Can win streaks be a thing?
This is already baked into Glicko2. If you’re consistently not performing at expectation (either above or below), your rating will change faster.
Kirrena Rosenkreutz
No it’s not, that’s not how Glicko2 works.
No because win streaks are mainly luck on where the losses are placed.
No it’s not, that’s not how Glicko2 works.
Care to point out where I’m wrong?
Kirrena Rosenkreutz
Are you for real? GW2 uses rating system invented for chess which is a 1v1 game? xD Anet keeps on surprising me.
Are you for real? GW2 uses rating system invented for chess which is a 1v1 game? xD Anet keeps on surprising me.
So what system would you propose? Rating systems for competition draw all their roots to chess rating systems. And many games use Glicko for rating systems.
Kirrena Rosenkreutz
Are you for real? GW2 uses rating system invented for chess which is a 1v1 game? xD Anet keeps on surprising me.
It does, that is why it is such an inaccurate system for a team game. It is a really terrible system for team games no matter what some pvp people try to tell you.
It might work if people would be forced to make static 5 man teams and the whole team gets a MMR. But it will never work for solo players in a team game.
There is also no system in place that forces lets say rank 1 and 2 to play against each other to figure out who is better. If you think someone is too strong just queue dodge him or play in off peak hours.
The whole MMR system offers nothing for people who are not top tier players. The rewards are either participation trophies (PvP tickets, gold, ascended shards of glory) or totally out of reach for average players. In previous seasons you could at least put in effort and work towards the seasonal legendary title. People say there were grindable. Maybe true, but at least they kept motivation up. The current system does everything to make me demotivated from putting effort in.
Also it is funny that some people who struggled (or even failed) to get out of Ruby in previous seasons are suddenly some hundred MMR higher than me. Talk about MMR being representative of skill.
Tl;dr: The current season is the worst one ever unless you like getting participation trophies handed over to you in record time (around 28hours to get all season exclusive rewards) or are a top tier player who can compete in top 250.
1st person worldwide to reach 35,000 achievement points.
(edited by Malediktus.9250)
Are you for real? GW2 uses rating system invented for chess which is a 1v1 game? xD Anet keeps on surprising me.
It does, that is why it is such an inaccurate system for a team game. It is a really terrible system for team games no matter what some pvp people try to tell you.
It might work if people would be forced to make static 5 man teams and the whole team gets a MMR. But it will never work for solo players in a team game.
There is also no system in place that forces lets say rank 1 and 2 to play against each other to figure out who is better. If you think someone is too strong just queue dodge him or play in off peak hours.
The whole MMR system offers nothing for people who are not top tier players. The rewards are either participation trophies (PvP tickets, gold, ascended shards of glory) or totally out of reach for average players. In previous seasons you could at least put in effort and work towards the seasonal legendary title. People say there were grindable. Maybe true, but at least they kept motivation up. The current system does everything to make me demotivated from putting effort in.Also it is funny that some people who struggled (or even failed) to get out of Ruby in previous seasons are suddenly some hundred MMR higher than me. Talk about MMR being representative of skill.
Tl;dr: The current season is the worst one ever unless you like getting participation trophies handed over to you in record time (around 28hours to get all season exclusive rewards) or are a top tier player who can compete in top 250.
this sistem works on 5 man teams (team score,not individual) and pure solo, (in large numbers the unique constant on all your matches are you and stadisitcally you get teamed with better an bader players). if you mix duo, trio with soloers, population problems in determinated hours, players doing manipulations on mm etc etc the stadistics get broken an system becomes garbage
Are you for real? GW2 uses rating system invented for chess which is a 1v1 game? xD Anet keeps on surprising me.
It does, that is why it is such an inaccurate system for a team game. It is a really terrible system for team games no matter what some pvp people try to tell you.
It might work if people would be forced to make static 5 man teams and the whole team gets a MMR. But it will never work for solo players in a team game.
There is also no system in place that forces lets say rank 1 and 2 to play against each other to figure out who is better. If you think someone is too strong just queue dodge him or play in off peak hours.
The whole MMR system offers nothing for people who are not top tier players. The rewards are either participation trophies (PvP tickets, gold, ascended shards of glory) or totally out of reach for average players. In previous seasons you could at least put in effort and work towards the seasonal legendary title. People say there were grindable. Maybe true, but at least they kept motivation up. The current system does everything to make me demotivated from putting effort in.Also it is funny that some people who struggled (or even failed) to get out of Ruby in previous seasons are suddenly some hundred MMR higher than me. Talk about MMR being representative of skill.
Tl;dr: The current season is the worst one ever unless you like getting participation trophies handed over to you in record time (around 28hours to get all season exclusive rewards) or are a top tier player who can compete in top 250.
this sistem works on 5 man teams (team score,not individual) and pure solo, (in large numbers the unique constant on all your matches are you and stadisitcally you get teamed with better an bader players). if you mix duo, trio with soloers, population problems in determinated hours, players doing manipulations on mm etc etc the stadistics get broken an system becomes garbage
The problem is a season only last 2 months. Most people will not play anywhere near enough matches even if it would be a 5 man party leaderboard or soloQ only.
