How the hell can be that I am 14th with a 8/8 record while for example ACT Tears is 5th with the same (8/8) record?
Or even Master Miyagi has got better position too (12th) with a worse (7/8) record than mines?

Or how can jz be 3th while Robin Hood 4th?
jz has got a 8/10 record while Robin Hood has got 10/11 plus Robin defeated jz!?

Just check out the rating of the players you defeated and you got your answer. Only oh noes got a high rating.
It's not very good, but you need to play fast and have strong opponents early on if you want to get ranked high.
Not a big fan of the rating. Many good players drop out after their first 1-2early losses cause they think they got no chance on a good rank anymore. This hurts a rating a lot too. A bit sad

Edited 10/27/2015 04:52:39

It is all even in the end, but during the season ratings can be very off.

I think it is not my fault that I was paired againts so weak players like Math Wolf or Jefferspin :P

But this is serious? Not enough if you play your best but you must write a wish letter to the Santa Claus?


Dear Santa,

I was so good, so please give me the followings:
please give me good enough opponents in the seasonal ladder (who will play all of 20 games at least).
please give me enough freetime to play fast enough my games to chance good pairings (and please give me opponents who play fast enough too)
please give me the important first moves
please give me the high 3vs2 capture % (some 2vs2 capture would be good too)
please give me the pick combinations with I can counter my opponents but they can't
please give me a red toy car

Love
Walter the cute

Yes as i said you need to get lucky with your early matchups. Also if you don't finish your early games fast, the ladder got no indicator wheter it should pair you up with good or bad players. Not a fan of this concept, but since it promotes playing fast i'm ok with it.

How the hell can be that I am 14th with a 8/8 record while for example ACT Tears is 5th with the same (8/8) record?

Ladder ratings are calculated using an ELO algorithm, not based on straight win-rates.

In the seasonal ladder, only your final rating matters, not ratings during the middle of a season. If you have a high winrate, as you do, the ladder will reward you by giving you games against higher ranked players. Which it did -- you got a game against the #2 person. Continue winning and you'll have no problem getting to #1.

Many good players drop out after their first 1-2early losses cause they think they got no chance on a good rank anymore. This hurts a rating a lot too.

There is a penalty for dropping out, but it's worth noting that your ELO rating is calculated before the penalty is applied.

That means there's a big difference between playing a player who has a poor rating due to dropping out of the ladder and playing a player who has a poor rating because they lost a lot of games. Even if they have the same rating, losing to the former won't hurt you nearly as much as losing to the latter. As long as the person that dropped out didn't suicide in any games, it won't hurt you.

There is a penalty for dropping out, but it's worth noting that your ELO rating is calculated before the penalty is applied.

It is true if this dropped out player defeated you, but what about if you are who defeated him? :P

If he would carry on the playing and get all the 20 games, it is high chance that his final rating would be higher because the + 40 points per game (and if he is a decent player that points would be much much higher because of his unrealized wins!). And because your rating is always change if your earlier opponents' rating change this means your rating will be hurt because of these dropped out players. Or I do not understand well the system?

Let me write a theoretical example: there are two players who are the best (with similar strenght and the rest players much weaker than these two one). One of them play very quick and got good opponents all the time and everybody of his opponents finish the season (got 20 games with many wins). The other (or his opponents) play much slower, and his first 5-6 opponents are very bad and many of his later better opponents left the ladder after few games. Of course both of them win all of their games.
These two players will be pairing againts eachother during the 20 games at all? :O
If yes and the second player defeat the first one, he will be the first for sure?

Or other, not so theoretical example: if a season would be only 11 games instead of 20, Robin Hood how can overtake jz if jz win his last game? Roobin Hood would have 10/11 record while jz would have 9/11 and Robin defeated jz, but I think Robin would stay behind jz :O
Is this ok?

And I know that impossible to create a perfect (in every aspects) ranking system but do you think this is the best achievable one?

^I completely agree with this. I think it wouldn't be that dramatic if borh players had only wins,but we had seasons where someone joined later and had a 19:1 record. He still had no chance on getting first. The rating system isn't the problem imo, but the pairing system is. Stalling an early lost game can make you end up 5-10 ranks higher due to the matchmaking system. No clue if there is a way to go around this problem with only 20 games played.

I don't really think there is a solution to this issue, since you can't control how fast people play. In an ideal world, everyone would finish each game before they receive their next one (so game 1 is finished by the end of day 3, game 2 by the end of day 6, etc.), but obviously that isn't a possible solution based on how multi-day Warlight games work.

You can game the system here, but it mostly encourages the top players to play quickly and join at the beginning of the season, which is a good thing.

20 games are way more than enough for the system not to be exploitable

we had seasons where someone joined later and had a 19:1 record. He still had no chance on getting first.

This is by design. I didn't want there to be any strategy where leaving/joining at strategic times gave you an advantage, so I made sure that the best strategy was to join the ladder before it started and stay joined to it for the entire season. If you join late or leave mid-way through, you're at a disadvantage, and it should be that way.

"The system is broken!"

"I know, I made it broken based on a set of assumptions I never actually validated. It works exactly like I designed it!"

Fizzer sounds awfully similar to a software engineer at Comcast.

Edited 10/28/2015 20:55:58

Fizzer, when do you need to join by in order to avoid the penalty? Is it before the season starts or before the first games are created?

