31-01-2003 04:01									#1
	[Attempt at] New PBEM rating system Support Apolyton

hi everyone

as promised i now started a new thread to propose and discuss new ideas for PBEM rankings.

according from what i read in the previous thread, i conclude that the most important and most desired priorities for such a rating system are these:

1.)
depending on the course of a PBEM MP game, the points gained from it should be subject of change, in order to keep players interested in continuing a game as long as they believe to have any chances.
this was formerly tried with the 9x turns evaluation, which had the problem that the points not just changed but even accumulated. so..

2.) the points from a game should not be depending on the duration of the game, especially not on the past turns in that game, but only on the current situation in the game, most probably only measured by power graph.
this is to prevent players from staying on top of the list just by one long game instead of maybe several games where they win rather quickly. also, there's no need then anymore to stretch a game for the points only.

3.) if possible, the race for x9 turns lead should be fought.

4.) newer players should not need to fight their way up from way down the ladder but rather start somewhere in the "average zone" of the rankings. (my personal opinion).

according to these requirements/wishes i suggest the following system:

- each player starts in the list with 100 points.

- if a player enters a new game that should be part of the ranking, he antes a certain amount of points, for example 10. if the player has less then 50 points he simply antes a fifth of his points.

- the game has to be entered into the ranking when it starts and not retroactively somewhere in between.

- the total points anted are put into one 'pod' associated with the started game. each players ante share is remembered with the pod and game.

- at each 10th turn (10, 20, 30, ...) a screenshot of the player's power graph is sent BY the game initiator TO the person managing the ranking list (possibly me) at that time and according to the power graph the game is re-evaluated into the ranking.

- this is done in the following way:
1.) the players points currently associated with the game are deducted from his ranking.
2.) the total pod points are divided among the players according to a special system. gavrushkas tables look good for that purpose. also, that saves me a lot of additional work.
3.) the re-calculated points are added to the players rating and then associated with the pod in the new way instead the old one.

- a player who quits the game is simply regarded as last for the next rating calculation. he takes the points that the last player at that time would have, gets them back for his rating and is then withdrawn from the game. then the remaining points in that game are distributed according to the table with one less player in the game.

- if the game ends, the ratings are recalculated like in the 3 steps above (that is like it was an evaluation turn), only that the points gained for the rating are now definite and cannot be redistributed via this game.

so much for the description.
if i imagine this right (and im quite sure i do) this system will have the following characteristics:

- x9 turn rush wont exist anymore because the gained points will be retaken and redistributed at the next x9 turn anyway. (only that x9 turn will then be x0 turns, which is effectively the same)

- you wont gain massive amounts of points by stretching the game and staying in the lead all the time.
the maximum number you can EVER gain from a specific game is clear at the start of the game via the appropriate tables. the duration of the game will only redistribute the same points anted at the very start of the game.
this is because at each x0 turn, the points are RETAKEN before they are given out again.

- newer players will start somewhere in the middle with 100 points. player who consistently lose (talking about many games here, not the same one over and over again) will drop down in the rankings. on the other hand players who consistently win, will rise in the rankings.

- no player will ever lose all points he has, cause he at best antes one fifth of his points and he gets some of them back even if he performs badly in the game.

- points you gain are always points someone else loses so the total number of points in the ranking list will be #number of players# times 100.00 points (except for a few rounding faults)

a typical ranking list, after including a healthy number of games, might look like this:

PLAYER #1 / 123.65 points / 12 games
PLAYER #2 / 118.45 points / 13 games
PLAYER #3 / 115.69 points / 8 games
PLAYER #4 / 108.60 points / 11 games
PLAYER #5 / 106.33 points / 9 games
PLAYER #6 / 103.47 points / 5 games
PLAYER #7 / 101.66 points / 3 games
PLAYER #8 / 100.00 points / 1 game (running that is)
PLAYER #9 / 97.65 points / 4 games
PLAYER #10 / 96.44 points / 16 games
PLAYER #11 / 95.38 points / 10 games
PLAYER #12 / 93.45 points / 7 games
PLAYER #13 / 88.70 points / 8 games
PLAYER #14 / 85.21 points / 11 games

so what do you think ?

as long as the game initiators send in the power graphs regularly i would be willing to organize the rankings.

also, if someone knows an EASY and FAIR way to incorporate the power graph difference (very far ahead or just a little) into the rating i would be grateful and probably implement that as well.

