Improving the Rating System

[cherieontop]

I would like to suggest some changes to the rating systemin regards to 1) the impact of winning Rock, Paper, Scissors on the chance of winning 2) Matches rating compared to Singles 3) the formula itself which doesnt factor in the opponents rating.

Ive collected some personal statistics for reference. Feel free to skip this and go straight to "Conclusion"

Last 100 Singles (603 to 1257 rating):
RPS won: 50
RPS lost: 50
Won Game: 77
Lost Game: 23
W/L ratio: 3.348 / 1
GOING 1st:
Won Game: 56
Lost Game: 4
W/L ratio: 14 / 1
GOING 2nd:
Won Game: 21
Lost Game: 19
W/L ratio: 1.105 / 1
RPS Won:
Won Game: 46
Lost Game: 4
W/L ratio: 11.5 / 1
RPS Lost:
Won Game: 31
Lost Game: 19
W/L ratio: 1.632 / 1

Conclusion:
⦁ the win ratio after winning RPS is 3.43 times HIGHER than the average win ratio.
⦁ the win ratio afer losing RPS is 2.05 times LOWER than the average win ratio.
⦁ the win ratio after winning RPS is 7.03 times HIGHER than the win ratio after losing RPS.
The rating system currently treats each result as if it was decided on an even playingfield. to offset the luck factor (of RPS) players would be required to play an unrealistic amount of games. thats why the rating distribution needs to be adjusted accordingly.

This is an example how the rating system would look for Singles while factoring in my stats:
win after losing RPS = 205% of current rating increase
loss after losing RPS = 49% of current rating decrease
win after winning RPS = 29% of current rating increase
loss after winning RPS = 343% of current rating decrease
im aware that this change could be exploited by players building decks that have a lower win ratio in favor of being less dependant on RPS. i suggest to change the rating distribution by a lower factor than what my stats suggest, but keeping the proportions similar:

⦁ win after losing RPS = 153% of current rating increase
⦁ loss after losing RPS = 65% of current rating decrease
⦁ win after winning RPS = 45% of current rating increase
⦁ loss after winning RPS = 222% of current rating decrease

Matches should be calculated by adding the separate results of Game 1, 2 & 3 (if there is one). The new system would be used for Game 1. Since siding eliminates the RPS factor, Game 2 & 3 would be calculated using the current system. Using my stats, an ideal distribution would look like this:
⦁ 2-0 lost RPS = 253% of current (Singles) rating increase
⦁ 2-1 lost RPS, W G1, L G2, W G3 = 253% increase
⦁ 2-1 lost RPS, L G1, W G2, W G3 = 200% increase
⦁ 2-0 won RPS = 200% increase
⦁ 2-1 won RPS, L G1, W G2, W G3 = 200% increase
⦁ 2-1 won RPS, W G1, L G2, W G3 = 145% increase
⦁ 1-2 lost RPS, L G1, W G2, L G3 = 165% decrease
⦁ 1-2 lost RPS, W G1, L G2, L G3 = 200% decrease
⦁ 0-2 lost RPS = 200% decrease
⦁ 1-2 won RPS, W G1, L G2, L G3 = 200% decrease
⦁ 1-2 won RPS, L G1, W G2, L G3 = 322% decrease
⦁ 0-2 won RPS = 322% decrease

Note1: Using this method, the k-factor in Matches is roughly double compared to Singles in order to reward players proportionaly to the time invested and fix the inequality between Single & Match ratings.

Note2: in case of a 2-1 or 1-2, the 1 loss/win is not considered as that would result in some cases that the winner actually loses rating, and vice versa. after all, its the goal to win the match, not every single game. So, 2-1 and 2-0 should be treated equally (besides the RPS factor ofc).

The final formula would look like this: K*(1-(1/(1+10^((OR-PR)/1500))))*RPS-factor
- PR = players rating
- OP = opponents rating
- For the K value (max possible rating change) i recommend around 25 to 30 for singles.
- The RPS factor for matches is already = 2 Singles, otherwise K would need to be adjusted for matches

[jonjones]

I made a thread on this earlier in February. The rating system is seriously messed up on DB especially for matches. It gives too few points for a match win and yet when you lose especially with high rating you lose a ton of points. I don't think the DB staff has made any changes to the rating system. I see the ratings for both match and single are still low and similar to what they were in previous formats. I wish there was a way for someone to contact BLS, the guy who created DN, and ask him what he did to come up with the rating system on DN.

[jonjones]

I also don't agree with the premise of the thread in that rating system needs to take in account rock paper scissors. It definitely needs to be changed for sure, to be similar to DN's where a match win gives way more points than it currently does now. On DN you would gain 30+ points regularly if you 2-0ed someone with a lot higher rating than you. Or 22+ points if you 2-0ed someone with similar rating than you. That doesn't happen on DB. I was the number one rank last format on DB in singles and the rock paper scissor wasn't that big of a deal for me. I simply used the best deck by far (spyrals) and built the deck to go second with 3 evenly matched and 7 hand traps so even if I lost rps it didn't matter for me if my opponent picked to go 1st because I always wanted to go second. Every single duel I opened evenly I won. A lot of my opponents chose to go first as well. Spyrals was so strong and consistent that even if my opponent chose for me to go first that was fine for me provided I didn't brick (which was very rarely for me). I think your deck is not built to go second. That and it's not that good of a deck. What deck are you using in singles btw?

The rating system on DB is more favorable to singles than matches and I don't like that. On DN it was the exact opposite which is why the match ratings on DN were always higher than the single ratings. On DB the top ratings in singles are higher than the top ratings in matches and they need to change that. Create a system that gives match players more points for a win and you will see a definite increase in the match ratings. It would also imo generate more interest in players to play matches since there is more incentive or reward for playing a match (more points for a win). It would get me back in the game as well.

