Pokémon TCG: Sword and Shield—Brilliant Stars

Designing a Fair Rating System: Part 1

Eevee, I wasn't saying that if you don't travel then its because you are lazy. Or anyone else who is disadvantaged by geography.

FS gave the following as a Problem and hence in part as justification for changing/replacing the rating system

FS post 1 said:
PROBLEMS:
-People in areas with less events are at a massive disadvantage.
Click to expand...

I don't accept the particular 'problem' as justification for altering the rating system. The world is split into regions to attempt to take account of the variation across the world. Not a perfect solution but it certainly helps and addresses the massive assertion.. Here's why I dismiss 'geography': It is inherent in a rating/reward system that you have to play. So no matter what changes are made to the rating system this geography feature will remain. The only way to make geography irrelevant is to scrap the ratings invites. Even then you have to be able to attend whatever replaces the tournaments that no longer count. So I agree with the following statement: 'People in areas with fewer events are at a disadvantage' but I don't agree with identifying it as a problem that can be fixed by changes to how the rating system is operated. Fewer events means much more uncertainty in rating. uncertainty is bad. Disjoint areas cannot be strictly compared but at least if the playing populations in the disjoint areas are large and comparable in number then there is some justification in the comparison between the disjoint groups.

So definately not dissing those who are hampered by geography, or age, or money, or whatever reason that is excluding potential Pokemon stars from shining. I offer no solution because I don't think there is one. Geographical advantage and ratings advantage go hand in hand. Having parents involved as LL/TOs/PTOs also conveys an advantage. There is no fix for the advantage from parental involvement just as there is no fix for being geographically unable to attend tournaments .

Now if there are other players nearby then all you need is a TO and a local tournament becomes possible. The geography problem can be made to go away with more local players. But I reamin convinced that geography will always disadvantage a quantity of players.
 
Last edited:
Here is my idea:

When you win, you gain points based on your rating at the time. For example, if you play a game and your rating is 1710 before the game begins, you would receive points, if you win, based on a scale between 1700 and 1750. Arbitrarily, 10 points for these parameters. The lower your rating, the more points you gain for each win. So, to invent a scale, between 1500 and 1550, you might gain 40 points per win, and between 1550 and 1600, you might gain 35 points. The reasoning for the scale is that lower-rated players ought to be rewarded for their skills, while higher-rated players should be expected to win, period, regardless of their opponent. Thus, a player between 1900 and 1950 might gain 5 points per win, and a player between 1950 and 2000 might gain 3 points per win.

When you lose, you lose a set number of points, regardless of your rating. The points lost should be a low number, so that highly-rated players are not put off from playing at late-schedule events, such as Battle Roads or Nationals, in order to secure their high rating. This will improve the player pool, and will also encourage players to try new ideas which, in the current rating system, is not a sane idea for highly-rated players. The set number of lost points could be 10 points, so that a lower-rated player receives proper credit for a decent showing, and a highly-rated player is not severely punished for an off day.

Sample1 of Scale in Action:

Player is rated 1822.
Game 1 Win +15 points. New rating 1837.
Game 2 Loss -10 points. New rating 1827.
Game 3 Loss -10 points. New rating 1817.
Game 4 Win +15 points. New rating 1832.

In this example, a player who is of a pretty good rating to start is not punished for a mediocre performance, but is actually given 10 points total for his effort!

If the same player won all of his games, then after Game 2, his points gained would fall from 15 to 10, so that after 4 wins, the player would end up with a rating of 1872.

Sample2 of Scale in Action:

Player is rated 1962.
Game 1 Win +3 points. New rating 1965.
Game 2 Loss -10 points. New rating 1955.
Game 3 Loss -10 points. New rating 1945.
Game 4 Win +3 points. New rating 1948.

In this example, a player who is of an excellent rating to start does not have his rating totally shattered, but loses only 14 points.