So either we need longer seasons (6 months?) with more incentives to play as many matches as possible or some kind of judging system where volunteers review random games and manually adjust MMR based on their judgement.
1st person worldwide to reach 35,000 achievement points.
Are you for real? GW2 uses rating system invented for chess which is a 1v1 game? xD Anet keeps on surprising me.
It’s modified to account for 5v5 differences.
Are you for real? GW2 uses rating system invented for chess which is a 1v1 game? xD Anet keeps on surprising me.
It does, that is why it is such an inaccurate system for a team game. It is a really terrible system for team games no matter what some pvp people try to tell you.
It might work if people would be forced to make static 5 man teams and the whole team gets a MMR. But it will never work for solo players in a team game.
There is also no system in place that forces lets say rank 1 and 2 to play against each other to figure out who is better. If you think someone is too strong just queue dodge him or play in off peak hours.
The whole MMR system offers nothing for people who are not top tier players. The rewards are either participation trophies (PvP tickets, gold, ascended shards of glory) or totally out of reach for average players. In previous seasons you could at least put in effort and work towards the seasonal legendary title. People say there were grindable. Maybe true, but at least they kept motivation up. The current system does everything to make me demotivated from putting effort in.Also it is funny that some people who struggled (or even failed) to get out of Ruby in previous seasons are suddenly some hundred MMR higher than me. Talk about MMR being representative of skill.
Tl;dr: The current season is the worst one ever unless you like getting participation trophies handed over to you in record time (around 28hours to get all season exclusive rewards) or are a top tier player who can compete in top 250.
this sistem works on 5 man teams (team score,not individual) and pure solo, (in large numbers the unique constant on all your matches are you and stadisitcally you get teamed with better an bader players). if you mix duo, trio with soloers, population problems in determinated hours, players doing manipulations on mm etc etc the stadistics get broken an system becomes garbage
The problem is a season only last 2 months. Most people will not play anywhere near enough matches even if it would be a 5 man party leaderboard or soloQ only.
So either we need longer seasons (6 months?) with more incentives to play as many matches as possible or some kind of judging system where volunteers review random games and manually adjust MMR based on their judgement.
more large seasons, or just directly the old permanent ladder ok
the judging…. mmmm … not to adjust points, to decide about infractions, players trolling, afk, or other manipulations, and to invalidate in these cases the outcome of the match if needed and punish the infractor
So anyone going to provide a real mathematical argument? Or just continue to make kitten up?
Kirrena Rosenkreutz
So anyone going to provide a real mathematical argument? Or just continue to make kitten up?
In theory, Glicko should take into account win streaks. As in, if you’re consistently winning, you are going up in rating, meeting higher ranked players, defeating them, repeat. However, Glicko also becomes more and more certain the higher up you go, so your individual gains from each match along the win streak tapers down, as again, should be expected.
Other games though, make an extra modification to quickly get out of a bad placement, by giving bonus rating. This is something in addition to Glicko, as far as I can tell. This means that your win streak doesn’t need to be as long to reach your target skill division. Without bonuses, hypothetically, you may need 30 matches to advance to your actual skill rating, with bonuses, that could be brought down to 15 matches, for instance. Actual numbers would depend on many factors, but the assumption here is, the player doesn’t have trouble consistently winning. The bonus rating changes should fix any anomalies in placement matches where a player gets placed too low and the system stabilizes too quickly.
http://imgur.com/n7oceII Bad image due to lack of scanner.
That should highlight some scenarios that could happen with Glicko. TRUE indicates players true skill rating, starting average is 1200, and HIGH/LOW represents misplacements. Of course, this assumes that a player’s true rating is above the starting average, and there’s an equivalent set of scenarios for if their rating is below average.
The red line indicates what should happen, in theory, for an average player whose true rating is 1200. The system should match them up in 50-50 matches and they’ll fluctuate around 1200.
The purple line indicates a proper placement, where they’re placed right at their true rating, and they fluctuate there.
The green line indicates a player who completely overshot their true rating, and through a series of win/lose streaks that got progressively shorter, arrived at their true rating.
Orange indicates a high misplacement. In theory, the player should hit their losing streak a lot quicker, unless they played it super safe and only did 1-2 matches a day to maintain a 50-50 winrate despite not belonging there. They may inevitably hit a big losing streak, and they’ll drop down to their true rating. How fast that happens depends on any extra penalties they accrue along the way (right now, there appears to be no extra penalty for being on a losing streak, aside from a big drop in rating by default).
The light blue line indicates a low misplacement. In this scenario, the win streak should happen a lot quicker to boost them out of that low rating towards their true rating. The win streak could cause an overshoot, but they should lose matches and start to fluctuate around their true rating.
The minor fluctuations are a result of the 50-50 win rate. It’s a pattern of win/loss/win/loss/repeat, or win/win/lose/lose, which is a more volatile case.
