    You know what automatically takes all of this and everything else into account? Win rate.

    I mean, Chivalry’s rank system should definitely still be kept in the game. I just don’t think competitive would feel like competitive if absolutely nothing was at stake.

    Edit: This is kind of unintuitive, but you know the competitive system in a game is good when people are getting mad that they lose. With Chivalry’s rank system, winning or losing really didn’t matter. It was just one giant blur of farming kills. That’s not what competitive is supposed to feel like.



  • I would suggest a players Rank could go down as well as up In any form of Mirage stats system:

    that would encapsulate the competitive system you mentioned just. behind the scenes other factors could be taken into account as you mention even win rate, ( more for ffa imo then team based modes) . It would be quite easy for TB to enable Rank to Decrease/Increase
    until if they choose to add a much more complex indepth system, even with a website breakdown. A simply Rank up/down would get
    the thing rolling even from the next round of testing infact. If it was very successful then they might not even need to introduce any
    other ranking system of stats: Also by this if your beaten by an enemy team or solo you might not be paired with them again
    until your Rank is built up again , thus keeping both parties interested plus with almost equal skill lever opponents etc…



  • @Wilt That’s definitely the key. They’re going to have to try stuff and see what works.

    Also, they haven’t revealed much about the future competitive system, but I kind of assumed it would only include team-based modes.



  • Although there are plenty of reasons why including individual skill as a factor is desirable, I would like to provide a few reasons why this system creates problems that are more detrimental to the competitive experience overall.

    Let’s look at Overwatch, which used a competitive ranking system during its first season that included personal performance. On characters such as Genji and Tracer (assassins/flankers), earning lots of kills/assists helped you to climb faster while also diminishing the impact of losses. However, for characters such as Mercy and Lucio (who are healers), they gained personal performance benefits through healing and reviving. This created a paradox for healers and other supportive playstyles. When their teammates did well, they didn’t take as much damage and these healers couldn’t heal them for as much. This meant that even when a Mercy main had the same winrate and contributed as much to the success of their team as a Genji main, they would not climb as fast because their personal performance (healing allies) was disincentivized when their teammates played well (and didn’t take damage). However, because healers are necessary for a competitively viable team, you basically ended up with one or two supportive players on a team who would have to “take one for the team” and play a character who would not allow them to climb as much as they deserved for winning. In turn, this led to fewer players choosing to play healers and more frustration among the last people on a team to pick, since they were often “stuck” with a bad role. How does this relate to Mirage, you might be wondering?

    For example, consider the supportive Entropist class (yes, I know the Entropist is unavailable this build and is getting reworked, but as of last build it had a healing grenade spell). Hypothetically, this player should be supporting teammates with healing grenades while also harassing the enemy and occasionally using parkour or the magic carpet to escape. If getting lots of kills becomes a means of climbing the ranked ladder faster, playing the Entropist is no longer incentivized. In fact, even if the Entropist was rewarded for landing healing spells on allies, it would be punished by this same paradox found in Overwatch in which a winning team simply doesn’t need as many heals as a losing team. This will in turn lead to fewer players choosing Entropist and there will be more frustration for players who enjoy supportive characters – by the way, this would include to an extent the Vigilist and her Iron Dome spell.

    Rewarding good stats also leads to a stagnation of the meta. Let’s say hypothetically that there is an Alchemancer player who has exceptional parkour skills and defensive positioning but poor aim and predictable attacks. This player might still find a niche with a simple winrate system. By flanking enemies and harassing them with ranged attacks then escaping with spells/parkour, this player could distract multiple enemies at once and thus enable the rest of their team to win an uneven teamfight. This is an example of a player who perhaps goes 0-5-2 using a subversive strategy to win, but this kind of strategy would be dismissed under a system involving individual scores. The competitive meta as a whole will stagnate once statistics become a sought-after factor in winning. Why would anyone practice a harasser parkour-heavy Alchemancer strat when they could spend their time learning to simply go ham as a Taurant?

    Lastly, competitive ranked modes in games like these will always bring a certain amount of frustration. Yes, it is certainly frustrating to go 25-6-14 and still lose a full -25 points because your teammates played terribly. But, if you’re consistently going 25-6-14 then you will almost certainly be climbing by default unless you are making really, really poor strategic decisions or (like mentioned above) you are flaming your teammates and making them play worse. Furthermore, the unintended consequences of personal performance systems will remove incentives to play supportive and subversive playstyles thereby causing the meta to stagnate.

    TLDR: Win rate should be the only factor in ranked modes. Individual performance will cause the meta to stagnate by punishing supportive and subversive playstyles.

    PS: For anyone who suggests that the meta will eventually stagnate anyway if there are no balance updates, go check out competitive Super Smash Bros Melee and you’ll see that players are still innovating and the tier lists are still changing with every passing year, even after over a decade without a single update to the game.



  • I thought of something else, too.
    Let’s say your team is completely demolishing the other. At the end of the game, you have to choose between winning it, or spawncamping for kills. Kills being at all a factor for your rank punishes players for trying to end the game. If winning is next to guaranteed, it’s always a better idea to just try to rack up as many kills as you can before the timer ends.

    There shouldn’t be mechanics that encourage this imho.