Why higher point per game values after placement?

No really good reason for this. I happened to get a string of afk players after my placements and got on a losing streak. Now I have to try to make back all those points with less points being gained per win. It's bad enough when these players screw you over for normal point values.

@"Rufo.3716" said:No really good reason for this. I happened to get a string of afk players after my placements and got on a losing streak. Now I have to try to make back all those points with less points being gained per win. It's bad enough when these players screw you over for normal point values.

The GLICKO-2 system used by GW2 has three relevant values to determining how your MMR changes at the end of a match:

1. "Confidence", or how well the predictions match reality.
2. The expected vs actual result. (eg: if you were predicted to win, did you win?)
3. Win vs Lose. Just a boolean, doesn't matter if you win by 1 point, or by 500.

Win vs Lose, number 3, determines if your MMR goes up or down.

The other two, confidence and expected vs actual, determine how big the change is. If you are expected to win, and win, or expected to lose, and lose, you get a small change. If you surprise the system (eg: expected to lose, but win) then you get a big change.

Confidence acts as an additional multiplier: it starts very low, at the start of placement, but rapidly increases as your MMR gets better predicted, so normally by the time ten placements are up the predictions are pretty good, and confidence is reasonably high. This means even the "big" changes when you surprise the system are smaller at the end of placement than the start.

If you get through placement with a super-inaccurate result, though, you will keep confidence low: you keep surprising the system, so confidence stays low, so changes are bigger. This helps move you as fast as possible into the right general area for your MMR -- if you are a "gold 1" quality player, you should get into the region of that fairly fast.

Then the predictions will be right more often, confidence will increase, and the size of changes will decrease. A surprise will still result in a bigger change than delivering what the MMR system expects, but nowhere near as big as during placement.

If you still have those huge changes, you are seeing what is hidden during placement: the big jumps. The only difference between a placement and a real match is that ANet hide the MMR changes so people don't panic about the size of them. Looks like you maybe get a chance to see it in action, thanks to placement sucking.

The good news is that placement is ten matches because that is normally about how long it takes GLICKO-2 to converge on a fairly good estimate, so you shouldn't have this rollercoaster ride for more than another ten.

I'm going to argue about this, but like the OP said, this is still too much luck based. If post placement games influence for +/-25/30, it tells placement games have no sense.I pop at 1200 with 5/5, get 10 loses then 10 wins. I end up at 1050.I pop at 1000 with 2/8. Get 10 wins then 10 loses. I end up 1150.

People going to say " but if you're good, you're going to climb back" which is true. But it keeps telling this is still too much luck based..

@"Abazigal.3679" said:I'm going to argue about this, but like the OP said, this is still too much luck based. If post placement games influence for +/-25/30, it tells placement games have no sense.

It tells you they are not perfect, but that doesn't mean "completely useless" either.

People going to say " but if you're good, you're going to climb back" which is true. But it keeps telling this is still too much luck based..

MMR is an estimate. Like any guess, it can be wrong, which is why it keep correcting all the time -- every single match changes your MMR, from the first to the thousandth match. This is why I think the idea of "placement matches" is bad: it hides the fact that MMR is very volatile early in the season, but it also creates the illusion that those matches are somehow "special" when, in reality, they are just like every other match.

I'm just saying I have a problem with the extra weight to the points after placement matches compared to the rest of the season. Losing 4 games at -25 points per game is a huge difference than losing 4 games at 12-15 points per game. Especially after those highly weighted games are over and you are stuck trying to gain those points back through normal point value games.

@"SlippyCheeze.5483" said:MMR is an estimate. Like any guess, it can be wrong, which is why it keep correcting all the time -- every single match changes your MMR, from the first to the thousandth match. This is why I think the idea of "placement matches" is bad: it hides the fact that MMR is very volatile early in the season, but it also creates the illusion that those matches are somehow "special" when, in reality, they are just like every other match.

True, but this is still complicated to base people upon 10 games ( or more likely 20 since the following can count for +/- 30) since luck is going to happen, even more with a low population/ poor matchmaking/ solo queue issues ( afk/toxicity/...).

@Abazigal.3679 said:

@"SlippyCheeze.5483" said:MMR is an estimate. Like any guess, it can be wrong, which is why it keep correcting all the time -- every single match changes your MMR, from the first to the thousandth match. This is why I think the idea of "placement matches" is bad: it hides the fact that MMR is very volatile early in the season, but it also creates the illusion that those matches are somehow "special" when, in reality, they are just like every other match.

Absolutely. Thankfully, those are as likely, statistically, to artificially inflate your MMR as they are to reduce it: both sides have an equal chance of getting the toxic AFKer who sulks in a corner because they were not carried to victory. Which sucks for the person getting them, but over time tends to balance out in your favor -- because you can only have 4 slots they can get assigned, the other team have 5. :)

@"Rufo.3716" said:I'm just saying I have a problem with the extra weight to the points after placement matches compared to the rest of the season. Losing 4 games at -25 points per game is a huge difference than losing 4 games at 12-15 points per game. Especially after those highly weighted games are over and you are stuck trying to gain those points back through normal point value games.

If you are only getting "normal" point values, then the confidence of the MMR algorithm has increased, which is to say: you are now playing at the level where you rarely surprise it. What you describe is exactly the visible effect of being placed too high, and then dropping to the appropriate level.

@"SlippyCheeze.5483" said:

If you are only getting "normal" point values, then the confidence of the MMR algorithm has increased, which is to say: you are now playing at the level where you rarely surprise it. What you describe is exactly the visible effect of being placed too high, and then dropping to the appropriate level.

No, It's like this every season. Every so many games after you play your placement matches the games are worth more points, wins and loses. There is no surprising the system, and it's no longer surprised when your point values drop. They drop gradually after so many games to where they level off and only vary by around 5 points or so on average.