You're right this is the hardest part in coming up with these rules. Before these rules are accepted, I suggest that we should try as hard as possible to define what is considered stalling and what isn't. I know it's a difficult task, but we should try as much as possible to outline it. This is especially important since you can't rely on one person to do all of the enforcing, so everyone that enforces it need an established guideline that's as specific as possible. The guideline can be changed over time, if needed to combat abuse, but there should always be a guideline.

What you should do is take the avg time per move in games won, avg time per move in games lost, and run a statistical difference of means test:

http://onlinestatbook.com/2/tests_of_means/difference_means.html

you could run this for, say, a difference of 24, 36, and 48, or some other (see below) hours, testing for a 90 or 95% significance bound in each case. Furthermore, the bound could be only one sided, as we don't really care about the upper bound in this case.

Obviously, you would have to claim that the time taken per move for playing in wins is statistically independent of the time taken per move per playing in losses, but this is probably true.

And, of course, the more data a player has accumulated, the more you can be certain there is some degree of stalling. (as standard deviation goes down).

So, how do we determine what difference of means, when statistically reasonable, count as stalling? I think the simplest way to do that is to go out and ask a group of players what they feel are cases of definite stalling, while also checking the forums for the more publicized cases of them as well. You then run the difference of means test on the select cases, find the lowest statistically significant difference, and take an average of them. You should not publicize this number, otherwise players could, in theory, manipulate the system so they could keep barely under the threshold. However, it would provide a robust and numerically based definition of stalling, assuming you take enough cases.

When we see stalling, we know it. We don't need a mathematical definition for it.
And if we're worried about false positives, let's have a second or a third pair of eyes check the game. It should be enough.

Still vague. You have to remember this has to do with the upper echelon of the ladder. Some players don't know, or even won't surrender until their last bonus broken. With the screwy ladder giving way to fluky runs, I think these players could enter the top 25. Saying "I know what stalling is" gets far more complicated when we look at the ladder as a whole.

Also there is discussion of trying to automate this, to me, I wanna see what defines as stalling under this automated system.

If people are unaware that they they should surrender when they don't have a chance to win, rather than play until elimination, that's fine. They will get a warning first anyway. Then they will know.

It's not a question of awareness. It's a question of skill.

When you are in the lower ranks of the ladder, It's the "Oh shit I'm at 5 base income or all my bonuses are broken" which warrants surrendering at this level.

Top level, You can say "surrender when you know you've lost".

Only stalling that's done is by top players for the reward of 1st place. So I wouldn't have any concern over application of the rule anywhere near the worser players. Keep it to keep the top players in check.

EDIT - I think I misread something above but the point still remains relevant.

Edited 9/5/2017 18:48:48

Just wanted to throw out there a recent example of obvious stalling.

https://www.warlight.net/MultiPlayer?GameID=13947548

Patience should be taught in school.

Edited 9/5/2017 22:59:17

Playing till elimination is one thing, but how do you explain stalling to someone who doesn't know what stalling is.

@Benoit, yeah, that's clear stalling. But how far do we go, are we gonna punish a player for going 2 days + on a probably lost game. This is too vague, and will likely make it hard to police the ladder as a whole.

Blame the game not the player.

I've mentioned it to Beren, but a not very formal rule against stalling is never going to fix it, either being too tight or too loose, unless gameplay mechanics change. Force double turns, use banked time... whatever works.

Specially because, realistically, a consistently slow player is an annoyance much larger than a selectively slow one.

You are given a set timer, and speed isn't a skill WarLight demands, so you are adding rules on top of flawed systems - masking the problem, if anything.

"When we see stalling, we know it. We don't need a mathematical definition for it.
And if we're worried about false positives, let's have a second or a third pair of eyes check the game. It should be enough."

I will add my own This! To this..

Regarding stalling, I think there should be an automated basic filter, which will detect potentially stalled games (a very lenient filter, which will trigger in the most obvious easily codifiable cases) and light an in-game warning for the game (with no other consequences), so that the player knows he is now a potential target for stall investigations, this in itself may speed up some games.

Or maybe a 2-stage system, with a "yellow" warning which is only visible to the player, and a "red" warning which also automatically triggers a flag visible by admins (and make this system be public knowledge). Then once a "yellow" light is on, you know you need to either move faster or surrender in order not to get into the "red" "under investigation" zone. Yes, those lights would kind of hint that you may be underestimating opponent strength, which is disclosing hidden information, but in those cases this must be OK.