If the same player won all of his games, the player would end up with 1974. This is not a tremendous increase, but neither is a bad showing a disaster for an player with an excellent rating.

I think this is a great and equitable way to work things out.

Thanks to Mike Reynolds, Matt Kirschner, Jordan Hill, and Matt Riddle for input.
 
Last edited:
they shoudl jsut charge 1-5$ for everyone at events and then bring back gym challenges and top 8 worlds qualifications. i hate the rating system. You have to hand you soul and socail life and everything healthy to a stupid card game. Pokemon USA promotes addicts and humble indiviaduals imo. I think this game is a great game, but it isn't like golf, basketball, or baseball. You can't make money off it. It's for fun, as much as i hate to say it.
 
KingGengar,

Your proposal scraps the zero-sum property. Zero-sum isn't a holy cow and so can be abandoned. However you need to be really carefull when designing non-zero sum rating systems because they are wide open to abuse and manipulation. Much more so than the existing elo based system.

What happens if any group of players take turns to beat each other.. they all rise up through the points under your proposal. When you design a non-zero sum system you pretty much have to start from the safeguards to protect against abuse before you get to the details of how to actually reward the players.

A little bit of non-zero sum is probably tollerable but to have the whole system non-zero sum may be impossible to achieve because of the high probability that someone somewhere will manipulate the system and the equally high probability that it will be difficult to detect.
 
NoPoke: So I guess what you're saying is that my idea is pretty good but (as usual) it's the manipulators of the system who ruin it for the rest.

I agree with your reality check.

But how do you think it rates in a moral universe?
 
Even in a moral universe POP would have to take great care to ensure that there was equality of opportunity. Much more than they have to do now under the non-linear elo system. This need for very careful balance of opportunity applies to any system that is a long way from zero-sum. Its not surprising that the Pro-tours have a fixed calendar/circuit and are frequently invite only. Formula 1 and Soccer spring to mind, but there are lots of examples.

Your proposal is non-linear which is good as that non-linearity constrains excess growth. You are addressing the potentially catastrophic loss of points that the high rated players risk. Both features are required of any open-ended tournament system where players control the amount that they play. Not using the opponents rating to decide the points gain is similar to what happens under the current system in the early season where everyone is 1600. I really don't know if it would be a good idea to ignore the opponents alleged skill for the whole season.

Here's a thought for you, instead of trying to make your proposal work over individual matches see if it can be adapted to whole tournaments. In a tournament you gain points for a win regardless of the opponents skill, much like your proposal. Losing has a fixed impact too much like your proposal. Of course tournaments are very nearly zero sum (byes upset this) so you may have to alter the players gains and losses through some renormalization process. That zero-sum property is rather helpful in keeping everyone honest. How about having the points that players gained or lost to be based purely upon their eventual win-loss record with a multiplier derived from whatever properties of the tournament that you wish to use eg. average player rating , number of players, age, number of rounds.

Don't give up on zero sum. Non-linear is good.
 
NoPoke, your points are well taken.

So, what is the solution or, more nearly, the compromise, if any?

It seems to me that no winner is decrying the number of points he or she gains under zero-sum, even if very low, knowing that such things make sense in the ratings stratosphere. However, that same system inordinately (IMO) crushes the daring player with an elite rating, so that the board leaders are reluctant to play late in the season, when ahead. Not that I fault such caution- it's prudent.

I don't care to improve the gains side of the coin, but only limit rating slippage on the loss side.
 
Solution? LOL.. err pass.

reduce the risk of a late season loss without the extreme of sitting out. Sitting out obviously reduces the risk to zero. Its not actually the risk of a late season loss that causes players to sit is more that such losses are excessively penalized in many players view. reducing the K value of the late season tournaments reduces the penalty but also the gain. The risk-reward equation remains unchanged and ought to result in exactly the same level of sitting in a completely rational world. But we aren't completely rational so I think a K reduction would actually have more players playing even though there is no logic behind the players decision. The risk can be reduced by switching to match-play for late season tournaments. Best out of three 45 minutes in the Swiss does allow recovery from dumb bad luck. Now if 45 minute rounds were practical at USA nationals then I would be advocating strongly for POP to recommend its use at all late season tournaments. *shrug* silly 700 player tournament!

