[Attempt at] New PBEM rating system - continued

[Mathemagician]

Originally posted by Mathemagician
so, for several finishing orders:

ABC : +4 / -1 / -3 // SUM = 0
ACB : +4 / -11 / +7 // SUM = 0
BAC : +9 / -6 / -3 // SUM = 0
BCA : +9 / +7 / -16 // SUM = 0
CAB : +17 / -6 / -11 // SUM = 0
CBA : +17 / - 16 / -1 // SUM = 0

sorry, misposted here. the numbers are correct, just not the order. it should say (so that the first score belongs to first player/letter in the list):

ABC : +4 / -1 / -3 // SUM = 0
ACB : +4 / +7 / -11 // SUM = 0
BAC : +9 / -6 / -3 // SUM = 0
BCA : +9 / +7 / -16 // SUM = 0
CAB : +17 / -6 / -11 // SUM = 0
CBA : +17 / - 1 / -16 // SUM = 0

now, based on the initial scores of 1050 for A, 1020 for B and 980 for C, A always gets +4 / -6 / -16 for places 1, 2 and 3. B always gets +9 / -1 / -11 and C always gets +17 / +7 / -3.

Originally posted by blackice
Now tell me those numbers posted are wrong again? You posted them based on your system...I just applied the final math and whoa it's flawed...

that i mixed up the order of the numbers still does not mean they add up in anyway for each *9 turn, blackice.
if you refuse to understand that and further refuse to explain to me what kind of flaw you see and why you
believe it is a flaw, i cannot help you.

10% system makes allowance for this yours does not and you simply dismiss it, not good.

if i dismiss anything, its because you only speak in half-sentences that dont explain the problem.
i suggest you explain one problem in detail in a post instead of alluding to all the problems at once without any insight what you mean.
didnt want to dismiss anything, just not understanding what you mean.

[ricketyclik]

What is not clear to me, and perhaps someone could simply and clearly tell me (I am too lazy to wade back through the voluminous posts) is

a) which system(s) only change points on a change of ranks?

b) which system(s) penalise early retirement, and how?

c) how are replacement players handled?

[Mathemagician]

Originally posted by ricketyclik
What is not clear to me, and perhaps someone could simply and clearly tell me (I am too lazy to wade back through the voluminous posts) is

i'll try to answer for those systems i have understood

a) which system(s) only change points on a change of ranks?

my system does that and as far as i understood it also the quadratic system of quinns that Solver explained.

note:
actually i think each system could be used in a way so that scores only change when ranks change (easily accomplished by 'replacing' the score instead of accumulating it).
the point is for which system this makes sense, i.e. its creator/representative wants it to be used. so again, mine and quinns' system plan to function that way.

b) which system(s) penalise early retirement, and how?

here i can only speak for my system. (non-exponential ELO as it was called)

in my system there is about equal penalty for early or late retirement. you are the assigned the last place of the remaining players. the only difference for later retirements is that another player may have retired earlier.

example:
for example, in a 5 player game using my system, the player resigning first would be counted 5th (thus last) for the rest of the running game. if that player chose to resign later though, after another player has resigned, that other player would be 5th and the newer player 4th and so on.

c) how are replacement players handled?

again, i can only speak for my system, as all your points can be deliberately chosen, independent of the mathematical implications.
for my system i havent solved that yet.
i'll be working on this and edit this post when i have found a way

[quinns]

Math, Solver, and blackice, I'll post all the polls if you like (Solver's a bit slow to the punch ), but please spell out each of your described systems in 50 words or less (for the voters). No emotional words, please, like "brilliant", "idiotic", "ridiculous", etc.

[blackice]

that i mixed up the order of the numbers still does not mean they add up in anyway for each *9 turn, blackice.
if you refuse to understand that and further refuse to explain to me what kind of flaw you see and why you
believe it is a flaw, i cannot help you.

Yup you did I guess now you see the "flaw" I was refering to with your math...
The other "flaws" have been explained serveral times now...I guess you will just have to wait and see them. I can not and will not explain them any more you simply do not get it. You look at numbers not a whole system, we have and we have tested and looked closely at the elo and quad. The flaws are there look beyond the numbers.....

ABC : +4 / -1 / -3 // SUM = 0 b loses 1 point for beating c (see CBA)
ACB : +4 / +7 / -11 // SUM = 0
BAC : +9 / -6 / -3 // SUM = 0 a loses for beating c?
BCA : +9 / +7 / -16 // SUM = 0 c gets 7 for beating a and 7 for beating b?: (see ACB)
CAB : +17 / -6 / -11 // SUM = 0 a loses for beating b?
CBA : +17 / - 1 / -16 // SUM = 0 b loses for beating a? (See ABC) but b loses 1 point for beating a?

"Flawed" again are you sure you have it right this time cause it still does not make sense?

Point again why would a top player play lower ranked players? Is it a rule to force them to play?

that i mixed up the order of the numbers still does not mean they add up in anyway for each *9 turn,

I simply assumed you intented this example for ONE ninth turn ranking...Was I wrong?

if i dismiss anything, its because you only speak in half-sentences that dont explain the problem.

