• 98 posts
  • Page 3 of 7
The_Bishop wrote:
Thanks @ dough_boy for clarifying, and yes...
The new rule is: one cannot lose more than 50 rating points per game

1st = Losing over 50 is too bad, but winning 160 in one fell swoop is not too good (??) I really can't understand the whole reasoning that led to such a rule.

2nd = The new rule only serves to help a handful of players that really don't need the help. And the worst is that points are deducted from legit winners (very often beginners).

3rd = Just do the maths. If a player deserves to lose 70 but he only loses 50 then it's a 20 points more for him, fine! When it happens 5 times then it's 100 points more, wow nice for him! 10 times is 200 points more, amazing. Here we play hundreds of games, and well, in a hundred it will be just 2000 points more. Wait... I feel like there's a couple of players which are currently running for a 5 digits score. I call it "score" because it is not a real rating anymore, it's just a score like in videogame sites.

4th = The more one grows in score, the more the 50 points-cap becomes helpful for him. Unfairness... Probably in the long run someone will even achieve a 6 digits score and we should begin to worry about how many digits the rating variable can actually contain in the code. I would never want, you know, a sort of "millionium bug" to occur, something that might cause the site to fall down.
«God doesn't play dice with the World» ~ Albert Einstein
Cireon wrote:
1. As staff member, you had insight in the research that Matty did into the actual numbers. We found that a rating limit of 50 had an effect on such a small number of games, that it wouldn't make a big difference in the grand scheme of things. Yes, there will be a really small percentage of games where a winner may not get all the rating they "deserve", but they will still get a lot of rating. Since this logic only applies to games with three players and higher (two player games have some special logic that makes their effective cap 30 rating lost, this has existed for many years now), it would probably enough to skyrocket them up the better part of an entire rank with one game! So yeah, that's the bad stuff, but it only affects a very small number of games, and in a very small way. It also affects a much larger number of games in a positive way: higher ranked players will actually play with newbies again because there isn't a gigantic risk, which leads to a much nicer mixing of players, and a better community for everybody.

To comment on why winning 160 rating in one fell sweep isn't too good. If you play 9 player games with players that are exactly as good as you, you will win one out of 9 games. Every game you lose (so 8 out of 9) you lose 20 rating each, so to break even, in that 9th game that you win, you earn all that 160 rating back.

2. As explained above, this rule wasn't put in place to benefit the highly ranked players. The benefit for them is actually super minimal if you look at the actual statistics and maths, which you don't seem to have done, given that you're throwing random numbers around in the post. It benefits the community as a whole, because people can actually play with other players of any rank much more. If you ask a newbie: would you rather miss out on a bit more rating once every 1000 games, or never play with highly ranked players, I think people would really prefer playing with highly ranked players instead.

3. That's not how the ranking system works. Rating isn't created out of nowhere, so it isn't "more points for him". Otherwise, yes, your arithmetic is sound. If a player is involved in a hundred games where they would have lost 70 rating , but they actually lost 50, there is 2000 rating that that players still has. To lose 70 rating in a game, you need to lose from a player that you have more than 3.5 times as many ratings as. The top players of this website have around 6000 rating right now, so for this scenario of 70 rating to happen, one of these players needs to lose against a player with 1700 rating or so in a >3 player game. Realistically, that doesn't happen very often.

4. It is unlikely somebody will ever reach a 6 digit score. The rating system is balanced in such a way that rating tend to remain in the 1-6000 range, because rating doesn't enter or leave the system (some rating trickles out due to some rounding complexity in team games, but it's insignificant to the website's scale). As soon as somebody hits 5 digits, they would win only 2 points per game per player from playing brand new recruits. Let's say they play experienced players (around 2000 rating) all the time: they would win 4 rating per player per game. That means for every game they lose (50 rating), they would have to win from 12.5 players. That could translate to winning half of all 7 player games, which is really really hard! Long story short: the system is designed to deal with this, and the 50 rating cap does not significantly influence its ability to self correct.

Regarding your "millennium bug" comment: don't worry. We can easily support rating values of up to 2 billion, and could be increased to 9 quintillion (that's a 1 with 18 zeroes) with a single code change.

