SE4 Rating System - Page 53

Slynky · #**521** March 14th, 2004, 06:23 AM

Quote:

Originally posted by Grandpa Kim:
Quoting myself:

quote:
A personal, 300 point drop does not scare me. Does it really scare any of you?

Well, to answer (gulp), yes, it does. Though my score reflect victories against opponents near my skill level (and losses to people I consider better than my skill level), I've spent quite a few games getting it that high. Not wanting to sound over protective of my score...I'd have to say I would be a bit apprehenisve risking "it all" on one multi-player game where I might have a bad position that, despite my debatable diplomacy skills, would result in a position I could not overcome. As an example, in Anklebiters, the first 3 people I met had the same colony type as me! While others might have been trading colony types, I had no such opportunity. To add to the misery, 2 had methane (so there was not wide ability to trade "breathers") and the other had NONE as I do. There are 7 rated players in that game competing. If I come in 7th among them, I would feel like it was a bit unfair to lose (estimating) around 100 points.

When it all comes down to it, fitting square pegs in round holes (

) like we are trying to do when we adapt a chess formula to a different sort of game, it's ONE game, not a Swiss System pairing of 7 rounds.

Having said that, I'd be willing to go for a 50-point gain (or loss).

I MIGHT feel differently IF all multi-player games were played on Geo's Balance mod with an uninterested party positioning all the players in fair starting positions (etc.) but that is probably not going to happen on the majority of multi-player games. Nature of the "beast" and all.

Another "addition" to rated games might be to add a rule that says: Before game start, if a player has changed his mind about having himself rated in any particular game, he may withdraw his request to be rated. This would give a person the chance to change his mind if the number of players requesting to be rated surpassed the limit of what he was willing to gain/lose.

Again, I just don't know...tossing out ideas.

geoschmo · #**522** March 14th, 2004, 06:30 AM

Scared, no. But I'd be a little ticked off about it probably.

I am not sure how we got off on the issue of fairness of starts. I guess that is a factor for some, but not so much for me. I think all that evens out in the long run.

Slynky · #**523** March 14th, 2004, 06:45 AM

Quote:

Originally posted by geoschmo:
Scared, no. But I'd be a little ticked off about it probably.

I am not sure how we got off on the issue of fairness of starts. I guess that is a factor for some, but not so much for me. I think all that evens out in the long run.

I think I started the "fairness of starts" comment. Fairness of starts over a series of, say, 10 games of 1x1 would tend to even out. But, in one game of a 10-player rated game, a bad starting position equates to having 10 bad starts in 10 separate 1x1 games.

Well, that's my thought, anyway...

.

Grandpa Kim · #**524** March 14th, 2004, 11:10 PM

Since my Last post I've come to realize there is quite a difference between one on one and multi-player.

Consider 2 scenarios:

20 rated players in a multiplayer game
20 rated players in a 19 game, one on one, round robin

In the first, there will always be one player out first effectively losing to all 19 opponents and one winner effectively beating all 19.

In the round robin the chances of one player winning (or losing) all 19 of his games is absurdly small. No-one in KOTH has put 19 victories together back to back, and I don't think it will ever happen.

This makes it both easier to win and easier to lose in multiplayer.

To accomodate this fact, I suggest computing the new Ratings normally then dividing by 2. This still leaves a substantial change in Ratings but recognizes that the same feat is much harder in one on one.

Slynky, remember that while you are playing that rated multi-player game you could easily play 20 one on one's. This should help smooth out the element of luck. Also you could be playing several, rated multi-player games and as Geo points out, things tend to even out. Some will give you a good start, some bad.

By playing many games, the element of luck diminishes. It is players like I, who play a small number of games who should be worried about luck.

Joachim · #**525** March 15th, 2004, 11:07 AM

Heh ya,
I really dont mind what system we end up with. But, it needs to be decided with the number of multi-player games going at the moment.
I would prefer a system that penalises the 1st loser in a 20 player game at around 3 times the loss of a one on one - so there is some extra penalty but not enough to stop people rating big games.

P.S, Slynky, I won my KOTH v Warlord Ragnarok, can I have some points so I have more to give you when you get a chance

Ragnarok · #**526** March 15th, 2004, 03:27 PM

Quote:

Originally posted by Joachim:

P.S, Slynky, I won my KOTH v Warlord Ragnarok, can I have some points so I have more to give you when you get a chance

Yeah, I forgot to report that one Slynky. Take away me points! It's all fun and games, right?

Slynky · #**527** March 16th, 2004, 01:32 PM

Quote:

Originally posted by Ragnarok:

quote:
Originally posted by Joachim:

P.S, Slynky, I won my KOTH v Warlord Ragnarok, can I have some points so I have more to give you when you get a chance

Yeah, I forgot to report that one Slynky. Take away me points! It's all fun and games, right?

I'll get to it tonight.

Still needing to come to an agreement on multi-scoring before we "cast the die".

geoschmo · #**528** March 16th, 2004, 03:27 PM

In the insterest of compromise though, I think Gpa's "Round Robin/2" suggestion is a step in the right direction. I also think Joachim has a decent point that losing a 20 player game should cost more then just losing one game, as the "fractional" system allows for.

The problem is how do you come up with a modifier that works for all number of players in a game? Dividing by 2 might give you a number you like for a 20 player game, but what about a 5 player game, or a three player game?

Here's a suggestion. Let's call it the "2/N" system. Take a 20 player game. The first player eliminated, you calcualte the points lost to each player as you would in the round robin system. But then you multiply each of those by a modifier which is equal to 2/number of players. In a 20 player game the modifier would be .1. So assuming equal rankings the player would lose 1.9x points. Each of the 19 winners would get .1x points. This is twice the amount gained in the fractional system. Plus keep in mind each player isn't getting an equal share here. x in this system would be a different number for each player, unless they all have the same ranking.

In this system assuming equal rankings the points would range fomr -1.9x for the loser, to +2.2x for the winner. Less on the high end but more on the low end then the straight fractional system.

The advantage of this system over a straight x/2 modifier is that the formula works for any number of players. 2/2 (1x1 game) = 1.

This system also has no point inflation since 100% of the points awarded to the winners comes from the losers. It has a little multiplication, but that's what we are wanting, to give a winner of a big game a little more then a 1x1 game.

primitive · #**529** March 16th, 2004, 03:38 PM

Reading down only Gramps seems to be against the principle of the �fractional� system, all the other reservations seems to be on the number of points on offer.

The easy way to increase the point-spread of the "fractional" system is by multiplying the results with a fixed number based on the game size. This can be just an easy formula like (1+ gamesize/20) or we could make a fixed table.

Or we could just go totally non-mathematical and make a table for different gamesizes. A little work to set up, but very easy for Slynky to use later

Slynky · #**530** March 17th, 2004, 01:36 AM

Games needing rating computation (in this order):

4 player multi-player (Alneyan, Belisarius, Gecko, Slynky)

Alneyan vs Primitive

Joachim vs Warlord Ragnarok

I can compute the Last game (it is dependent on nothing else) but if I compute the second game without doing the first one, Prmitive gets less points (possibly...and probably only 1 point less).

So, should I wait for what appears to be an agreement on rating multi-player and then do the second game or go ahead and go ahead and compute the second game (which, I suspect, is mostly up to Primitive).