Ratings - Not Rewarding Good Play

[meganium45]

Steve, that is confusing.

Your kid lost twice, the other kid lost once.

How is the ratings point issue even confusing here.

What the ELO system does is reward you for beating better competition (Ratings points wise) and not give you bonuses for winning competitions against lesser competition (Ratings Points wise)

Was winning a small Regional even close to winning one of the 4 "Big" Regionals points wise?

Probably not.

Should it have been?

It is a catch-22. Do you want to go 7-1 against weak competition, winning the event, but only gaining 4-6 points per win and losing 34 for the loss, or play against elite competition, where you go 7-2, lose in the top 8, but gain 20 points per win and 20 points per loss???

Probably not.

Vince

 

[Robbgobb]

Mmmm, now time for a far-reaching view on the topic of the system not valuing player skill. However, to really get a good feel for the issue, let's take a look at this season's K-Values, which I feel are critical to fairness:

Battle Road Fall: 8K
City Championships: 16K
State Championships: 32K
Regional Championships: 40K
Battle Road Spring: 8K
Nationals: 44K

What I think is good about these k-values

-Progression. You have smaller events being worth less, and larger events being worth more. BR's and Cities are clearly not enough to earn an invite, which is good.

-Consideration of "trapped" players who cannot attend nationals...Hence Regionals being worth only 4K less than "the big show."

What I think could be improved about these K-values

-Further recognition of event similarities/differences. While Battle Roads were nerfed with widespread approval, POP should still recognize that these events are not significantly different from city championships...Or at least not 8K's difference. Now...I don't know about you guys, but when Ludicolo wins four battle roads, four cities, and T4's states/regionals, I think he deserves worlds; however, because of the Battle Road nerfing, he may not get that, even though he has displayed a long-lasting ability to succeed.

-A more even spread of K-values. Although POP owes us no explanation for its K-values unless it impacts our contribution to the brand, certain elements of the structure seem random, such as the 16K jump between Cities and States.

Tentative Proposal

Battle Roads: 12K
City Championships: 16K
State Championships: 24K
Regional Championships: 32K
National Championships: 40K

Note: I do NOT claim to be an expert of the ELO system, or even of general fairness in the system (after all, Battle Roads are worth half of a state)...However, some clear-cut formulation of my message is essential to going anywhere with it. Take this with a grain of salt!



P.S. Let's bounce some ideas back and forth with this K-Value thing, but I have a VERY important issue to emphasize about the ELO system later: the resistance-fueled phenomenon of "5th is better than 4th." Just keep that concern in mind, because POP _has_ to implement some solution to the ranking system's biggest failure.

There were kids in 10- already up in the 1800s before States. That means that with Spring BRs that they would have a chance at getting an invitation. That is not a good measure of the rankings. That is why the whole rankings stink IMO. This is a bad system and something new needs to be tried if the people in charge have their bosses telling them a system is needed. I know people who use to play that I never see any more from all the age groups. We have some new faces but I know from talking to some of those that they just don't like POP play any more.

Worlds is a big prize. People have not liked my suggestion of making each level of Premier Tournaments be worth so many points for 1st and trickles down for people who made the cut. I would like to see that system and see what happens. If we stick with this system though I would prefer to see just Tier 2 tournaments count towards ones rankings. Luck is going to be around no matter what. I just am tired of seeing people not having fun because having a bad day really is more than just a bad day with the way things are now.

 

[Scizor]

If I had the opportunity to go to as many Cities as I bet a majority of the T25 went to, I'd definitely be way in the T25. CCs being 16K as Kettler pointed out gave a lot of people high ratings. I started off 1600 at one and gained 81 points for 6-1, so I should know how much they meant. My 21-4 record at States and Regionals is probably among the best in North America, but it doesn't show because of the lack of BRs and CCs I attended.

