I would like to bump this. Especially if there will be new seasonal ladders introduced, I think there should be a fundamental change to the shelf life for ladder games. 3 months seems like far too short of a time frame, considering that some games can take upwards or even longer than a month to complete.

There's a lot more I can add here, but I feel like it may come off wrong or people will take the wrong meaning from it, so I will hold off.

I still say the best idea I've heard is using your last 100 games as the benchmark instead of a time frame.

I have not heard that idea, but I agree that it sounds awesome. One of the current flaws with the system is that the number of games you play affects your rating pretty severely (generally better to play less), and with games constantly falling off your page, this is constantly a pretty relevant factor. Setting games to be removed based on a static number rather than a time frame would be a really good fix to the problem. I like it.

Again, agree with almost all the suggestions made in this thread. My personal favorite options being a longer shelf life and/or the removal of a time frame.

|> *One of the current flaws with the system is that the number of games you play affects your rating pretty severely (generally better to play less)*

I've told you this wasn't true before, but I guess you didn't believe me. So I made a graph. Hopefully this will bust this myth forever, as there's clearly no correlation:


Playing fewer games makes your rank more volatile, but it doesn't make it better. This myth comes from the fact that you notice players who's volatility gives them a good rank, since everyone watches the top 10 more closely. You don't see all of the players when the volatility works against them.

Excel's trendline (not shown) was pretty much flat - in fact, it even angles down slightly. So if anything, playing fewer games hurts your ranking slightly, but it's probably within the margin of error.

I actually do believe you, and I believed you then as well. We're not arguing the same thing, though, really. I'm not arguing that you can't be a high ranked player by playing a lot of games, or that you can't be a low ranked player by playing only a few games. Playing less games makes your rank more volatile, and volatility is basically the key to "success" in a rating system where games quickly fall in and out of relevance. If your goal is to be consistently be ranked in Xth place, where Xth place is basically your accurate ranking based on true skill, then you would want to play as many games as you possibly can. If your goal is to spike for a period of time and hope to hit really high, even if short-lived, then you want to play as few games as you can reasonably can. For most people, success on the ladder is measured by "how high you can get" not "how long you can maintain a stable rating" which is why volatility and less games is the ticket to win for most.

Does it ultimately matter that much? Not really. However, I think if a new seasonal ladder springs up based on the survey, it would be cool to have the original ladder become a little more consistent. Basing a player's rank on total number of games played rather than games played over a time frame would remove a lot of the volatility in the system.

I know I argued for this a long time ago, and basically gave up on it because nobody cared or was on board with it. However, I think with a seasonal ladder where player rankings will be very swingy based on the nature of the ladder, it's the perfect time to move to a more consistent rating system for the original ladder. It seems as this time I am not alone in hoping for these changes.

I suggest to add Ladder tournaments every three months. In this tournaments will be involved first 10 players at the moment. Tournament should be robin type. Games inside this tournaments will be count as any regular Ladder game. However, minimum played ladder games aside of this tournament would be still active. This will slightly incrase the ladder games played.
I think that this will give more accurate skill measurement at the top of the ladder and give additional motivation to players to get into top ten and participate in one of this tournaments.

I agree with Impaller. Though players improve and expiring games would better rank improved players. I suspect this is one reason why the Fizzer of Oz (the Godfather of Warlight) implemented the current system.

Tournaments: Top 16 (like NBA Playoffs) or Top 12 (like NFL Playoffs, in which best 4 have a bye the first round) would be enjoyable.

What I'd like more than any change: 4 starting armies instead of 3.

Picks are important. But: the less starting armies we get, the more important picks are. 1 more starting army would give the skillful deployment and use of one's armies more weight. Winning or losing on picks (I'd define 'picks' here as (a) what you pick, (b) how you order your picks vis-a-vis the opponent's picks, (c) who gets what pick first) alone happens more in a 3-pick setup than it does in a 4-pick setup.

A 4-pick setup would:
- slightly decrease the importance of pick ordering (ie, which of my picks are 1, 2, 3, 4, 5, 6, etc.): if you only get 3, which of your picks you get and which you don't get matter greatly; with 4 spots, this is less decisive.
- decrease the importance of who gets first pick: same reasoning.