Lets consider a hypothetical case where the red line, the average player, improves in skill significantly over many matches, and their true rating is indicated by HIGH. The problem now though, is that the system has stabilized to the point where their gains and losses aren’t particularly high. They go on a long win streak, but because their gains are low, their win streak slope would be far less steep than light blue. Assuming they maintain that win streak, it will take them a really long time to reach HIGH. This is where the win streak bonus should kick in, to allow that player to get a large jump (steeper win streak slope) to get to HIGH. They may overshoot it, but again, they should be able to fluctuate until they settle at HIGH.
The win/lose streaks should effectively change a player’s volatility to allow for quicker changes, before settling again, but only for the truly anomalous cases where placements put them either too high or too low, and for any player who surpasses their true rating significantly.
One caveat though, is the frequency of wins/losses. If you only play 1 match every 3 days, and win all of them, that shouldn’t be counted as a win streak, as you’re playing it too safe. Same goes for losing streaks. Win/lose streaks should factor in how often you’re playing too; if you play 10 matches per day and win all of them, then that’s a fair win streak.
The win/lose streaks should effectively change a player’s volatility to allow for quicker changes, before settling again, but only for the truly anomalous cases where placements put them either too high or too low, and for any player who surpasses their true rating significantly.
This is basically what I was referring to. If you’re getting a win streak like the OP indicated (and it was against even opponents), he’ll start to increase faster. But like you said, such a huge shift is unexpected and should stabilize quickly.
The point is that Glicko2 does account for anomalies and that a special “win streak bonus” is not needed.
Kirrena Rosenkreutz
(edited by Exedore.6320)
I played over 10 matches today (soloQ) and won 80-90% (try harded a bit) of them. But my MMR barely increased, it is really frustrating.
In old seasons I would have gone up half a division by this because of win streak bonus.
1st person worldwide to reach 35,000 achievement points.
I played over 10 matches today (soloQ) and won 80-90% (try harded a bit) of them. But my MMR barely increased, it is really frustrating.
In old seasons I would have gone up half a division by this because of win streak bonus.
Well the old system was a grind with safety nets at every level and not based on skillful team play so there is that to consider, Legendary in s1-s4 meant next to nothing as with all the divisions during those seasons.
I played over 10 matches today (soloQ) and won 80-90% (try harded a bit) of them. But my MMR barely increased, it is really frustrating.
In old seasons I would have gone up half a division by this because of win streak bonus.
Who you play is just as important as the outcome. If you beat people with lower ratings than you consistently, your rating won’t (and shouldn’t) go up much. If you consistently beat people equal or above you, your rating will rise.
Kirrena Rosenkreutz
I played over 10 matches today (soloQ) and won 80-90% (try harded a bit) of them. But my MMR barely increased, it is really frustrating.
In old seasons I would have gone up half a division by this because of win streak bonus.Who you play is just as important as the outcome. If you beat people with lower ratings than you consistently, your rating won’t (and shouldn’t) go up much. If you consistently beat people equal or above you, your rating will rise.
So should I not play against better people if I win many games in a row to prove myself? But I think my rating is moving slow because I already played 294 matches this season, so the algorithm considers it as settled.
1st person worldwide to reach 35,000 achievement points.
So should I not play against better people if I win many games in a row to prove myself?
Yes. And the system does do that.
But I think my rating is moving slow because I already played 294 matches this season, so the algorithm considers it as settled.
It only stays settled if you perform to the expectation of your rating. If you have unexpected outcomes (such as defeating people equal to or higher than your consistently), your rating moves up and the amount it increases with each move and the amount it moves each time will increase slightly.
Keep in mind that it’s unlikely for someone’s performance which is “well known” to change dramatically (multiple hundreds of points) overnight. That’s why the system doesn’t all of a sudden bump you by large amounts; instead it tracks your performance vs. the expectation, but also takes into account that it is less sure of your true rating.
Kirrena Rosenkreutz
Keep in mind that it’s unlikely for someone’s performance which is “well known” to change dramatically (multiple hundreds of points) overnight. That’s why the system doesn’t all of a sudden bump you by large amounts; instead it tracks your performance vs. the expectation, but also takes into account that it is less sure of your true rating.
I disagree, if someone paid a pro league player to duoQ with him and follow his instructions, they could probably get top 250 at any time. I already had someone offer me to help me get rank 1 in a future seasons for 5000g for example.
Even if you do not hire a pro league player, playing with duoQ instead of soloQ can drastically change your win rate for better or worse over night.
1st person worldwide to reach 35,000 achievement points.
I think once you get 3 or 4 wins in a row your MMR should start increasing at a faster rate. Not really seeing that now…
The win/lose streaks should effectively change a player’s volatility to allow for quicker changes, before settling again, but only for the truly anomalous cases where placements put them either too high or too low, and for any player who surpasses their true rating significantly.
This is basically what I was referring to. If you’re getting a win streak like the OP indicated (and it was against even opponents), he’ll start to increase faster. But like you said, such a huge shift is unexpected and should stabilize quickly.
The point is that Glicko2 does account for anomalies and that a special “win streak bonus” is not needed.
While a bonus system isn’t needed, and Glicko corrects anomalies by itself, the bonus system would correct anomalies faster than just glicko by itself.