#26 14.01.2012 15:48:53

Marduk
Administrator
From: Rotterdam, Netherlands
Posts: 151

Re: Importing the ranking

I'm glad you spotted those mistakes. Yea this is exactly the kind of system that should be automatic, by hand it's very easy to make a mistake and it's impossible to correct many games later.

Kryon wrote:
Marduk wrote:

I'm not sure if we should allow players to get a new start with a new nickname. If you played many games it's quite a challenge to keep a decent score (1000 points after 10 games is quite good). If players can be reset to the default 1000 points very easily then this would be unfair to those who keep the old records, even if those records are weighing them down. I think as soon as we find out that a player has played before, we should let that player choose which name he wants to use and then keep that name. When the ranking algorithm is calculated automatically this wont be a big problem (by hand it would be tons of work to change the scores when it turns out that an old and a new ranking entry are actually the same person).

Although I agree that if we allow name changes, players who lost their initial games would simply use a new name to have a fresh start with 1000 pts. However, once a player wins his 1st game, he earns a lot of points and he'll continue using the same name until he loses many consecutive games and drops below 1000. Also, it is impossible to know whether the two players are the same person if the names are very different (e.g. akfaew and dude) unless we check the IP and even the IP might change. So I suggest we do allow name changes for people who want it and fix it for people who don't (e.g., people who had to slightly change their name due to new website which do not allow special characters and space in the name).

True it's hard to police the system to prevent players from changing their name on purpose. Hm this is a difficult decision, if we allow this then the meaning of the scores changes a lot (at least the lower half of the scores, they're most likely to take a fresh start after initial losses). But it's possible that this already happened before and I didn't notice it. Well ok let's allow it, but I hope as few people as possible actually make use of it.

Offline

#27 14.01.2012 16:46:16

akfaew
Administrator
Posts: 620

Re: Importing the ranking

Kryon - thanks for the data. It's been imported.

All that is left is to write the automated ranking.

Offline

#28 14.01.2012 20:06:38

Kryon
Player
Posts: 337

Re: Importing the ranking

akfaew wrote:

Kryon - thanks for the data. It's been imported.

You're welcome. It feels good to contribute to longturn smile

I can see that all previous games are now displayed. A few points:

- Can you list the games in order? It'll look much better.
- The idler column is not really necessary. The idlers have been counted as losers because we have no formal definition of idler and almost no idler records for previous games.
- Is it necessary to show the games without winners (due to server crash, dispute, rule violation, etc.)? (e.g., LT0 LT1 LT3 LT5 LT8 LT14 LT19 LT22 LTXIV LTXV LTXVI LTXVII LTXVIII)
- It seems the names have been imported correctly, but some of the game modes are not correct. For example:

LT18 was a no alliance/lone wolf (N type) game but is listed as teamless (A type). (see http://old.longturn.org/ltxviii-set-diplomacy)
LT24 was a T (team) game but is listed as A game. (see http://old.longturn.org/lt24-team-game )
LT9 was a T but listed as A (see http://old.longturn.org/topic/lt9 )
LTXIII was a T but listed as A. (see http://old.longturn.org/ltxiii-teams-selecting and http://old.longturn.org/ltxiii-no-teams - the no teams vote failed)

The most important one is LT18 which is the only no-diplomacy (lone wolf) game which terror won and got 2240 pts! In a no-diplomacy or team game, p is always 10%, but in an alliance game p varies from 0% to 15%. So if (t)error is not fixed, terror gets 3360 pts instead of 2240 pts.

Last edited by Kryon (14.01.2012 20:21:21)

Offline

#29 15.01.2012 09:08:24

akfaew
Administrator
Posts: 620

Re: Importing the ranking

Kryon wrote:

- Can you list the games in order? It'll look much better.

I need start/end dates first for this.

Kryon wrote:

- Is it necessary to show the games without winners (due to server crash, dispute, rule violation, etc.)? (e.g., LT0 LT1 LT3 LT5 LT8 LT14 LT19 LT22 LTXIV LTXV LTXVI LTXVII LTXVIII)

OK, done. But what about LT22, i found no information about this game.

Kryon wrote:

LT18 was a no alliance/lone wolf (N type) game but is listed as teamless (A type). (see http://old.longturn.org/ltxviii-set-diplomacy)
LT24 was a T (team) game but is listed as A game. (see http://old.longturn.org/lt24-team-game )
LT9 was a T but listed as A (see http://old.longturn.org/topic/lt9 )
LTXIII was a T but listed as A. (see http://old.longturn.org/ltxiii-teams-selecting and http://old.longturn.org/ltxiii-no-teams - the no teams vote failed)

fixed

Offline

#30 15.01.2012 09:23:55

Kryon
Player
Posts: 337

Re: Importing the ranking

akfaew wrote:
Kryon wrote:

- Can you list the games in order? It'll look much better.

I need start/end dates first for this.

