Unofficial GLOBAL Ranking

[AlexAdam]

handicap = (turns - turns of #1)/game speed

Now we 're talking! You are absolutely right, I was thinking something along these lines myself. I'm going to get back to you on that, kind of busy right now

 

[TyfoonTurk]

problem with that approach is that players with less games have a better average and bad games have a huge influence at your handicap.

 

[tommynt]

yes, not submitting (bad) games doent make me (or anyone) a better player

 

[AlexAdam]

That is why you should not rate players according to a total average. Instead, score may be calculated based on handicap, but for each game on its own. Then, this score, which will always be positive, will be added.

 

[Half Nelson]

The handicap idea looks like a good one if someone is prepared to set it up. Looking back over the games I didn't submit, 4 were losses (2 deity, 2 complete disasters:- so guilty as charged on this issue) and the rest were invalid due to hardware / software issues causing me to replay turns. The only TSG scalp that I didn't eventually take after restart was the German deity game, although I only managed a domination victory with the France deity game (rather than the designated culture victory).

 

[TyfoonTurk]

Here is the top 10 of the handicap as implemented now (the average turns you need more then the winner):

extra.subm.handicap
1 tommynt 29 17 2
2 DaveMcW 177 23 8
3 ssjos 200 9 22
4 Neuro 386 15 26
5 Mazer Rackham 377 13 29
6 CharonJr 101 3 34
7 Max M. 104 3 35
8 Tabarnak 997 28 36
9 Attaturk 447 12 37
10 slugster 394 10 39

Extra is the the sum of all the extra turns, calculated as mentioned before.
Submissions the number of submissions.
Handicap is extra/subm, which translates to

The problems mentioned before (less submissions means better average/not benificial to submit bad games) are still here.

That is why you should not rate players according to a total average. Instead, score may be calculated based on handicap, but for each game on its own. Then, this score, which will always be positive, will be added.

I don't understand?

 

[AlexAdam]

I don't understand?

I mean, for each game, every player should get points according to his handicap for that particular game. The less the handicap, the more the points. Then, all the points from all his games will be added together to produce his final score.

However, before trying to work out a formula for that, I think it would be very helpful to answer the following questions:

1. What should be the point ratio between the 1st place and the last place winner? In the table I produced, this ratio is a whooping 12 to 1. 1st place winner gets 12 points, last place winner gets only 1.

2. Should the winners of medals get extra bonuses, or should the point system be entirely normalized accross the table according to handicap? (1st place winner obviously gets 0 handicap).

Your thoughts?

 

[Maxym]

Losing a 2nd city is recoverable, losing your capital is pretty much game over.

Most likely, but I would play it out anyway just to see what happens

 

[DaveMcW]

I ignore ranking systems that I don't think are perfect. But here is one I would actually pay attention to:

• Score = 100 - (100 * (finish_turn - fastest_finish_turn) / turn_limit)
• Losses, retirements, and wrong victory condition get 25 points for submitting.
• Ranking Points = Total scores from last 6 months, weighted towards recent games.
• Old scores decay at 16.7%/month

Note: These numbers can be adjusted depending on how much you want to reward extended participation.

 

[AlexAdam]

Score = 100 - (100 * (finish_turn - fastest_finish_turn) / turn_limit)

y=100-100*x/450... Using a linear function like this doesn't give enough credit to the fact that the faster you get, the harder it is to get even faster. For example, at the Immortal level, the difference between a 190 and a 200 turn finish is much more significant than the difference between a 350 and a 360 turn finish.



I suggest we use a logarithmic function instead like the following:



This is: y=100-10*log2(x+1)

I'm all for applying a "decay" factor for the older games.

 

[Ribannah]

Instead of using a scoring formula, you could use ratings to create a global ranking.

Ratings increase or decrease based on the difference between actual and expected result, where the expected result is calculated from the current ratings of all participants in the new game.

 

[DaveMcW]

The difference between a 190 and a 200 turn finish is much more significant than the difference between a 350 and a 360 turn finish.

I don't see much difference. They both finished 10 turns behind the leader.