I want to finish off by clarifying one thing: you're not wrong. This rating limit is not without its disadvantages. However, all credit goes to @Matty here who has done a bunch of research into the actual effects this change would have. These effects turned out to be really small. Putting them side by side with the positive effect of having more high level players interact with the newer players on this site, this was a trade-off we were willing to make. We will monitor what happens with the rating values over the next few months. If it turns out the problems with the rating cap are bigger than we expected, we can always revert.
“This is how humans are: We question all our beliefs, except for the ones that we really believe in, and those we never think to question.”
- Speaker for the Dead, O.S. Card
Cireon is online.
The_Bishop wrote:
First of all, checking 2 games upon a million cannot be called "doing a research":
Matty
I have browsed through a couple of finished games from slackbatter, who is the best player rating wise on this site for a while now, and he does play newbies, but not that much. He rarely loses more than 50 rating too (usually between 30 and 45).
Source: https://dominating12.com/forums/6/suggestions-feedback/3287/rating-disparity-ratio/post/59504#post-59504

And the so-called research has already been denied by the person concerned:
slackbatter
I've had games where I lost over 70 points
So different from saying "rarely more than 50" and some simple maths can show how many points a 6000+ can actually lose.

Let's say player A has 6400 rating, he plays against eight guys having 800 rating each and one of them wins:
the average rating of the losers is (6400+800+800+800+800+800+800+800) / 8 = 1500;
A loses 20*6400/1500 = 85 rating points (rounded down), capped at 50 by the new rule;
all the other losers lose 20*800/1500 = 10 rating points each (rounded down);
the winners receives 85+10+10+10+10+10+10+10 = 155 rating points, reduced to 120 by the new rule.

Notice that poor guy: he received only 120 points, and if he would have won against same rating players he would have been rewarded with 160 full points. That's absurd.
So it is better for him not to play the champions of the site: if one of them enters his games then he won't get a full reward in case of win. In a "very deep research" of 5 minutes I found at least 3 games where each winner got rewarded around 20 points less than normal. The virus is just multiplying itself creating hundreds of unfairness. It's only a small percentage of games, that's true, but still that small percentage harms the system in its entire.

Then the second Cireon's point is already denied by the previous maths: high ranks are encouraged to play low ranks (sure they lose no more than 50), but lows are totally discouraged from playing highs, they'll be actually rewarded with less points.
I think there should be a warning for this, something like:
"A very high ranked player joined your game and this can cause you to win less points than expected, play it at your own detriment".

The third. I know the rating is not (in theory) created out of nowhere, unfortunately in this case it's just stolen from poors in order to feed the riches. And true, they are not "2000 more points for him" but "2000 points not lost by him", it's kinda matter of philosophy but I second this. And I mirror this sentence: "That's not how the rating system works" with the same sentence, because of this:
Cireon
To lose 70 rating in a game, you need to lose from a player that you have more than 3.5 times as many ratings as.
?? Absolutely not true: in games with 3 or more players the rating of the winner has no part at all on determining how many points the others lose!

Fourth. No entries and no leaves of rating does not ensure that the disparity among players cannot grow. Instead the new rule is just ensuring the opposite. And still, for every new user we actually create 1000 points out of nowhere, this can cause quite an inflation of the rating over time, whatching the situation along the years I'd say the inflation occurred but fortunately is just a tiny. And again still the same mistake... The points awarded are fixed: we win (we won really) always the same, whatever the ratings of our opponents are. So it is not possible for anyone to win only 2 or 3 points per player, not even for one having a 9 digits rating. Excluding small deductions, due to the approximations, we win: 160 in a 9p game, 140 in a 8p game, 120 in a 7p, 100 in a 6p game, and so on... The losers "pay" for our win dividing the "cost" in proportion to their ratings (just as showed above). If one of the losers is a very high rank benefiting from the 50-point cap, then we win less points than usual: that's the worst part of this "rule".

I think the rating system we have (as it has been until May 24th) is just excellent, it represents the skills of a player for real (more or less some fluctations) and it is not something you can find as good as our in every site. The mathematical formula is simple and easy and by itself it menages to keep the balance of the entire system. No caps, no adjustements are required. For me it's the most valuable heritage left by 4myGod to the site (including Africa 1890 also). There's no reason for making any changes, especially for just one complain over a bunch of years!

Everybody here is saying he finds no issue on playing "mixed games". Slackbatter is the one that more than anyone else knows (for personal experience) how the rating works at the higher levels (being on the top for several years). And he's just saying that he feels like he can actually get higher by playing beginners only and someone keeps saying he wants to incentivate him even more by allowing-forcing him to steal points from others. He doesn't want such a help, God!