Half sentences? LAND IS A MAJOR PART OF ANY WIN OR LOSS ASK ANYONE WITH A HALF A BRAIN. Clear enough? You just dismiss this fact WAKEUP. How's that for half sentences? Skill and land placement are two different things PAUL, YODA, any world ranked top player will tell you this, let alone any average player. Why is this so difficult for you to understand? Why do you wave it away with your magic wand? A FLAW...

ricketyclik

A) 10% does not do this it was tested and it was flawed. You wager 10% at your current ranking and that is it for the rest of the game.

b) 10% does this by dropping one full ranking in your over-all ranking, if you are first over-all you are now second over-all. The game penalty, there is none but the remaining players do get a bonus, i'll explain: Until a replacement is found say in a four player game the scores on the ninth turn are counted as a three player game. Meaning more points earned for each remaining player until the dropped player is replaced.

c) In the 10% system the new player is non-rated for two full 9th turn rankings. More than enough time for thier playing style to show through. After that they become ranked in the game, great opportunity for a newbie to jump in and get the feel for things. Also two full rounds of bonus for the remaining players for the BS of finding a new player.

Again I see numbers Math that do not add up. I see numbers but no system, rankings are a system not just numbers...

So tell me what problems do you have with the 10% SYSTEM? I for one would like to hear them? I mean you came up with a new one but clearly do not understand the old one? So why the push for a new one? Why not just support the old one? That goes for Solver and Quinn too...Solver just did not like it, unknown why? Quinn it was personal, that's understandable but time has passed and I have yet to see anyone clearly state why the 10% system was not good?

Partisan politics? It could'nt be so what is "flawed" about the 10% system? No emotional words, please, like "brilliant", "idiotic", "ridiculous", etc.

[checkMate]

sorry, misposted here. the numbers are correct, just not the order. it should say (so that the first score belongs to first player/letter in the list):

ABC : +4 / -1 / -3 // SUM = 0
ACB : +4 / +7 / -11 // SUM = 0
BAC : +9 / -6 / -3 // SUM = 0
BCA : +9 / +7 / -16 // SUM = 0
CAB : +17 / -6 / -11 // SUM = 0
CBA : +17 / - 1 / -16 // SUM = 0


Now I get it That is much more like it should be.

[checkMate]

OOPS, I think you were in a hurry. BAC can't be right.

[blackice]

3 Player game the 10%Formula: ..1ST..54%..2nd..31%..3rd..15% Formula...
Starting for the three players:
A : 10.5
B : 10.1
C : 9.7

Wager:
A : 1.05
B : 1.01
C : .97
Total: 3.03 for the pot.

ABC : A 1st) 1.64 / B 2nd) 0.94 / C 3rd) 0.45
New over-all ranking :
a: 12.14
b: 11.04
c: 10.15

Wager:
A : 1.05
B : 1.01
C : .97
Total: 3.03 for the pot. This remains until games end...

ACB : A 1st) 1.64 / C 2nd) 0.94 / B 3rd) 0.45 turn 9
New over-all ranking:
a : 13.78
b : 11.49
c : 11.09

BAC : B 1st) 1.64 / A 2nd) 0.94 / C 3rd) 0.45 turn 18
New over-all ranking:
a : 14.72
b : 13.13
c : 11.54

BCA : B 1st) 1.64 / C 2nd) 0.94 / A 3rd) 0.45 turn 27
New over-all ranking:
a : 15.17
b : 13.13
c : 12.48

CAB : C 1st) 1.64 / A 2nd) 0.94 / B 3rd) 0.45 turn 36
New over-all ranking:
a : 16.11
c : 14.12
b : 13.58

CBA : C 1st) 1.64 / B 2nd) 0.94 / A 3rd) 0.45 turn 45
New over-all ranking:
a : 16.56
c : 15.76
b : 14.52

Now lets just say because this game is full of Diplomacy (Backstabbing ) C and B gang up on A for another round...

CBA : C 1st) 1.64 / B 2nd) 0.94 / A 3rd) 0.45 turn 54
New over-all ranking:
c : 17.40
a : 17.01
b : 15.46

A is no longer in first... Now A, B or C could have other games or not, the point being is no one is assured a spot. B could have bad land but is still not out of the race by any means...

The rankings are always on the move, very (Pro)active to say the least. Now we all know in another 6, 9th turn rankings the whole game changes, I.E. sols cannons and muskets let alone wonders and land. Land included here Math because while you may have had forest for your fast start your opponent had mountains. His mines will out-produce your forest in the next 100 with out a doubt... We all know this...Right?

The point being with this system the rankings are very active and no one can be assured a constant ranking be it one game or ten.

You will also note all players gain points not lose them, we call that incentive... This helps to ensure no one quits a game which should be the main idea right? I mean all points add up so why drop a game? Thats part of what we call a "system" which is part of the "rankings" I.E. math, formulas what ever you choose to call it. One is useless without the other...

Now look at this the same players starting a new game at turn 54 with new players:

c : 17.40
a : 17.01
b : 15.46

Wager for game:

a : 1.74
b : 1.70
c : 1.54

Pot for new game: 4.98 as opposed to 3.03 for the first one.