	31-01-2003 06:08									#2
	Support Apolyton, buy Civilization: The Boardgame

Looks good and fair to me, Mathemagician. I will support it.

	31-01-2003 06:55									#3
	Tired of ads?

Is the redistribution of points based only on the relative ranking PG ranking for each x0. Or does it depend somehow on the players most recent delta in the Power Graph?

If the players leads for the first two x0 for example doe his rating rise twice? But the rising some how saturates after so many wins at x0.

Perhaps a simulation showing power graph settings for three or four turns and the rating results would help see how it is working? But it sounds good.

	31-01-2003 07:02									#4
	Tired of ads?

no, once the new rating day comes around, the previous ratings would be forgotten

	31-01-2003 07:13									#5
	Support Apolyton, pre-order Civilization IV

So for example if on turn 10 players 1 2 3 are ranked
as 1 2 3.

There ratings might be (just for example)
1 2 3
105 100 95

If the order in PG stays the same for the next 100 turns
the rating at the end will be
1 2 3
105 100 95

Then suddenly at turn 110 player 2 goes in the lead
so that the order is 2 1 3. Now the PG will be

1 2 3
100 105 95

?

	31-01-2003 07:16									#6
	Got spare money?

exactly. thats the idea. only that the total rating gets modified also by the other games a player is in.
but right !

	31-01-2003 08:28									#7
	Increase Your PM Length

I like it

	02-02-2003 20:33									#8
	Support Apolyton

What happens if someone takes over someone elses game? For example if you are a very bad player and you take over a nr.1 spot from someone else do you get his ranking? Or vice versa.

	02-02-2003 21:11									#9
	Suffering from ads?

hmm. good question m.darkheart.

i guess we can either find a mathematical system to incorporate this (which would be rather hard i guess) or we simply dont allow overtaking someone else's position in "official" games.

	02-02-2003 22:00									#10
	Got spare money?

quote: dont allow overtaking someone else's position in "official" games.

That would cause much trouble.

	02-02-2003 22:02									#11
	Full PM-box? Change here!

quote: Originally posted by Solver quote: dont allow overtaking someone else's position in "official" games. That would cause much trouble.

im open for suggestions though.

	02-02-2003 22:22									#12
	Support Apolyton, buy Civilization: The Boardgame

Sure . But for some reason I am not going to be erupting with suggestions. Good luck, though .

	02-02-2003 22:45									#13
	Support Apolyton

well, here are some ideas:

- you cannot take over someone else's position

- the leaving player takes his points when he leaves, the joining player isnt rated

- the leaving player takes his points when he leaves, the joining player antes the same number of points the leaving player gained, so his break even point is the position where he joined (that is, more risk for the joining player)

...more ideas to come...

	04-02-2003 08:02									#14
	Help yourself to an AD-FREE life

I', not sure I ee what would be the point in rating the game at each x0 using this system.

	04-02-2003 15:20									#15
	Support Apolyton, buy Civilization III: Complete

quote: Originally posted by Mathemagician well, here are some ideas: - the leaving player takes his points when he leaves, the joining player antes the same number of points the leaving player gained, so his break even point is the position where he joined (that is, more risk for the joining player) ...more ideas to come...

I support you Math. And I like the third idea .

But players have to agree to be rated BEFORE the first turn is launched. Else we would have the same "problem" as always: Bad starting position = cold feets = I do not want to be rated in this game .

I would like it more these way: Bad starting position = cold feets = Ok, I can't loose more point than I already have put into the pot - so let's play the game.

Edit and added:

Even all challengegames could be included here - though it would cost me lots of points

	04-02-2003 18:55									#16
	Help yourself to an AD-FREE life

Sounds good. It does take care of some of the inherent problems of the current system.

But have too many people lost interest in rankings for it to take off? I wonder....