De-emphasize ratings. Ratings were just for fun in the past. The USA could go back to that approach and just have tournaments giving out invites. It is probable that the USA doesn't need ratings invites at all to solve the 'problem' of how to award invites, elsewhere in the world I'd say that ratings invites remain a good idea. At least until our OP catches up with the USA.

Change the risk-reward equation. Your proposal is an example of such. I've suggested an alternative modification over on the professor forums. http://forums.pokemon-tcg.com/showthread.php?t=3160 To explicitly account for a degree of luck in the calculation of the stake. (The stake is what pop call the number of points transfered from loser to winner). It turns out that including luck as a factor in the calculation of the stake also reduces potential loss to the higher rated player but without a big impact on the gain. including luck introduces an asymmetry that moves the risk-reward balance closer to where it probably should be. The proposal I made also retains the zero-sum property.

late season loses have a much bigger impact than early season loses. I don't know if this is right or wrong. If you want the prize then you have to win on the day that counts or even the days that count. Its no good pointing to past glories.

So not so much 'Designing a fair rating system' more 'Designing a fairer rating system'
 
Last edited:
As you say, I wouldn't be bewailing the current rating system's fairness if there weren't prizes on the line. If just for fun, then the inequity may cause a little angst, but no proper concern.

It's a catch-22, or perhaps a paradox, that any hard feelings towards fairness of the system is directly tied to high anxiety caused by the prizes. However, I wouldn't want to eliminate the possibility of World's trips for that concern.

POP might say, "OK then, grin and bear it." Which we do. But hope springs eternal, and by hashing it out on the thread, perhaps we'll hit on something more mathematically-pleasing. And, who knows.... maybe POP sees one of these proposals and finds it endearing and practicable.

So onward we must go.
 
I agree this needs to be sorted.

It would also be good if every country had the same amount of invites via Nationals. Maybe top 2 as this would guarantee that every country had a reasonable amount of players at worlds.

Closed events also needs to be looked at. How can it be a fair playing field when players are excluded from events based on where they live. If they want to travel - why can they not do so!
 
I aggree that cities K value is too high but battle roads should be higher...
1.)Because by the time battle roads roll around, all the high rank players are sitting on their ratings so there is no change towards the top of the ratings, but only in the middle/bottom because the points won are so low they dont effect much.
2.)And because people know this they dont attend. so there are less points to be won
Its like if you go undefeated at battle roads you might gain 30 points... well if you are at 1700 it wont make aa difference to those at 1900+. but if you can gain 50-60 points per battle road, those players cant afford to sit and not play because you will catch up to them, and since they have to play they give up more points or gain more points which makes the other high ranked players have to continue to play as well to keep up as well. true they may win and shoot even higher in the ratings, but to do that they will have to play and risk those points. WHen the k rating was announced for battle roads 60% of the pokemon players just stayed home because it would not effect them at all i mean 4 rounds with a top 2 cut?
but thats just for battle roads.
 
MukMan: Redesigning the K-Values for Battle Roads might get some folks excited. By the same token, the Battle Roads were "easy points" for some because of the low turnout. In a way, it gave reg'lar folks a chance to catch up to the supahstars. So there's a catch: higher K-Values means more domination by better players. Do we want that or do we not?

I like your idea, but I still think that K-Values on the loss side need to be mitigated without reducing the possible gains from victories.
 
It can really ruin your day...
You think ou yes, I've won something ... and then:
1757.29 1476.31 You 5.96 1763.25

BETTER RATINGSYSTEM
 
You must log in or register to reply here.
Back
Top