Now this also ensures new games and participation for oldtimers and newbies alike...It also assure you can not sit on your butt in one game and expect to be top dog...

Seems to me and others, that is what we are after here is it not? If you doubt it check out the challege ladder Quinn started People want to play and they are not forced to do so per se.

10% has less rules more incentive and more "positive" activity in the rankings. It has been used for what 5 years now? It was designed with CTP and most important is easy to use and a complete "system". By use I mean maximizing the system to make the greatest gains in the rankings..Without quiting games and all the BS that goes with It. The formula is easy and the rules minimal "easy system of rankings".

[blackice]

sorry, misposted here. the numbers are correct, just not the order. it should say (so that the first score belongs to first player/letter in the list):

ABC : +4 / -1 / -3 // SUM = 0
ACB : +4 / +7 / -11 // SUM = 0 c gains 7 for for beating b?
BAC : +9 / -6 / -3 // SUM = 0
BCA : +9 / +7 / -16 // SUM = 0 C gains only 7 for beating A? (see ACB)
CAB : +17 / -6 / -11 // SUM = 0 A loses 6 for beating B? and C (see BAC)
CBA : +17 / - 1 / -16 // SUM = 0 yet B wow loses 1 for beating A? Yet gains 7 for beating B? (see ACB)

Nope still not right B may have lost to C but for pete's sake he beat the top player and loses points? FLAWED...Maybe I need three D glasses?

Now I get it That is much more like it should be.

I am still miffed the math is still not working for me anyone else?

Seems like a top dog should never play a rookie. I guess you would make a rule they can't? Or is the rule you have to accept all challenges? In which case we have lives too we have what 30 active players?. Or is it you can only challenge two or three above your ranking? More rules in which case decide because numbers are only part of a "ranking system"

I for one with this system would never play a rookie anyone else? So little to gain so much to lose so little time... So I guess you have to make a rule to force people to play anyone...

Now if there is no rule we have packets of people in small groups playing each other with close rankings, "community" is lost...heirarchy, top dog, little newbie, bottom feeder. We should avoid this at all cost, should'nt we?

[ricketyclik]

Actually, Blackice, it seems to me with the 10% system with the pot capped, there is no change in score without a change in rankings.

Correct me if I am wrong.

[Mathemagician]

Originally posted by blackice
ABC : +4 / -1 / -3 // SUM = 0
ACB : +4 / +7 / -11 // SUM = 0 c gains 7 for for beating b?
BAC : +9 / -6 / -3 // SUM = 0
BCA : +9 / +7 / -16 // SUM = 0 C gains only 7 for beating A? (see ACB)
CAB : +17 / -6 / -11 // SUM = 0 A loses 6 for beating B? and C (see BAC)
CBA : +17 / - 1 / -16 // SUM = 0 yet B wow loses 1 for beating A? Yet gains 7 for beating B? (see ACB)

the points each player gains/loses are already summed up (not over 9* turns, just for the theoretically 'seperate' results between each 2 of the three players).
so, just taking this one example you criticized:

CAB : +17 / -6 / -11 // SUM = 0 A loses 6 for beating B? and C

C, as the worst player beat both higher ranked players, thus he gains +17.

A 'lost' to the weaker C and 'won' against the also weaker B.

losing to a weaker player = lose many points
beating a weaker player = gain few points


lose many points + gain few points = still lose a few points = -6

B 'lost' to the stronger player A (lose a few points) and to the weaker player C (lose many points).
summed up, he loses points twice, for a total of -11, which is not as much as C gained (as both of C's beaten opponents were ranked higher, the wins thus more unlikely)
had B lost to two weaker players, he would have lost more points. you can see that from the -16 A loses when losing to everyone else.

[i]it works analogously for the other examples.

--
i posted that cause the number table just hit my eye.
i dont bother reading all your other stuff, blackice. sorry.
80% rants and 20% information isnt worth it. period.

[Solver]

Yeah, Quinns, spelling out each system in 50 words is what we need and what I have some trouble doing .

You could certainly describe the Quinns quadratic formulas, could you not, having created them ?

[quinns]

Okay Solver, here goes -- in 50 words :
---------------
Quadratic Rating System: The change in a player's rating, done every 10 turns at the *9 mark, is calculated as though the player is playing individual games against all opponents. The rating change is based upon:
Probability of lower rated player defeating higher rated player =
[1 /(2 + (1st_Player_Rating - 2nd_Player_Rating)^2)].
--------------

[Mathemagician]

alright, Mr.Solver - but be aware i was always bad at school at essays and the like.

"The Non-Exponential ELO system"
by Mathemagician


The basis of this system is to evaluate a player's rank and relative skill in one game with one formula, as well as to have all players' score gains/losses sum up to 0.
the score is re-calculated every mark *9 turn, but not accumulated.
score = 10 * (r(i) - s(i))
r(i) depends on your rank in the game
s(i) depends on your score/skill before the game

[Solver]

Yeah, I pity the guys who can't understand any maths .