LongTurn League - official

[Corbeau]

Current ranking table!

Old thread: http://forum.freeciv.org/f/viewtopic.php?f=20&t=89912

For now, LTL is my "private" project that nobody else seems interested in, but I'll go along with it because I find it interesting and everybody else is free to ignore it If someone else finds it interesting, I will share the joy

The purpose of LTL is to try to keep some kind of score and create persistent ranking based on individual performance of players in Longturn games over longer periods of time. The main criteria for a Longturn game to be included in LTL is: No Alliance Victory. The reason for this is in the origin of the idea: to keep track of individual performance with as little interference by other factors such as popularity contests, luck with choosing allies, tagging along etc. As a colateral benefit, this reduces the effect of out-of-game actions such as delegation, pre-arranged alliances, collective grudges etc.

The first game I included in LTL was LT40.

The second game I am including is LT48 that just started.

Games are not server-specific and any game can be included as long as it is made public and available to all players (and that there is No Alliance Victory).

The calculation for LTL points is as follows:

Points for the first game (LT40) were calculated using the formula:

Points = ROUND( (inverse_rank / N)^2 * SQRT(N)*10 )

and then normalised so that the best player gets 100 points.

The original intention was to adjust the score according to the number of ACTIVE players, but I failed in that.

LT48 will be scored using a variation:

Points = 100 * ROUND( (inverse_rank/N) ^ 2) * SQRT(total_score/12,000)

where
- N is an average number of active players taken at turns 10, 30 and 50
- inverse_rank is almost self explanatory: the best player gets N, second best gets N-1 etc.
- total_score is the sum of all players' scores
- 12,000 is the slightly rounded (from 11,917) total score from LT40

The current ranking list (with rankiong from only one game at the moment) can be seen here:
https://docs.google.com/spreadsheets/d/ ... sp=sharing

Basically, I am still testing the waters so the scoring for the third game (whenever it happens) may slightly vary from this, but not too much.

UPDATE: Scoring for LT50 (if it happens) will use the same method as scoring for LT48.

[Lexxie]

I have been in favour of creating an ELO or similar system for a very long time. Sadly, in a game of diplomacy and alliances, making the formula require no alliances self-decapitates the project into disinterest and uselessness. However, the difficulty of figuring how to weigh alliances into the ranking has been the obstacle that stopped further progress. So it is hypothetically a good first base project, to isolate a system independent of alliances.

But let's not fool ourselves. In a FFA no-diplomacy no-tech based game, there is still a large amount of diplomacy and quasi-alliance behaviour going on.

Isolating this out makes the question a bit trivial, since what we are left with is a situation which others have already created and critiqued and peer-reviewed the formulas for.

To create true interest in the project, we must tackle a way to weigh the effects of 'externality', rather than pretend it does not exist.

For example, an inferior player does well in a game because a superpower (whether allied or not) eliminated all his immediate neighbours. OK fine, we find a way to discredit the player for this (or blindly ignore it and hope for sample size to even it out later.) But now think the other way. A moderate level mechanical player has superior results in the game because of best-in-class coalition-forming skills. How to even tackle and weigh such issues? I have some ideas but first we have to lay the foundational groundwork before any building commences.

I want to be clear. I am very in favour of a good project here, but it should not be built on formulas which have biased or naive assumptions baked into it. Now that these reservations are out of the way, I will also repeat that it's a valid approach to FIRST try to isolate the maths for the most simplistic case first. That is, once a healthy robust formula exists for the simplist case of 1v1 games, we can proceed to integrate more into it.

So to conclude, let's get a healthy 1v1 formula. That opens the gates for the possibility of this ever going somewhere.

How will such a formula look? Let's get into the real foundations here.
1. Existing ELO formulas - very good and tried and tested BUT only measure win/loss/draw. Inadequate as a base formula in large multiplayer games where there may only be one first place nation and everyone else is "less than first."
2. Formulas which incorporate any kind of score mean or median - problematic. Different map sizes, ruleset variations, tilesperplayer, and so on, will create radically different scores. There can be games where a 2000 player is ranked 1 and games where that same score is #14.
2a. Other problems: games finish at different times. A dominant player could lose in the very end and have a pathetic score. Or, a bad player could have neglected military strength and pumped a high score in economics, then easily be run over; in which case a summation or averaging of score over the length of the game would give this player a higher rating than the person who deliberately and fairly WON by not following this strategy.

First step forward: A way to relativise and wash away the "absoluteness" of game score values or summation of game score values, into a simple coefficient from 0.0 to 1.0. Without this foundation, the project already takes mutated DNA that can never grow properly.

[Corbeau]

I'll just repeat the first paragraph here.

For now, LTL is my "private" project that nobody else seems interested in, but I'll keep pushing it because I find it interesting and everybody else is free to ignore it If someone else finds it interesting, I will share the joy

Cheers.

[Corbeau]

A new game, LT50, is about to start. It has several options for endgame and some of them will exclude alliance victory. IF alliance victory doesn't happen, I will use the score for the LTL and use the same calculation method I used for LT48.

[wieder]

For the first time LT50 also has the mid-game rankings.

There is a lt.net specific great wonder called Atlantic Telegraph Company that will give all the players embassies with everyone. The effect is not lt.net specific and it's nothing that fancy. The new stuff comes from everyone getting the embassy with everyone once someone decides to do that + rank the players.

Once the ATC wonder is built, all the players are ranked by score and they get the mid game rankings for playing LT50. It's been a problem that the players are only ranked by the outcome of the game. The ATC wonder can be built with espionage (was with electricity) meaning it's taking place when there is about 1/3 of the game left.

[Corbeau]

Adjustment: LTL scoring will be applied for LT50, without an "if".

If the game ends with "alliance victory", mid-game scoring will be used.

If the game ends with score-victory, this final score will be used and mid-game scoring will be ignored.

Also: on T20 there are 23 active players.