I can see how joining late could give an advantage. It gives you more time to practice on the template and makes it easier to review the play styles of your opponents without them being able to see your play style.

I agree on this with Fizzer. it's a good Thing if players get an indirect Advantage if they're playing the season from beginning. It's also good that Players who finish their games fast benefit. In my opinion it's just too easy to manipulate the ladder with misbehaviour, but i don't see a way around it.

Don't Forget the seasonal ladder gives everyone only 20 games each season (not saying it should change) and it'd be impossible to make early matchups fair for everybody. If you Play on a good Speed, you won't suffer a lot in Rating.

JZ: it gives you 40 additional Points for every finished game, so theoretically if you join 20 days before the season Ends, you might get enough games. However you get a disadvantage in matchups and maybe you will never be able to contest for the top ranks.

Edited 10/28/2015 21:07:09

It's not about whether or not you play at a good speed, it's about the speed at which the opponents who are losing against you play. If you have people who play slow when they are losing then you are going to get worse matchups, which will affect your score for the rest of the season.

that's what i meant with "easy to manipulate the ladder with misbehaviour". one guy stalling a game can make the difference between 1st and 2nd rank

The thing is that attempts to correct these issues often create new issues in and of themselves. For example, the 1v1 ladder has an issue where people who're good lose a lot of games by boot when real life happens and suddenly inflate a lot of rankings (creating a sort of chain reaction). One naive way to correct that would be to weigh losses-by-boot less than regular losses, but that's gameable.

So is the stuff above- if finishing fast gives you more points, I could simply stall in a matchup with you (if I've lost) to hurt you. If starting early gives you more points, then I get a leg up on better players who play more games (and beat me) just by starting early. It doesn't make the rating significantly less skewed and just pisses off some players.

I'm curious as to whether Fizzer ran simulations before going with what he has now. At least that's a way to validate some assumptions given others (hopefully weaker ones).

Edited 10/28/2015 22:05:58

What I do not understand that why Fizzer doesn't use the community for the developing. I do not mean for coding but for example in a brainstorming form the community could find really cool solutions for existing weak points like in this topic, I am sure. This is a brain game thus there are many brilliant brain here who would help, I am sure.

And I do not say the present seasonal ladder ranking and pairing system is very bad (maybe it is really the best achievable one), but I would give it a try to find anything we can improve it.

For example what do you think to solve the stalling issue with a "seasonal time bank"?
I mean the boot time of the individual games should be reduced a lot in the seasonal ladder but create a "seasonal time bank" what means after the boot time reached in a game the time start to use from this pool. Stalling a game in the beggining would be not a good idea because it would mean you will have less time for the harder games or when you won't have enough time to play.

Or an idea for the pairing system:
what about if the pairing is not based on the present points but the present winning percentage first then then only after the points?

But this is only two first sight idea from only one (but very brilliant) man, I am sure we can find the best achievable solution and maybe Fizzer will be unaccustomed and surprize us to implement it.

And I do not say the present seasonal ladder ranking and pairing system is very bad (maybe it is really the best achievable one), but I would give it a try to find anything we can improve it.

Considering seasonal is only 20 games, couple of hundred players and has to start and end at a specific point, it has been working very well so far considering those constraints. I encourage you guys to find a coherent system with those constraints.

what about if the pairing is not based on the present points but the present winning percentage first then then only after the points?

So if I join and win 1 game, I have 1-0 record and 100% win rate, I'll be likely paired with #1 player...

For example what do you think to solve the stalling issue with a "seasonal time bank"?

That's a decent idea, but again it's all about specifics. What boot times and what bank would you propose? Also keep in consideration the participation rate, if the boot times are too harsh people will not be as likely to join the new ladder.


Also I'd like to say that the system is not designed to only solve 1st place vs 2nd place, it is used to rank all players. In the past when I was looking at ladders (unless the settings were incredibly ridiculous) most of the time top 10-20 consisted of players that I expected to end up on top. Almost all exceptions consisted of people who joined late (like Fatih going 19-1 but joining very late).

@Walter Ego
It is definitely a good thing to search for a better solution for any ladder, however in that particular case I am a bit sceptical if there is a way around the constraints of that particular ladder. Still, banking boot time has not been utilised a lot outside of rt games and rt ladder so there is room to experiment there.

I agree that Fizzer should have a system (like it does now) that encourages people to participate from the start. I like that it does that now.

I do think stalling is a bit of an issue, but that would be even worse if you did some sort of bank system IMO. They would just stall longer and know they have to take their turns faster later. Also, there are gray areas on what is stalling. Playing slow is not stalling, but taking extra turns when you have lost would be considered stalling. I think a lot of times just public shunning takes care of stalling abuse for the most part.

I would still like to see the auto join issue fixed. I am not a fan of being auto joined, because you end up with a lot dropping off. I would like to see more focus on the preparation of a new ladder by Fizzer though. It feels like he just throws a template up there quick, puts a blog post up, and that is it. There is very little build up to it, so people drop off when they realize they don't like it. Why not just expose it more in advance so the right people can join/leave and we don't have these complaints at all (does nothing about stalling, but fixes the other issues). Maybe some polls on what the community want? We have past seasonal ladders that are SORT OF the most played templates but with small differences (like Turkey and Guiroma). Something like the luck not being what most people use. So if there was a little more of a process on setting up the next template, you'd have a lot less complaints later.

Edited 10/29/2015 16:08:24