Looking at my formula again though, it could be amplified a bit. The spread in results tends to be within 200 turns of the leader, so it gives 60 points for even the slowest win.

Maybe 100 - (200 * (finish_turn - fastest_finish_turn) / turn_limit)


Some more formulas for losers:

• Wrong victory condition: Same as winning, but divide by 2. Maximum of 50 points.
• Lost game: 20 * finish_turn / turn limit.

 

[AlexAdam]

I don't see much difference. They both finished 10 turns behind the leader.

You misunderstood my example. Let's use real data to make this clear. In GOTM43 tommynt finished in 127 turns. I finished in 210 turns, that's 83 turns difference. Now let's take into account wloff who finished in 293 turns, same distance from me than my distance from tommynt.

According to your formula the score would be:
1. tommynt = 100
2. AlexAdam ~ 82
3: wloff ~ 64

...as you see, tommynt's distance from me is of equal value to my distance from wloff: that's 18 pts. I don't think this is fair. I think tommynt's distance from me is much more important than my distance from wloff.

According to my logarithmic formula, the score would be:
1. tommynt = 100
2. AlexAdam ~ 36
3. wloff ~ 26

See my point?

 

[DaveMcW]

And my updated formula would be:

1. tommynt = 100
2. AlexAdam ~ 67
3: wloff ~ 33

I think that's quite fair. My goal is to make the rankings relevant for everyone, not just #1.

 

[tommynt]

Dave, not submitting games doesnt make a good player ... so if the rating should be about how people do in GOTM overal the raking system done is a very good one imo- submitting should be awarded. You submitted often and even some not so good games - that doesnt make you a worse player.
U still did do good more often as every1 else.

Apart that its all kinda meaningless anyway.

Half nelson might be best example - he deserve a overall good spot - just for submitting even subperfect games - he keep activity up and therefore does good for the GOTM overall.
Doing good often is better as doing great once

 

[AlexAdam]

Dave, not submitting games doesnt make a good player ...

I think Dave would agree, that is clear to all of us... We agree that ranking should not be based on an average of how well you do in your games, so that you would benefit from not submitting subpar games just to keep your average high.

The question is what exact method we should apply to a) make GOTMs even more competitive and b) be fair to all.

Soon enough I will recompile with both my formula and Dave's, so that we will have 2 lists we can then decide on which one makes more sense.

 

[Monthar]

According to my logarithmic formula, the score would be:
1. tommynt = 100
2. AlexAdam ~ 36
3. wloff ~ 26

With rankings like that I wouldn't even bother submitting unless I was certain I had the fastest win.

 

[Arilian]

Ohh, I am begginning to regret, I did not finish math in the university

 

[TyfoonTurk]

whatever way you put it, comparing scores is useless if not everybody participates in every game. compare it with the competition of your national sport, skipping games is not an option. however, we are handicapped by a thing called real life, which makes it impossible to do every game.

i still think the scoring system should refrain from absolute, trivial scores like 100 points for a victory. instead going for a handicap score introduces an intuitive and comparable score.

it's also possible to introduce seasons (like for each patch), where everybody starts at 0 at the begin of a new season.

 

[TyfoonTurk]

For the Gods and Kings season, 20 persons have submitted all four games. Implementing the handicap system leads to the following ranking:

	name		subm.	vict.	extra	handicap
1	tommynt		4	4	2	1
2	Minime		4	4	258	65
3	Derfel_82	4	4	273	68
4	Morcar		4	4	295	74
5	gkreitz		4	4	300	75
6	Arnold_T	4	4	301	75
7	Halcyan2	4	4	306	77
8	Dreadnough	4	4	311	78
9	pontias		4	4	319	80
10	Thresian	4	4	348	87
11	Suntechnique	4	4	358	90
12	Gamewizard	4	4	396	99
13	kamikazees	4	4	446	112
14	chaptak		4	4	481	120
15	DrZ		4	4	516	129
16	Cromwell	4	4	610	153
17	csaapproved	4	4	725	181
18	xiziz		4	2	213	107
19	dmccarthy	4	2	400	200
20	schmiddi	4	1	221	221

The handicap is the average extra turns you need more than the winner.