Edited 9/6/2017 19:51:20

(Haven't played on the ladder in years, so take everything I say with a grain of salt)

These all look solid, but I share others' qualms about the stalling rule. Seems like stalling only exists because there's a mismatch between 3-day boot and how long players are actually expected to take. Why not just change that? If you have the ability to overhaul the ladder, you have the ability to revisit some Warlight standards. Change it to some variant of 100% banked time w/ 24 hour turns + 2 days banked initially (but 3 day absolute limit), combine that with a UI change on the dashboard that lets you sort games by time remaining to play (shouldn't be too hard if you know the game's boot time rules + the time the game is waiting on you for, right? banking might complicate things and make it harder to handle on the client-side, but I'm sure you can figure out a fix for that almost trivially), and that should take out most forms of stalling at an almost conceptual level.

Having the rule manually enforced will probably mean it's only enforced in egregious cases or at high levels, unless you dedicate some resources to it. I see stalling as something akin to cheating in a class, where there might not be a clear-cut, perfect definition for it (although classes usually have some pretty well-defined rules there) but through some automated detection and some manual checking on the part of professors and TA's it's caught fairly consistently in organized classes. But that makes the whole system dependent on the effort of some maintainers (hopefully not Fizzer himself, since he's up to other interesting things), and I don't think that can be relied upon.

Regardless, this new system for ladders seems like a good one. Glad y'all are revisiting the rules.

TL;DR: If we're revisiting standards, why not just get rid of 3-day boot and go with a banked solution that closely matches how long players are expected to take?

Edited 9/6/2017 22:33:37

TL;DR: If we're revisiting standards, why not just get rid of 3-day boot and go with a banked solution that closely matches how long players are expected to take?

regardless of everything else, +1 to this. Not sure { 24h + 100% bank + 2d initial bank + 3d max bank } is the best option, but something similar to this may work

Having the rule manually enforced will probably mean it's only enforced in egregious cases or at high levels, unless you dedicate some resources to it.

and +1 to this as well. With a side note that probably most people only care about stalling at high levels, and high levels is what sparked this.

One way to account for what you call "stalling" would be, for example, a proper use of banked boot time.

+1

As I have written elsewhere:

I) All these rules are unnecessary with a better rating system.

Our real problem is the BAYESIAN ELO SYSTEM. It does only work for tournaments with everyone playing the same amount of games and getting all opponents or getting them assigned via swiss system. Bayesian elo system is not made for such a system like a ladder. It is just a poor system for a ladder.

BAYESIAN ELO SYSTEM IS THE PROBLEM.

All the changes are only an inefficient attempt to fix a bad rating system.

II) Rules itself:

1.) A player can't start playing on the 1v1 ladder with an account if they have another account with unexpired games. Games expire after 5 months. Violations will be punished with a warning and then bans.

Agreed, allthough the participation in the ladder may drop and the effect may even be negative. But worth a try.

2.) A player may only play on the Seasonal Ladder with one account per season. Violations will be punished with a warning or ban on all implicated accounts.

Agreed.

3.) A player may not play on the Real Time ladder with an account if they have another account which is currently ranked. Accounts become unranked after 3 days of inactivity on the ladder. Violations will be punished with a warning and then bans.

Agreed.

4.) A player may only play on a team ladder if there is no team consisting of the same players (not accounts) with unexpired games. They may join the ladder on a new team if at least one of their teammates is a new different player or if one or more of the players on the old team is not on the new team. Violations will be punished with a warning and then bans.

Like rule 1 for teamladders, allthough you can trick the system by using another puppet mate in your team.

5.) Egregious stalling can be reported and the game will be reviewed by Fizzer and the mods to see if it is a blatant attempt to thwart fair-play by stalling a loss. Automated tools will be used when possible to confirm stalling. Enforcement will err on the side of caution to result in the fewest amount of false positives. Violations will be punished with a warning and then bans.

Hard to detect and there is a huge grey zone. Good luck trying. And I want to stress, this does not even solve the problem itself, as runs will also be good in this system. For example I can play slower against good opponents and speed up as soon as I have an advantage. That's not really stalling...

6.) Manipulation of the force-finish in the Seasonal Ladder is not allowed, and the use of vacations towards the end of a season leading to force-finish will be considered especially suspicious. Violations will be punished with a ban. Fizzer retains the right modify results in egregious cases as has been done in the past possibly resulting in the loss of games and/or trophies.

Hard to detect, but worth a try.

Edited 9/11/2017 18:45:25

I don't think bayeselo is as big of an issue as you make it out to be.

It may not be the absolute best choice, but the real issue IMO is stil matchmaking, not the way ratings are calculated.

Tighten matchmaking and you'll get far more accurate ratings.

