1 |
@GoogleFrog I hate it when you come down on me like this without data to back it up. @Godde and other top 10 players have a point here with them sometimes not being matched, but being top 10 is a tiny edge case that we can handle specially.
|
1 |
@GoogleFrog I hate it when you come down on me like this without data to back it up. @Godde and other top 10 players have a point here with them sometimes not being matched, but being top 10 is a tiny edge case that we can handle specially.
|
2 |
\n
|
2 |
\n
|
3 |
This system only affects players who are playing a streak of MM games, the first game of the day is always handled as if the system didn't exist. Afterwards it tries to average out the win chance of the player. I tested the system in practice with @flap, who lies somewhere in the upper mid. Without the system, he would mostly play against newbies and win game after game, without much elo to gain. The system allows such a player to get occasional matches with the high elo crowd whenever enough newbies have been beaten up. In the end it leads to each player seeing a roughly even distribution of stronger and weaker players during an evening. Something that was often not met due to the bimodal elo distribution.
|
3 |
This system only affects players who are playing a streak of MM games, the first game of the day is always handled as if the system didn't exist. Afterwards it tries to average out the win chance of the player. I tested the system in practice with @flap, who lies somewhere in the upper mid. Without the system, he would mostly play against newbies and win game after game, without much elo to gain. The system allows such a player to get occasional matches with the high elo crowd whenever enough newbies have been beaten up. In the end it leads to each player seeing a roughly even distribution of stronger and weaker players during an evening. Something that was often not met due to the bimodal elo distribution.
|
4 |
\n
|
4 |
\n
|
5 |
I
went
ahead
and
downloaded
the
recent
MM
matchups.
Before
the
change,
the
average
win
chance
deviation
of
a
player
was
~9.
2%
(
14000
games)
.
Since
adding
the
system,
it
has
dropped
to
~7.
4%
(
2300
games)
.
The
most
active
player
in
the
2000
battles
before
and
after
the
change
was
@brianfreeman,
his
average
win
chance
(
compensated
for
elo
at
the
time
of
the
game)
went
from
32.
0%
to
39.
4%.
Second
most
active
was
@Tiptushi
who
went
from
64.
3%
to
57.
0%.
For
the
average
player,
the
system
is
clearly
working.
To
show
that
this
doesn't
come
at
the
cost
of
killing
the
matching
times,
the
MM
activity
compared
to
other
game
types
also
increased
by
~30%,
but
this
is
probably
due
to
other
factors.
|
5 |
I
went
ahead
and
downloaded
the
recent
MM
matchups.
Before
the
change,
the
average
win
chance
deviation
of
a
player
from
50%
was
~9.
2%
(
14000
games)
.
Since
adding
the
system,
it
has
dropped
to
~7.
4%
(
2300
games)
.
The
most
active
player
in
the
2000
battles
before
and
after
the
change
was
@brianfreeman,
his
average
win
chance
(
compensated
for
elo
at
the
time
of
the
game)
went
from
32.
0%
to
39.
4%.
Second
most
active
was
@Tiptushi
who
went
from
64.
3%
to
57.
0%.
For
the
average
player,
the
system
is
clearly
working.
To
show
that
this
doesn't
come
at
the
cost
of
killing
the
matching
times,
the
MM
activity
compared
to
other
game
types
also
increased
by
~30%,
but
this
is
probably
due
to
other
factors.
|
6 |
\n
|
6 |
\n
|
7 |
The only downside that I have heard from other players, is this example of @Dregs after a loss streak not matching with @Godde after a win streak. This can be fixed with the exponent, by adding a special rule for the top players, or by having them play in a competitive ranked custom host. We don't have to dismantle the system for this.
|
7 |
The only downside that I have heard from other players, is this example of @Dregs after a loss streak not matching with @Godde after a win streak. This can be fixed with the exponent, by adding a special rule for the top players, or by having them play in a competitive ranked custom host. We don't have to dismantle the system for this.
|