Really at this point I don't know what else can be said to restore the rules.

We can make a deal.
As soon as someone achieves the goal of the 5 digits the new rule will be removed, everybody being allowed to exploit the 50-point cap at his best... At least it would be funny!
Even better, we make a poll and we let the majority decide if they want to restore the rules immediately or once the 5 digits are achieved.
«God doesn't play dice with the World» ~ Albert Einstein
Cireon wrote:
The_Bishop, you are honestly getting a bit frustrating at this point, selectively quoting things to make your point. slackbatter saying he lost more than 70 rating on some games does not mean he denies the research. We never claimed there wouldn't be games adjusted by this new rule, we just said that those games would be very rare. I am also sure that when Matty says he researched "a couple" of games, he doesn't literally mean two games, as "a couple" is often used to mean "a few", especially by non-native speakers of English.

Your example isn't quite correct I'm afraid, but it is close enough. Yes, you're right, in that very extreme case this cap has a pretty big effect. Nobody on this site actually has 6400 rating though, and them playing a game with just people of rating 800 is extremely unlikely. I am sure you could cherry pick a handful of other examples where the rating cap affects the final stance of the game, and if you read back, we have never denied so. However, in the large picture, the number of games that this would happen is very small. Since at this point you don't seem to believe me on my word, I decided to actually give you some data. Below is a histogram of the number of games people have lost recently:

Histogram (click to show)

As you can see, there is indeed a small increased bump at 50 points due to the cap, as expected. However, of all "losers", only 0.30% has their rating capped. Looking at games in general, we saw that around 0.68% of games were affected by this rating cap. So let's back up for a second: about 1 in 150 games are affected. So a player may get a bit less rating once every 150 games on average. That doesn't sound so bad if you take the advantages into account.

Instead of trying to debunk your other points, I want to respond to only one of your quotes:

unfortunately in this case it's just stolen from poors in order to feed the riches

I don't even quite know how to respond to this. First of all, there is no "rich" or "poor" here on this website. Some people are better at playing Risk than others, that's the truth. Second, you imply that this all is a scheme to make the highly ranked players even more highly ranked. I am not sure where you get that idea from. If somebody wanted to not lose points to newbie players, they could already use rating limits and only play against higher level players. This cap is only there as a safety net for players who want to play with newbies. Also, interestingly, when doing the research for the chart above, one of the games I found is game Game 1017558. Here we had maafi and slackbatter (two high level players) playing with a bunch of lower level players. Maafi won, but because the average rating was so low, slackbatter lost only 50 rating. This means that maafi, another high level player, lost out on some rating there. So this goes to show that everybody is affected by this equally.

As soon as someone achieves the goal of the 5 digits the new rule will be removed, everybody being allowed to exploit the 50-point cap at his best
The fact that you end your post with sarcasm makes me not want to take you serious any more.

Even better, we make a poll and we let the majority decide if they want to restore the rules immediately or once the 5 digits are achieved.
You seem to imply that the rules will be restored. They will not, unless we observe problems with them in the future.
“This is how humans are: We question all our beliefs, except for the ones that we really believe in, and those we never think to question.”
- Speaker for the Dead, O.S. Card
Cireon is online.
evil_house wrote:
I agree with The_Bishop.: No cap on points lost.

Moreover, I do not have clear what rating system is used, but I would go in favour for using an standard one (ie in chess, glicko rating system).
dough_boy wrote:
Personally I play to have fun.

I don't understand what the ultimate reason for the max point loss change was. It has always been a % of your rank correct? So you shouldn't look at it like man, I lost 70 points to someone who lost 20 points.

I know you have stats to back up how many are affected or not affected, but the fact that anyone can be affected means it is one too many. Anyone asking for there to be a cap doesn't understand how it works in the first place.
Virtuosity98 wrote:
Note: I am neutral in this topic - I am neither in favour of or against the cap on rating losses. I just want to clear up some confusion as to why the cap was introduced (it wasn't to protect the highest-rated players from heavy losses).

The reason that the cap was introduced was because there is very little integration between high-rated and low-rated players on this site. High-rated players did not want to play low-rated because of the increased randomness of games with new players - more games end because of a mistake/suicide/turns missed/rank-targeting. Due to the rationale that high-rated players cannot use their "skill" to influence these highly random games, for the vast majority of high-rated players it is simply deemed not worth it to risk so many points.
The cap is designed to encourage more integration between the new players and the experienced ones.