	04-02-2003 19:57									#17
	Increase Your PM Length

quote: Originally posted by checkMate I', not sure I ee what would be the point in rating the game at each x0 using this system.

good point.

just to keep the ranking list up to date. actually nothing else. you might as well only rate games that are finished, but then i wouldnt have any work to do.

okay, maybe if we find some fair way that the anted points depend on your current rating somehow, the continuous rating would still make sense.

	04-02-2003 21:51									#18
	Put an end to popups!

quote: Originally posted by Mathemagician - the leaving player takes his points when he leaves, the joining player antes the same number of points the leaving player gained, so his break even point is the position where he joined (that is, more risk for the joining player)

I agree with Birdman about this third option, Mathemagician. I think this will work the best. And I also concur with Birdman's "cold feet" comments above.

In regard to, "Why rate every ten turns instead of just at the end of the game?" --- Because most multiplayer pbem games take years to finish, not weeks or months, and it helps to keep interest in the games while they are underway.

Keep up the good work Mathman!

	04-02-2003 21:54									#19
	Increase Your PM Length

One more comment ... I think there should be more of a punishment for leaving games. Maybe some of the leaving player's points should be distributed amongst the remaining players??

	04-02-2003 21:57									#20
	Support Apolyton buy from Amazon

quote: Originally posted by quinns One more comment ... I think there should be more of a punishment for leaving games. Maybe some of the leaving player's points should be distributed amongst the remaining players??

good idea. i will try to come up with something on that

also, the cold feet thing is obvious. i didnt even think about handling that differently.

	05-02-2003 07:15									#21
	Support Apolyton, buy Alpha Centauri

Hard as I try I just can't think of a way to improve. I seem to want to give some benefit for winning or maybe even just improving from x0 to x0, but it all boils down to someone being able to accumulate points by somehow sandbagging a good position.

	05-02-2003 07:46									#22
	Support Apolyton buy from Amazon

Perhaps this is a good time to describe the US Chess Federation Rating system.

The rating formulas are here: (Warning Math phobics should not click here. Actually perhaps only Mathemagician should check this)

http://math.bu.edu/people/mg/ratings/approx/approx.html

The following page shows the current distribution of all players. 99.5% are below 2400, all exept 5% are above 100.

http://www.uschess.org/ratings/ratedist.html

One thing that might make sense is to have a slighly longer interval say 25 or 50 turns. Count them as full games in terms of rating. And count it as a win if the lower player can make up more than a certain % of the ground over a given interval, or if the leading player can increase their lead by a certain percentage. Otherwise the interval is considered a draw. However, I don't know whether the exact number from the PG can be obtained to make these determinations.

	05-02-2003 07:54									#23
	Full PM-box? Change here!

well, actually im quite math phobic myself checkmate
or at least ive become so over the past months...

anyway, this system looks very similar to the Magic: The Gathering rating (where i actually came from before joining the CTP community)

i see 2 main problems with this rating system (without having a closer look that is):

1.) its too complicated too use without an automated system
2.) it only evaluates duels, that is 1 on 1 games, where one player is clearly the winner and the other one the loser, with the biggest discrepancy coming from a draw.

thus, i dont see how we (i ?) could include this into the system we are looking for, since both reasons above are big problems here.

	05-02-2003 07:59									#24
	Increase Your PM Length

We could introduce the notion of draws if we had a band of PG difference that was considered to small. But as for the rating formulas. Yes too complex.

They used to be quite simple minded.

You got 16+/- 4% of the difference in rating for
wining. And lost the same for losing.
You got 4% of the diffrence (given or taken) in the
event of a draw.

	05-02-2003 09:17									#25
	Support Apolyton, buy Galactic Civilizations

MY GOD another one trying to equate chess to civ will it ever end... Again different formula same problems all have been covered before...

Chess formulas do not work here without endless complicated rules to cover the CALL TO POWER GAME and it's endless variables that CHESS simply does not have... DOES anyone get this yet?

Look instead of trying to use chess as a basis of a ranking system here use this game as it's basis for a ranking system...

Gav's system works, it always did only a few minor gliches and a few chess ranking system, control freaks killed it.