 

[Squirtle]

90 804000017 1 1851.82 1553.16 You 6.08 1857.90
91 804000017 2 1857.90 1637.62 You 8.78 1866.68
92 804000017 3 1866.68 1797.45 Opponent -23.93 1842.75
93 804000017 4 1842.75 1566.74 You 6.78 1849.53
94 804000017 5 1849.53 1696.24 You 11.71 1861.24
95 804000017 6 1861.24 1668.91 Opponent -30.06 1831.18
96 804000017 7 1831.18 1677.63 You 11.69 1842.87
97 804000017 8 1842.87 1721.13 You 13.27 1856.14
98 804000017 9 1856.14 1902.09 You 22.63 1878.77
99 804000017 10 1878.77 1837.77 Opponent -22.35 1856.42

It feels like i'm being punished for making top cut here. As you can see the round before I got knocked out, I was fortunate to play the person who was undefeated and got a decent win. Had I played someone else I would have lost around 14 points on the day.


While the system is alot better than last year, there are still room for improvements.

 

[SteveP]

Vince, the 6-1 player "over-achieved" according to the ELO system, thus the high points.

Determining who's better and who should win is often very subjective. Now, add the element of randomness pervasion in Pokemon, and you've got a rating system that is only slightly accurate, at best.

And finally, even at the Regionals level, premier points mean very little outside of the region. If you have only 1 top player in your region, that player will be ranked higher globally than players who have multiple top players their region.

I think the author of this topic is just trying to point out something I mentioned a while back in a different topic. Playoff perfmance is not adequately recognized in the rating system. I couldn't agree more.

 

[meganium45]

Without your top cut you would have lost points on the day!

Your points were gained in your REAL top cut matches.

If you had lost T16...then you would have really been pounded, as you would expect being decently ranked with a 5-3 record.

This system is working.

Vince

No way Steve.

How do you explain the MS Valley having 3 players in Seniors who ALL Played at the Regionals, and are now ranked #6,7 and 8 in NA after the event?

Easy, they beat the players they were expected to beat, and lost to, well, each other.

The risk if you have one shooting star player is heavy is the head that wears the crown. The points for winning become lower, and the points for losing become higher.

The best areas have 2 players that rocket up together, and a consistant group behind them.

The better your competition, the better your ranking will end up, if you win most all of the matches you are expected to win...

But if you get hit with the "Grinch" factor, a newer, low rated player who gets it during a series of events and takes games off of high ranked players, then your high ranked players are going to tumble.

Masters, well, those are a mess because it really is an any given Sunday mentality.

Steve, sure playoff performance is recoginized. You are naturally playing higher rated players, so there are more points available for every single game. The rewards are increased, but the risk is increased as well.

Vince

 

[pichu bros. rox]

It feels like i'm being punished for making top cut here. As you can see the round before I got knocked out, I was fortunate to play the person who was undefeated and got a decent win. Had I played someone else I would have lost around 14 points on the day.


While the system is alot better than last year, there are still room for improvements.

just wondering, what was your explanation for losing to a guy at 1668 and losing 30 points?

 

[Aggy&Co.]

the rating system stinks and we all know it. too bad we can't change it.

 

[rerisenphoenix]

As Kettler pointed out, the K-Values should be adjusted. This is the only problem I really see with it. This is what I'd recommend:
BR's---12 K
Cities---16 K
States---32 K
Regionals---36 K
Nationals---48 K

I agree with his decisions for BR's and Cities. However, I believe States should be a little higher than only double a BR, and Regionals at least double a Cities. My reasoning for weighing Nationals so heavily is just that...IT'S NATIONALS! It is the Pinnacle of tournaments (with exception to Worlds, of course), and, as such, deserves a much higher weighting!

 

[Squirtle]

just wondering, what was your explanation for losing to a guy at 1668 and losing 30 points?

Opening hand:

3 Banette, 2 Gyarados,Night Maitenance,Shuppet.



I was still going to win but lost by 1 prize on time.





Vince, my grapes aren't that sour. Your taking it the wrong way. With my current rating, I was expected to win the event and didn't - That's why I only got 4 points on the day. What i'm saying is that i'd like to see some bonus points or something for just making top cut.