So the question is, which do you value more?
Increased interplay between high and low ratings at the expense of "fairness" in 1/150 games
OR
Increased "fairness" in 1/150 games at the expense of increased interplay between the high and low ratings

I think the negative effect of the cap is quite limited, and the potential benefits could well outweigh the cost. BUT if we see after a few months that the high ratings are gaining too much or exploiting the cap, and the data back this up, I'm sure the cap would be adjusted.
It is now Day 8. Please submit your Lynch vote, as well as any Role-specific Day actions you wish to perform (countdown).
Day Actions:
• #LYNCH [player], #NO LYNCH, #ABSTAIN in forum thread.
• Role-specific actions (via PM with V98).





elysium5 wrote:
What Virt said and thanks for saving me a lot of typing!
"Bad Deadpool... Good Deadpool!"
dough_boy wrote:
Did you ask the high rated players if they will now play lower ranked players because there is a cap? If you did I would assume they would say no, because all of the other issues still remain. Are "we" tracking data to see if more higher rated players are playing more lower? Do they even know that the cap exists?

I don't play 1v1 because of the rating issue. I don't play because it is too much of a crap shoot due to the drop.

Also, I join games based upon the settings I like, not who is in it, or what their rank is.
maafi wrote:
I am also neutral. I had no problem with the rating system as it was and I have no problem with changing it slightly if it means games are filled quicker and more are created. But I think Bishop is right to question the change and the reasons behind it. I'm just not clever enough to understand the mechanisms and implications.

And I too join games based on settings - and also if a friend is already in it. I used to create games with an approx. 2k rating limit but not any more. And I don't play lower rated players to win easy ratings, nor do I avoid them for fear of losing a large number of points. It is about enjoyment of the game. Yes, lower ranked players might make more mistakes and yes, that might cost me a game and a bunch of points but it also adds a little bit of the unknown into the mix - a little bit of added jeopardy if you like. The game Cireon referenced was created by slackbatter and I joined it because it had a smiley face next to it. I didn't consider who else would join. I'm just upset to learn that one of the few times I beat slackbatter the points I gained from him were limited to 50. Unfair! Perhaps there is something in what Bishop is saying...
Let’s play Twister, let’s play risk
Cireon wrote:
dough_boy
It has always been a % of your rank correct?

The calculation for losing points is (your rating / average rating) * 20, where average rating is the average rating of all players in the game. The lost rating is then evenly distributed among all winners.
“This is how humans are: We question all our beliefs, except for the ones that we really believe in, and those we never think to question.”
- Speaker for the Dead, O.S. Card
Cireon is online.
The_Bishop wrote:
Ehm, sorry Cireon it looks like I have done some wrong considerations using a wrong formula.
It doesn't change much in my single example but it can change in the whole, 5 digits doesn't seem anything possible. My apologize, I hope you'll forgive my sarcasm and we'll pass over.
Still I cannot think at this rule as anything fair, because it seems to favorize just 6 or 7 for the detriment of all the the others.

I saw this game with Maafi Slack and a bunch of beginners you mentioned, and it is not the only! I think Slack actually "owes" at least 40 to maafi for the unfairness of the rule. I didn't say beginners only are harmed by the new rule, just mostly. I am also discuraged to play Slack or current Dominator Alex bcoz of the new rule, if many beginners are in the game they'll benefit the new rule discount.

Anyway i think i'm going to do a lot of reconsiderations in the next days... A small mistake can do a huge difference, but for now I'm still opposed.
«God doesn't play dice with the World» ~ Albert Einstein
maafi wrote:
My comment about the 50 points being unfair was a joke. Just in case it’s misinterpreted.
Let’s play Twister, let’s play risk
The_Bishop wrote:
@maafi, I don't think you should consider it as joke, it's just unfair for real. Actually slackbatter owes you few points. Probably you don't care about it, neither me, neither slack. But we all have to care to the system being Fair, and this new rule doesn't make any improvement in that direction, mostly deprovement.

I wanted also to quote this, which i consider very important.
dough_boy
Personally I play to have fun.

and this also:
slackbatter
The best way to get more points is to have fun, and the best way to have fun is to win!
«God doesn't play dice with the World» ~ Albert Einstein