Math your idea to have a base amount of points will work with Gav's system, your reworking of the entire system based on the CHESS way will not.

It has been covered a long time ago.
I do not mean to harsh here but dancing madly backwards to accomidate CHESS in CTP is hardly a new system. Been done and it failed Gav's did not it worked and more closely represented CTP than a chess based system ever will. This game has quantum times the variables to base a ranking system on skill alone...

Thier has to be compensation for land etc on and on all the variables skill alone does not cover. Just ask any millitary commander. He'll tell you yup we had skill, yup we had chess training but....Problems are simply part of the game. In other words they are not problems they are to be addressed...

If you do not address them you have simply neutered the game and the whole idea of playing for a rank.

Enough said good luck...

Black

	06-02-2003 11:12									#26
	Got spare money?

Sorry for the testy facts...

The reality is your concept of base wagers has addressed the only non factual misconception of Gav's system. The other complaint of playing 60 games to lead the leader board has not been addressed.

If we use you system and I play 50 games and you one...I lead the leader board...end of story back to chess based rankings. Which do not work here because quite frankly this is not chess.......

With your fixed points it addreses skill in the same way Gavs did with less ratical changes to the rankings, slow really.

The concept of more time between rankings creates quite an imbalance, quinns concept of nine turns is sound for many reasons.

I will suggest for the s&g that you seriously look at the cap you propose on the wagers with the fixed gains. Super impose that on Gav's system and see what you come up with...

To me on the surface I think it solves many issues and makes for a more solid base for all the variables civ has, let alone pbem...

After all Gav's system was conceived with civ in mind not chess and certainly not skill alone. That concept is proven to be flawed when it comes to civ, to many variables chess simply does not have.

Take the time to have a look at what I have been saying and what was just said.

Food for thought...

	06-02-2003 11:29									#27
	Increase Your PM Length

quote: Originally posted by blackice The other complaint of playing 60 games to lead the leader board has not been addressed. If we use you system and I play 50 games and you one...I lead the leader board...end of story back to chess based rankings. Which do not work here because quite frankly this is not chess.......

you do NOT. you only lead the board if the total points out of your 50 games is higher than the total points gained from the one game, which, assuming the other player leads the one game, would could for example be 1 lead for yourself and a little more than average for the remaining 49 games.
if you lose all your 50 games you will be WAY down on the ranking..

quote: The concept of more time between rankings creates quite an imbalance, quinns concept of nine turns is sound for many reasons.

wasnt it 10 turns in the old AND new system. just the last cipher of the turn number is different.

quote: I will suggest for the s&g that you seriously look at the cap you propose on the wagers with the fixed gains. Super impose that on Gav's system and see what you come up with...

didnt really understand that one...
s&g ?
cap ?
wagers ?

	06-02-2003 15:55									#28
	Support Apolyton, buy Galactic Civilizations: Deluxe Edition

Here we have a "pool" of same size during the whole game distribuated between the players according to their rank in the PG - unlike what we had in the first ladder, where the current no. 1 was matched with no.2, then 3, then 4 etc.

Then no. 2 was matched with no. 3 etc... etc....etc.

So NO Blackice - I do not see Maths suggestion is comming even close to a chess-system.

	08-02-2003 09:39									#29
	Put an end to popups!

quote: if you lose all your 50 games you will be WAY down on the ranking

If I have time to play one game only and am winning at it, the other player has time to play 8 games and is winning in 7 you mean to tell our ranking would be the same?

quote: wasnt it 10 turns

No nine it keeps the board moving and somewhat balances the leader board. If you just do it once a month some games do 20 turns some do 9.

quote: didnt really understand that one

Take your wager idea and use Gavs rankings would work quite well.

Yes I see Birdy I do like the idea as I have told Math before. As with Gavs system with wagers this one with the fixed amount.

	13-02-2003 01:16									#30
	Support Apolyton, buy Galactic Civilizations: Deluxe Edition

Any progress Math?!

If you need some "real" stuff, I have an excel spreadsheet with a lot of REAL gamesresults, that you might want to use to check your fomulars.

About 25-33% of the players are not active anymore - but you should still be able to use the results as a "real" test of how it could have been.