Topic: Widelands tournament 2017: subscriptions started (and random chit-chat)

I actually calculated the elo rating that would result from the tournament, and it took less than one hour. I did it by settting the starting conditions in a datasheet (1500 for scoring 50% of points, minus/plus 200 points for every extra point scored) and then manually adjusting all the values until the expected score for everyone was very close to what they actually did achieve (to within 0.12 points, with the exceptions of janus and epicspartan who only won against the forfait and could not be fitted well). The resulting rating, while of dubious accuracy, has a surprisingly large distribution. For reference, a difference of 200 points means the stronger player is expected to win 75% of times, a difference of 350 points that he is expected towin 90% of times (actually to score 75% or 90% of points, including draws, but since drws in widelands are virtually nonexistant...).

So here's the rating as found. it does not include variations for the second phase, otherwise there would be a big boost of notabilis a small boost for me and worldsavior, and a large loss for sirver

If I know him well, einstein may want to develop a program to automatically fit the rating to a better degree of accuracy than my manual fitting, in which case I have no problems sending him the datasheet.

Anyway, those values aren't of much worth, since 5 games (4 for those who got a forfait, 3 for epicspartan and janus who forfaited last turn, 2 for toptopple who retired) are really not much to cconstruct a statistically significant rating.

Top

Quote

@Elo-List:

Intéressant... Implementing a ranking would be nice indeed...

einstein13 wrote:

WorldSavior wrote:

einstein13 wrote:

I understand that you want to play, but probably it will be better if you change the idea of whole tournament into well prepared games (it would be pseudo-tournament) :).

Would the games in the tournament not be well prepared? What do you mean with "pseudo-tournament"?

The tournament is well prepared. I don't have any wrong feelings about that! But if there is not enough people, we can make something not official. Idea:

Make some ranking with names and "points" for winning. More games you win = better place you have. Everyone can join any time. Example, we decide that first game will be on Saturday, 8 PM CET. Some players agreed on that (at least 3) and everyone can join. After all 6 people met, have a nice game, and everything is recorded. Next week Wednesday the same (different time). So there is another game. After some time (2-3 months?) we can see results where not everybody played every time, but there are a few people who are on the top of the list.

Aha, okay... So you had an idea about a ranking, king_of_nowhere had one and I thought that a ranking would be good, too. I'm wondering if someone would be against such a ranking.

king_of_nowhere wrote:

einstein13 wrote:

Make some ranking with names and "points" for winning. More games you win = better place you have.

I actually considered implementing an elo rating system (https://en.wikipedia.org/wiki/Elo_rating_system) for widelands. I didn't because we didn't play often enough.

On the other hand, while I am unemployed I have a lot of free time I may just use to do something of that sort...

EDIT: there is also the question of whether such a rating system should separate different win conditions.

It could seperate the different win conditions, but also provider one further list where everything is combined, right?

If not both is possible, I would suggest to separate the win conditions

einstein13 wrote:

I don't think that Elo rating will be good for us. This ranking was created for chess: game where you can definitely say that there are strict rules: two players; win, lose or tie. In Widelands it is much more complicated system. You have multiple players at once with win-lose part, but not every loser should be punished as much as others.

Created for chess, yes, but also adapted for soccer...

The game "Age of Mythology" had or has a ranking which also includes team-matches of two teams with identical numbers of players. I don't know how it works, but that shows that a ranking can also be possible for more matches than 1vs1

EDIT:

king_of_nowhere wrote:

WorldSavior wrote:

I would prefer a tournament where each round can have the duration "one week" maximally. Two or even 4 weeks are a long, long time.

So you prefer a tournament with large maps, where often one cannot finish a game in one evening

Well, you can combine a large map with an 4-hour-wincondition and everything is fine

, and you would also like that tournament to finish all rounds in one week.

Haha, no, it should not be necessary that it's faster than one round per week, should it?

Wanted to save the world, then I got widetracked

Top

Quote

king_of_nowhere wrote:

@ ranking: the problem with team games does not exist in this case, as all the games considered were 1v1. league of legends adopted a ranking system (which, if you ask me, was much better than the current tiers, for at least it meant something) where the team ranking was likely the average of all members of the team. yes, one could perform well and still lose cause of noob teammates, or he could feed all the game and be carried by the rest of the team. statistically, those even each other out, and in a large number of games you still get the individual level with good accuracy. It worked because teams were shuffled all the time, and there was a large number of games.

That is the problem, number of games. elo ranking works when you have many games to even out statistical oddities, so the more games, the more it is accurate. we just don't play enough. though if my plan of two tournamments every year kicks off, it may become feasible for the recurring players. Dividing the ranking for win condition is even less feasible: how many wood gnome games were ever played in tournament? if memory serves, none before the 2016 tournament.

Well, many people might play much more often than in tournaments...

Well, you can combine a large map with an 4-hour-wincondition and everything is fine

yeah, but the point of playing on a large map is to go really big, and terminating the game after 4 hours defies that. though there will be an instance of this happening, but it's not what i mean with "large map, long game"

I'd say: The point of playing on a large map is to play normal Widelands

Wanted to save the world, then I got widetracked

Top

Quote

@Normal Widelands
Widelands is different for any person. For me, normal game is when you have enough time to build up whole economy. Then the fight is between economies & military part. Making game faster, you can see for example situation, when only one person manage to make whole economy (but not working yet), and the second wins because he/she managed to make simplified economy working (f.e. producing lvl2 soldiers in huge amount).
I understand that some people like that, but for me it is not the "real game". I rather like big maps, where you can fight and defend for a long time and make the economy working and producing best soldiers.

einstein13
calculations & maps packages: http://wuatek.no-ip.org/~rak/widelands/
backup website files: http://kartezjusz.ddns.net/upload/widelands/

Top

Quote

Wow,either I haven't checked the forum in a while or I need glasses... either way,I'd love to play!

Top

Quote