Examples of recent runs:

" PJ017 defeated Laith 13735188 6/30/2017 13:00:35 Laith: 1828
PJ017 defeated ALE 13722340 6/30/2017 10:58:36 ALE: 1542
PJ017 defeated Le Count H 13734063 6/29/2017 15:17:39 Le Count H: 1459
PJ017 defeated The Girl 13716985 6/29/2017 09:36:27 The Girl: 750
PJ017 defeated Grey Poupon 13716983 6/29/2017 09:36:25 Grey Poupon: 805
PJ017 defeated clamburglar 13717495 6/27/2017 08:02:16 clamburglar: 1197"

That top game shouldn't have been made. I'm not bashing PJ or saying he's bad, or even overrated. What I am saying is you shouldn't be matched with someone rated over 1800 because you beat people rated 1542 and lower. It should be more of a gradual increase, where you never play someone who is potentially out of your league.

Same goes for:

" Darkpie defeated CharlieBLue 13979724 8/24/2017 05:42:36 CharlieBLue: 1887
Darkpie defeated Anakin Chigurh 13866839 8/19/2017 16:07:18 Anakin Chigurh: 1893
Darkpie defeated Boubou 13857732 8/17/2017 18:18:01 Boubou: 1971
Darkpie defeated Laith 13852185 7/25/2017 14:01:59 Laith: 1828
Darkpie defeated Titorelli 13837345 7/24/2017 01:50:29 Titorelli: 1674
Darkpie defeated Stidsen 13837033 7/22/2017 19:24:12 Stidsen: 1068
Darkpie defeated Boja 13836628 7/19/2017 19:21:44 Boja: 1483"

You shouldn't be playing 1674 after 2 wins, and you shouldn't be playing 1828 and up after 3 wins.


In those two scenarios, those are actually good players, so it wasn't super harmful, the cases where it's truly an issue are like this:

"Hobo Jones defeated Blister 14068916 9/8/2017 23:59:43 Hobo Jones: 1834
Boubou defeated Blister 14064195 9/7/2017 02:49:01 Boubou: 1971
Bigchps147 defeated Blister 14066073 9/6/2017 23:41:03 Bigchps147: 1876
Wyindywidualizowany Indywidualista defeated Blister 14041413 9/5/2017 21:11:10 Wyindywidualizowany Indywidualista: 1848
Anakin Chigurh defeated Blister 14040949 9/5/2017 01:24:21 Anakin Chigurh: 1893
sloppyfatginger defeated Blister 14037832 9/4/2017 11:13:43 sloppyfatginger: 1775
Blister defeated Kha Khan 14009912 8/30/2017 09:26:53 Kha Khan: 1381
Blister defeated Ivan the Awesome! 14033679 8/29/2017 15:39:33 Ivan the Awesome!: 1406
Blister defeated JM 14015886 8/28/2017 22:56:25 JM: 1489
Blister defeated NZPhoenix (AHOL) 14000078 8/25/2017 07:13:13 NZPhoenix (AHOL): 946
Blister defeated kirsche123 14000077 8/24/2017 02:18:24 kirsche123: 1226
Blister defeated Moustache cash stash 13986042 8/22/2017 05:14:57 Moustache cash stash: 938
Blister defeated [FCC]Mortar man 13986043 8/22/2017 05:14:57 [FCC]Mortar man: 839:"

He won 7 games vs. people rated under 1500, and then all of a sudden gets obliterated 6 games in a row by people rated between 1775 and 1971.

He should never have gotten any of those matches, let alone so many in a row. Things like this completely ruin the integrity of the ladder. Those were all lose-lose games for the opponents. If they beat him their rating was likely to drop, if he beat one of them on a fluke they'd have had their rating obliterated.

Another step in the right direction would be to make someone instantly get removed from the ladder when they get booted from a game.

They could still rejoin, it would not be a punishment, it would however prevent massive bootstreaks from crippling not only their rating, but everyone elses they had beaten etc.

RTL has the option to remove a booted player and same should be followed by 1v1 ladder too.
Also what will happen to players with unexpired games on multipleaccounts from 5 months of this announcement-or when is the announcement date?

Will these rules be mentioned while joining ladders?

I've had once on my alt account (the one in Blitz now) that i was trying to start a run with 5 games but on the first time games were made i only got 1 game and my opponent there also just joined the ladder and surrendered against me directly when game started basically. This would give both of us a rating and for me that was already like 1600 so i got 5 opponents from 1500 to 1800 or so at that point as games 2-6 so rating system isn't really the best if you look at it that way as well.

That same run after like 3 or 4 games finished i was playing someone that was in the top 5 at that point on the 1v1 ladder just to give you an idea of how insane it can get. Never finished that run, ended in a 7-5 in the end but i feel like half of the names most would recognize.

I agree. I got Grona game when I was a newbie. As a result I played vs. 7 people in my first 20 games that had been rank 1. I lost 6 of those 7. Grona shot to rank 1 and carried me with him which resulted in me getting obliterated.

I think the "default" rating for people joining should be much lower than 1500. 1000 seems more reasonable.