Clint has agreed to take a look at the Player Rating System I'm
working on, and if people seem to like it, add it to the current PRS.
I want to create a rating system that most players will pay attention
to, and try to score well at. It wouldn't be a listing of "best
players", as no system can measure that. It would simply be a
listing of the players that play a lot, in challenging games, and
still manage to win a lot.
Ifm looking at a system that uses 3 numbers that are multiplied
together to get final ranting/ranking.
Adjusted win %
Based on last 10 positions completed. Win = 1, Loss =0, Transfer = -
1, Hard Drop = -2. Divide this by games complete (1-10). Every
player is given 1 free win to avoid division by 0 and give new
players a rating even before theyfve gotten their first win. (Bunch
more rules about neutrals waiting too long to flip, very short games,
etc, but Ifm trying to keep this post brief... Well briefer that it
could be.)
Game Activity Adjuster (GAA) that counts each position played as 1
point per game season, reduced by 5% per season. This counts how
active a player you are. Those that play the most will score the
best.
Where n=0 is current game season and n=19 is 5 years ago.
GAA = " PositionsPlayed(n) * (20-n/20)
[Side note. Brad Burnett suggested updating this every turn, but I
wasn't sure how I could get the data. I've realized it could use
the .xml files generated with the .pdf.files. Also, there is no
reason to lock it into a 5% seasonally decline except to encourage
active playing and give newer players a chance to move up
eventually. Comments encouraged.
I like the number being about 10 for 1 game at a time for 5 years and
about 25 for 3 games at a time for 2.5 years. It may actually be
easy to program "turns processed" per season/6 aged at 5% per season.]
To score well, play a lot and win a lot.
So far, so good, yes? No?
Here is where it gets ugly.
The third number I want is a measure of game difficulty. It would be
a combination of player experience and nation balance. I think that
playing 11 v. 14 (20% more players on one team) should count a lot
more than playing on a team that averages 4 years of experience
against a team that averages 5 years(20% more expereince).
I call this Opponent Strength Adjuster (OSA).
I want to update OSA every turn so that it reacts to nation
elimination and neutral shifts. Fortunately, this info is available
in the .xml files now sent out with turns. (Palantir is a GREAT
addition to the game. If you haven't checked it out, do so.)
My most recent attempt is to start everyone with an Opponent Strength
Adjustment of 1, and then modify it based on:
1) % difference in number of active positions
2) 1/10th the % difference in experience.
This amount of change would be per year normalized to 26 turns a
year. 30% fewer positions, for 26 turns should get you about 30%
increase in rating. 30% less experience, for 26 turns, would get you
about 3% increases in rating.
Problem is, that 50% less (half) is 100% more (double) for the other
side. 75% less would be 300% more. The % difference breaks down
then one side has 200-300% more experience than the other team. Even
using 10% of the difference could get 20-30% change in a year.
Thatfs as much as playing 14vs.11 for a year. An experienced player
could get on a team of new players, and see his OSA soar.
Thereafter, if he only played against teams with similar experience,
hefd never see his OSA fall.
Therefore, Ifll always use the "less than" number (dividing the
difference by the larger) and move up and down by that much.
The way to reduce a number by x% is to multiply the number by (1-
x%). The way to increase a number by x% is to multiply by 1+x%.
Since I want x% per 26 turns for nation difference, the formula is 1
+/- x%/26. For 1/10th experience difference the formula is 1+/-
x%/260. Actually compounding will result in slightly more or less
than x%, but too bad.
Therefore, the formula for OSA would be
···
+++++++++++++++++++++
Player Experience = GAA, but without the 5% per season aging
IF opponents are more experienced
Experience Adjustment = 1 + (sum of opponents experience - sum of
allies Experience)/ sum of opponents experience *260).
IF your allies are more experienced
Experience Adjustment = 1 - (sum of allies Experience - sum of
opponents experience) / sum of allies experience *260).
If opponents have more nations
Nations Adjustment = 1 + (# of enemy nations-# of ally nations)/ (#
of enemy nations*26).
If opponents have fewer nations
Nations Adjustment = 1 - (# of ally nations-# enemy of nations) / (#
of ally nations*26).
New OSA = Old OSA * Experience Adjustment * Nations Adjustment
++++++++++++++++++++++++++++
The idea is to punish experienced players for stomping newer players,
encourage experienced players to have newer players join their teams,
encourage neutrals to pick sides to balance a game, and account for
playing in unbalanced games.
Neutrals aren't counted except for FA games beyond GT12 where they
are considered a 3rd alliance and the number of enemy nations is
averaged before being compared to the number of friendly nations.
Examples:
Playing 26 turns (against players with similar experience) with a
14/11 split in your favor (opponents have about 20% fewer nations and
experience) would result in a 21% drop in your rating. OSA goes from
1 to 0.7893. This is a huge shift, but a 14 vs. 11 game shouldn't go
a full year.
Playing a 12 v 12 game, for 26 turns where your team averages 5+
years of playing 3 games at a time (60 XP) against a team that
averages 2 year of playing 2 games at a time (16XP) (they have 70%
less experience) results in a about 7% drop in rating over 26 turns.
OSA goes from 1 to 0.9292. This may seem like a small change, but it
is almost the same as the difference between winning 6 of your last
10 games instead of 7 of your last 10 games.
Flipping these examples around, an 11/14 split against (with equal
experience) and you hold on for a year, would increase your OSA from
1 to 1.2721.
Being on a team with 16XP average vs. a 60XP average (in a 12 vs. 12)
would increase your OSA from 1 to 1.076
So, my questions are:
1) Do we need opponent strength as a factor in a Player Rating System
or is win % and game activity enough for "bragging rights".
2) Is this getting too complicated? Is there a better way to account
for game difficulty?
3) Am I not adjusting enough for experience?
4) Since neutrals working for an alliance before flipping can't be
tracked, should we ignore nation and experience difference as
too "cheatable"? Another "cheatable" problem would be to create sock
puppet accounts (against the rules) to lower your team experience.
Would this be a problem?
5) Would you ignore EVERY Player Rating System no matter how well it
is designed?
6) Do you think the current system is good and not worth the effort
to try to improve it?