As you can see I did lose 30 points in one game. The system doesn't account for bad draws or bad luck.. Or simply losing on time when you have clear board control

 

[Prime]

Even if the k values are adjusted, it won't fix the crucial problem in the system.

The system doesn't account for bad draws or bad luck.. Or simply losing on time when you have clear board control

This isn't chess. This game has luck, and has less restrictions on time (which leads to more problems). I don't see how one system can work for two totally different games.

I agree that you shouldn't gain that many points for playing someone well below your rating, but there has to be a cutoff or something so people don't end up earning 15 points through wins and losing 60 points through losses.

 

[Aggy&Co.]

the whole idea that a players ratings adjust based on the win percentage of your opponent is ludiculous. the end result should be affected by what a player won/lost and nothing else. each placing should be given a point value and it shrinks the farther you go down the pool.

By the same token losses should be taken off the top, less points removed for the first round loss and it grows for each subsequent round you play in.

 

[Pablo]

My win/loss record is in my signature, it's below 80%: 77.9%, but I'm in the approximate range and I did 15-1 two States which is no easy feat. I'd like to think I am deserving of the place I currently hold.

 

[bulbasnore]

I feel like this system is one that is used because the old way was one that the people we know that is in charge of the game was told by their bosses to change. They then picked something that would not take too much work and shoved it on us.

Don't "trust your feelings, Luke" on this one. Room for improvement. I think the leadership will provide any they can make. In many ways the OP folks in PUI seem to have much more control over OP than the old MTs did under WotC, so I don't see imposition from above, I certainly wouldn't call them lazy (not within swinging distance) and I think their delight is to show the players a good time & reward achievement, not to shove things on them (except the occasional silly hat).

 

[NoPoke]

I have a Luck fix. its a Bayesian modification to the existing elo calculation of stake.

 

[Lawman]

just wondering, what was your explanation for losing to a guy at 1668 and losing 30 points?

A ''grinch" who hadnt played all season long. A good player who got T16 @ Nats last year.....you may have heard of one of his decks....Mario. Yes, it was KingGengar and he finished 2nd at this regional w/o playing a single tourney all season long. A 1600 player taking big bites outta people's points.

Keith

 

[meganium45]

Keith, Nationals will be full of Grinches. (Low ranked players who are actually good)

I Know of a Senior who will be coming in with a 1600 record, who went to Worlds last year...

He is going to take a bite out of some players.

There are several other Masters Players, myself included, who are just above 1600, and are going to hit people for some points. (Jimmy Ballard, among others)

We get to play, it is the nature of the beast.

Vince

 

[Cheenou]

hey where did you see your ratings at?.. for me at go-pokemon.com the ratings arent up for regionals yet

 

[NoPoke]

The only fix for the grinch effect is provisional ratings. eg http://www.daysofwonder.com/en/play/ranking5/ where the points exchange is scaled by number of games played upto 20 at which point you are on the established rating. So a grinch entering Nationals has their K value reduced in the folowing way

R1: K/20 = 2.2
R2: K/10 = 4.4
R3: K3/20 = 6.6
R4: K/5 = 8.8
R5: K/4 = 11

etc... Grinches with provisional ratings dont crucify you.

The Grinches themselves do okay, assume the opponents the grinches meet are all 1800 and on the establish ranking.

Grinch: 1600
R1: Grinch gains 200 points = 1800!
R2: Grinch gains 66.7points = 1867
R3: grinch = 1900
R4: Grinch = 1920
R5: grinch = 1933

Note how the Grinch rises rapidly but as the rating is provisional it wouldn't qualify for an invite and also does little damage to the opponents.

 

[ryanvergel]

I'm a master in the 1600s. Very low 1600s. But that's what happens when you go 5-3(5-2 swiss losing t2 to banette, losing t5 in mirror , then losing in top8 to an autoloss while never seeing a supporter or legitimate draw) at states, only go to 2 BRs 0 cities, and then bomb at regionals (4-3).

I'm going to grinch a few players at nationals. I can't get too many more supporterless games lol.