| 1 |
I
was
interested
in
this
claim
that
the
"recent"
overdrive
changes
cause
"elo
inflation".
|
1 |
I
was
interested
in
this
claim
that
the
"recent"
communism
changes
cause
elo
inflation.
|
| 2 |
\n
|
2 |
\n
|
| 3 |
Since the popular approach of just claiming things on anecdotal evidence doesn't seem like something that would actually produce any knowledge, and i don't really have an agenda to push by any means necessary, i've decided to try to check data. Since i'm an unrepentant egoist with an unchecked data privilege, and since my teams elo doesn't seem to have changed all that much in the last few years (it is generally much more stable than my 1v1 elo), i've decided to check whether the "recent overdrive change" can be seen to have affected my teams win rate.
|
3 |
Since the popular approach of just claiming things on anecdotal evidence doesn't seem like something that would actually produce any knowledge, and i don't really have an agenda to push by any means necessary, i've decided to try to check data. Since i'm an unrepentant egoist with an unchecked data privilege, and since my teams elo doesn't seem to have changed all that much in the last few years (it is generally much more stable than my 1v1 elo), i've decided to check whether the "recent overdrive change" can be seen to have affected my teams win rate.
|
| 4 |
\n
|
4 |
\n
|
| 5 |
1)
First,
i
needed
to
date
the
"recent
mex
overdrive
changes".
I
assumed
that
any
commit
that
changes
the
mex
overdrive
would
appear
[url=https://github.
com/ZeroK-RTS/Zero-K/commits/master/LuaRules/Gadgets/unit_mex_overdrive.
lua]here[/url].
On
this
assumption,
the
latest
candidate
commit
would
be
[url=https://github.
com/ZeroK-RTS/Zero-K/commit/4b507eed1f5bf2d7b85d04e1a794ec39133e01fd]on
Jan
5,
2015[/url].
The
earliest
date
on
which
the
change
would
have
affected
my
winrate,
however,
would
have
been
on
the
next
day,
when
it
[url=https://github.
com/ZeroK-RTS/Zero-K/commit/e71e59530edca676bbfdfe4fb1e22ac7461dd9cd]went
live[/url].
|
5 |
1)
First,
i
needed
to
date
the
"recent
mex
overdrive
changes".
I
assumed
that
any
commit
that
changes
the
income
distribution
would
appear
[url=https://github.
com/ZeroK-RTS/Zero-K/commits/master/LuaRules/Gadgets/unit_mex_overdrive.
lua]here[/url].
On
this
assumption,
the
latest
candidate
commit
would
be
[url=https://github.
com/ZeroK-RTS/Zero-K/commit/4b507eed1f5bf2d7b85d04e1a794ec39133e01fd]on
Jan
5,
2015[/url].
The
earliest
date
on
which
the
change
would
have
affected
my
winrate,
however,
would
have
been
on
the
next
day,
when
it
[url=https://github.
com/ZeroK-RTS/Zero-K/commit/e71e59530edca676bbfdfe4fb1e22ac7461dd9cd]went
live[/url].
|
| 6 |
\n
|
6 |
\n
|
| 7 |
2) Second, i needed a table of my teamgame battle results, with small teamgames (in which i'm likely to only play with highly competent people),duels, etc filtered away. [url=http://hastebin.com/tocecebari.sql]That[/url] was easy enough to acquire. It seems i've played no games on fifth or sixth, so this minor dating difference is not relevant.
|
7 |
2) Second, i needed a table of my teamgame battle results, with small teamgames (in which i'm likely to only play with highly competent people),duels, etc filtered away. [url=http://hastebin.com/tocecebari.sql]That[/url] was easy enough to acquire. It seems i've played no games on fifth or sixth, so this minor dating difference is not relevant.
|
| 8 |
\n
|
8 |
\n
|
| 9 |
3) Now, it's eyeballing (and statistics, if anyone cares) time! It seems i didn't play that many games, so the results are not going to be very reliable. Still, they are going to be better than calling names and explaining how i am superior to everyone else (that doesn't need explaining, anyway).
|
9 |
3) Now, it's eyeballing (and statistics, if anyone cares) time! It seems i didn't play that many games, so the results are not going to be very reliable. Still, they are going to be better than calling names and explaining how i am superior to everyone else (that doesn't need explaining, anyway).
|
| 10 |
\n
|
10 |
\n
|
| 11 |
Simply eyeballing the data doesn't seem to suggest i've started winning much more in 2015 in general (except in the last week where i get a nice 5-streak). So let's do some simple rate comparisons.
|
11 |
Simply eyeballing the data doesn't seem to suggest i've started winning much more in 2015 in general (except in the last week where i get a nice 5-streak). So let's do some simple rate comparisons.
|
| 12 |
\n
|
12 |
\n
|
| 13 |
There are 77 games in this dataset played after 2015-01-06; of them, 36 were won, giving a win rate of ~46.7%.
|
13 |
There are 77 games in this dataset played after 2015-01-06; of them, 36 were won, giving a win rate of ~46.7%.
|
| 14 |
[tooltip=46/86 stands for the UN resolution A/RES/46/86: Elimination of racism and racial discrimination. This is not a coincidence, because nothing is ever a coincidence]There are 86 games in this dataset played between 2015-01-06 and 2014-01-06; of them, i've won 46, that is ~53.4%[/tooltip].
|
14 |
[tooltip=46/86 stands for the UN resolution A/RES/46/86: Elimination of racism and racial discrimination. This is not a coincidence, because nothing is ever a coincidence]There are 86 games in this dataset played between 2015-01-06 and 2014-01-06; of them, i've won 46, that is ~53.4%[/tooltip].
|
| 15 |
\n
|
15 |
\n
|
| 16 |
It seems that my win rate has slightly [b]decreased[/b] after the overdrive change. If we consider this difference significant, then it conflicts with the hypothesis that elo has inflated significantly after 2015-01-06 due to the overdrive change, because it would pit me against players with an inflated skill rating, and elevate my win rate.
|
16 |
It seems that my win rate has slightly [b]decreased[/b] after the overdrive change. If we consider this difference significant, then it conflicts with the hypothesis that elo has inflated significantly after 2015-01-06 due to the overdrive change, because it would pit me against players with an inflated skill rating, and elevate my win rate.
|
B422542 6 on Red Comet (Multiplayer)