Thus: The pick-skill-chance balance (currently about 50%-40%-10%) would give slightly less weight to picks, slightly more weight to skill, and also reduce chance (1 more starting army is like insurance against unlucky picks vis-a-vis the enemy and unlucky pick ordering -- ie, who gets first pick). If the goal of a 1 v 1 is to see whose overall skill set (ability to make good picks & ability to make good moves) is best, a 4-pick setup would better measure the totality of one's skills.

The best boxers rarely lose to average, decent or fairly good boxers. They find a way to beat inferior opponents. Lucky KOs rarely happen to the best boxers. But they happen rather frequently in Warlight 1 v 1s. If the settings were better, Warlight 1 v 1s would better resemble a boxing match: only the best 10 opponents (or people rated 1800+) would have a real chance to beat the best (Impaller/Retro/HHH/Zaeban/Fizzer/Byt/Etc.). Take Impaller as an example. He rarely loses. Yet certain games he has lost should've been wins. A clearly superior player losing to a guy rated 1600-1700 because of chance/luck? This is evidence that the settings should be adjusted. I don't follow chess, but does a guy rated 2000+ ever lose to a guy rated 1600-1700? Also, take away a couple of those losses (which sting the rating more than other losses), and Impaller's rating would be 2100+, which I think would be a fairer measure of his skills.

On a related note: afaik the shelf life for 2vs2 is still 3 months, while it was said that it would increase to 5 months. A bug or a feature?

Also, I'm (still) a fan of weights depending on time. Letting games within 3 months count max, 3-6 months for 60%, 6-9 months for 30% or something like that (enough variants possibly) should be easy to implement. (easiest way: games 0-3 months count as "3 games", games 3-6 months count as "2 games", 6-9 months as "1 game" should do the trick. Basically: for every game finished, a virtual copy of 3 months old and a virtual copy of 6 months old is added). A player needs to have 30 games finished to enter the ranking, which can be 10 recent games, of some recent and a bunch of old games.

I wanted to see how the continuously decreasing rankings would look like (together with crafty) a long time ago, but never found the time. :-(

I'd like to bump this and add my two cents.

I think everybody would agree that when a player first joins the ladder, and his rating is based on only 10 games, then it is not necessarily an accurate reflection of his skill.

Now, with a ladder game shelf life of 3 months, suppose:
* A player plays in 2 games at a time.
* Although the boot time is 3 days, the game moves at a pace of 1 turn per day.
* An average game lasts 20 turns.
This would mean that, in 3 months' time, you would finish only 6 games. While in reality a lot of people play faster than this scenario, the scenario is quite realistic.

While I agree with the idea that games should expire, I think that there are many ways to improve upon the exact expiry criteria (currently set to a sharp cut-off at 3 months). I would propose that games are weighted so that they count less the older they are, but this is just one suggestion among many good ideas that have been posted in this thread.

Whether there is a correlation between the number of games you play and your rating is irrelevant - the problem is volatility. I don't think anybody can disprove that the rankings are volatile with the current system, and I don't think it is a good thing.

- I consider the last 100 games (or whatever the threshold should be) a good limit,
- with superior weight for the recent ones (say +++ in the last 2 months, ++ from 2 to 6 months, + from 6 to one year) and a full expiration after a longer period (say a year).
It's just a mix between different ideas widely brought to the forum, though this way active play is still more rewarded than past ones and expiration of active periods is not so hard to accept. Moreover, in order not to limit abuse on the top of the ladder with players wanting to stay high with no addition of games, a player would stay in the ladder only if he plays above a minimum number of games each sliding month, like 4 games started in the last period of 1 month or something like that.

And I'm a big fan of the idea of, well, "Masters series" if you may say it so : periodic tourneys of high-rated players, either RR of elimination (RR if not too large, with 10 it doesn't seem too demanding), possibly with a link to the tourney on the home page...

Afaik, last 100 games is not possible. Then a game would at a certain moment count for one of the players, but not for the other, which would bias the rankings.