I meant using the game number: LT2,LT4....LT29. Only LTXIII uses roman numerics and it is played between LT13 and LT14. The other roman numeric games do not have winners.

akfaew wrote:
Kryon wrote:

- Is it necessary to show the games without winners (due to server crash, dispute, rule violation, etc.)? (e.g., LT0 LT1 LT3 LT5 LT8 LT14 LT19 LT22 LTXIV LTXV LTXVI LTXVII LTXVIII)

OK, done. But what about LT22, i found no information about this game.

I still see all the games listed. It seems Maho has skipped LT22. After LT21, he named the new game LT23.

akfaew wrote:

fixed

Thanks but why hadn't you used the game type info from the data I prepared?

akfaew wrote:

All that is left is to write the automated ranking.

For this, it is important to run the games in sequence in the order they were played which can be simply known from the game number. No need for game start/end times.

We need the start/end times only for suspension of ranks for inactive players but I propose changing the last 3 yr. rule to the last 10 games as Marduk and I discussed above. That way we won't need start/end times. What do you say?

For alliance games, are you going to use my continuous formula P=(Q-2)^0.5*5 or Marduk's discrete Q/P table?

If you share the script you wrote, Marduk or I can double check to make sure there are no errors.

Last edited by Kryon (15.01.2012 09:30:07)

Offline

#31 15.01.2012 09:41:49

akfaew
Administrator
Posts: 620

Re: Importing the ranking

Kryon wrote:

I still see all the games listed. It seems Maho has skipped LT22. After LT21, he named the new game LT23.

They are listed, but there is a field 'rank' that determines if the game counts.

Kryon wrote:

For alliance games, are you going to use my continuous formula P=(Q-2)^0.5*5 or Marduk's discrete Q/P table?

If you share the script you wrote, Marduk or I can double check to make sure there are no errors.

There is no script yet, I wrote an email to you about that

Offline

#32 15.01.2012 11:51:40

Kryon
Player
Posts: 337

Re: Importing the ranking

Here's the sample script I wrote in a generic language. It uses the continous P/Q equation I suggested. Any comments/suggestions are welcome.

INPUT DATA:

GN(i) = game number of i th game (e.g. 2,4,.13,XIII,14,..,28,29)
GM(i) = game mode of i th game (T, A, or N)
L(i) = names of losers of i th game
W(i) = names of winners of i th game

OUTPUT DATA:

S('player') = rank score of 'player'

CODE:

p=0.1 ;percentage lost from losers' scores 0.1=%10
S(:)=1000 ;start every player with 1000 pts

for i=0,size(GN) begin

  if GM='T' or GM='N' then p=0.1
  if GM='A' then begin
     nW=size(W(i))
     nT=nW+size(L(i))
     Q=nT/nW
     p=(Q-2)^0.5/20
  endif
 
  totpts=0

  for j=0,size(L(i)) begin
    ptslost=S(L(i,j))*p
    S(L(i,j))=S(L(i,j))-ptslost
    totpts=totpts+ptslost   
  endfor

  ptswon=totpts/size(W(i))

  for j=0,size(W(i)) begin
    S(W(i,j))=S(W(i,j))+ptswon
  endfor

endfor

Last edited by Kryon (15.01.2012 11:54:39)

Offline

#33 17.01.2012 19:13:10

Kryon
Player
Posts: 337

Re: Importing the ranking

By the way, for LT9, I found a few inconsistencies between the player/winner lists of Marduk's data and website data. See:

http://old.longturn.org/lt9
http://old.longturn.org/lt9-game-over-v … eam-marduk
http://old.longturn.org/lt9-started

Can the two team captains (Marduk or Maho) remember all the winners? I am guessing the difference is due to idle players who were replaced.

Last edited by Kryon (17.01.2012 19:15:06)

Offline

#34 17.01.2012 22:27:01

Marduk
Administrator
From: Rotterdam, Netherlands
Posts: 151

Re: Importing the ranking

Kryon wrote:

By the way, for LT9, I found a few inconsistencies between the player/winner lists of Marduk's data and website data. See:

http://old.longturn.org/lt9
http://old.longturn.org/lt9-game-over-v … eam-marduk
http://old.longturn.org/lt9-started

Can the two team captains (Marduk or Maho) remember all the winners? I am guessing the difference is due to idle players who were replaced.

Correct, in earlier games it was more common for idle players to be replaced within the first few turns, in that case the substitute player could be counted as a winner (but only if he took over within the first 10 turns of the game).

Idler take-over can be a bit tricky: if someone went idle and was then replaced in the first 10 turns I may not have counted the idler as having taken part in the game (even though counting it as a defeat for the idler would be justified). If someone took over an idle accounter after 10 turns and went on to be defeated, I probably counted the defeat for the original player (the idler) rather than deduct points for the substitute player; if he won I probably didn't count the victory for either the substitute or the idler (but I don't think this every happened, taking over after 10 turns was usually quite chanceless). Only a victory for a substitute player who took over within the first 10 turns would be counted. There are some inconsistencies here, would be good to straighten those out. Anyway idler substitution only really happened in the first few games, and having a substitute player win a game was very rare.

Offline

Board footer

Powered by FluxBB