[cherieontop]

I made a thread on this earlier in February. The rating system is seriously messed up on DB especially for matches. It gives too few points for a match win and yet when you lose especially with high rating you lose a ton of points. I don't think the DB staff has made any changes to the rating system. I see the ratings for both match and single are still low and similar to what they were in previous formats. I wish there was a way for someone to contact BLS, the guy who created DN, and ask him what he did to come up with the rating system on DN.

The match problem is also my concern, i intended to fix this by giving the equivalent of 2 Singles victories for a match. ive noticed that on DB the rating of the opponent plays a lower factor, meaning a win vs the #1 ranked player will not give much more points than vs a 100/0 beginner. a loss at the beginning of the format resulted in a very low decrease no matter the opponents rating. the higher you get, the more a loss is punished even if its vs a much higher opponent which makes no sense.
https://i.imgur.com/eQwOoAB.png
https://i.imgur.com/vE3I3X0.png
so i agree that the DB system needs work too (besides the RPS factor).
I know everything about the DN rating system, it wasnt perfect either. It used a modified ELO system and it rewarded a 2-0 win more than 2-1 (which was bad), it also factored in Wins & Losses of the opponent which was unnecessary and caused inconsistencies. The only argument i see for copying DN is for comparison of rating records. Ideally, just use a ELO system that only looks at your rating vs the opponents rating. Then factor in RPS on top of it. like so:

25*(1-(1/(1+10^(( +/-Rating Difference)/1500))))*RPS-factor

[cherieontop]

I also don't agree with the premise of the thread in that rating system needs to take in account rock paper scissors. It definitely needs to be changed for sure, to be similar to DN's where a match win gives way more points than it currently does now. On DN you would gain 30+ points regularly if you 2-0ed someone with a lot higher rating than you. Or 22+ points if you 2-0ed someone with similar rating than you. That doesn't happen on DB. I was the number one rank last format on DB in singles and the rock paper scissor wasn't that big of a deal for me. I simply used the best deck by far (spyrals) and built the deck to go second with 3 evenly matched and 7 hand traps so even if I lost rps it didn't matter for me if my opponent picked to go 1st because I always wanted to go second. Every single duel I opened evenly I won. A lot of my opponents chose to go first as well. Spyrals was so strong and consistent that even if my opponent chose for me to go first that was fine for me provided I didn't brick (which was very rarely for me). I think your deck is not built to go second. That and it's not that good of a deck. What deck are you using in singles btw?

I think your case was very specific where the best deck is a going second deck. But normally this is not the case. Of course decks adapt to Singles format to have a better Game 1, but NOT a more consistent Game 1. Players will play the deck that gives them the highest overall win ratio, whereas the rating system is intended to reflect the win chances of each game individually. Keep in mind that my stats just serve as an example. DB staff has access to all the data needed to figure out an accurate factor for everybody.

[Smitcholin]

1) You don't need to adjust the amount of rating won or lost based on who won RPS because each player is just as likely to win RPS as the other so there is no fundamental flaw putting one player at a greater advantage= than the other.
2) GLEKO rating system should be implemented. Gleko rating system is the system used in Chess, go, TF2 and several other competitive games. Gleko rating system is the most accurate way to measure a persons skill.

[cherieontop]

1) You don't need to adjust the amount of rating won or lost based on who won RPS because each player is just as likely to win RPS as the other so there is no fundamental flaw putting one player at a greater advantage= than the other.
2) GLEKO rating system should be implemented. Gleko rating system is the system used in Chess, go, TF2 and several other competitive games. Gleko rating system is the most accurate way to measure a persons skill.

I kind of agree but while its not absolutely necessary it would make for a more accurate representation by minimizing the luck factor. the goal of any rating system is to represent each individual match, not the average.

However i no longer think this is a priority. the main issues are that 1) matches reward is too small compared to singles which has turned off a lot of players and 2) the formula itself which barely factors in the level of opposition at all.

*Glicko uses a very complicated formula, im not sure how to apply this to DB to be honest. so the idea is for a players activity to affect the amount the rating changes after a game. active player = low rating change. I can already see some problems with this method and not really any benefit. What did you have in mind? Do you think Elo wouldnt be a better alternative for DB?

[XtevensChannel]

We've tweaked how points are calculated a bit. Let's see if you guys notice a difference

[Lost in Illusions]

losing 3 games = -100 rating
winning 3 games = +33 rating

unfair

[cherieontop]

We've tweaked how points are calculated a bit. Let's see if you guys notice a difference

Great! i'll put it to the test.

[cherieontop]

We've tweaked how points are calculated a bit. Let's see if you guys notice a difference

I appreciate the effort but sorry i dont notice a difference. Literally none of the problems got adressed... how is it possible to gain less against an opponent with much higher rating, more wins, more experience and better win ratio? (in both cases the RPS result was the same.) Please use a strict ELO formula the like the one i suggested because the current system is unfair.
https://i.imgur.com/CiRUKF3.png (single)
https://i.imgur.com/7C5UqmB.png (match)

[jonjones]

I still don't see much change either. Why not make BIG CHANGES to the Duelingbook rating system to make it similar to DN's? I know you guys can do it. What if you make a poll on here and ask people if they would like the rating system to be similar to DN's? In this link I played a few matches under the new rating system and the last 3 matches I won by 2-0 over people with similar rating to me yet I only gain 17 and 16 points. On DN I would have gained 25+ points depending on the W-L ratios of both players. If you can come up with a rating system that awards or deducts points based on just the ratings of both players then that would be good. Don't factor W-L ratio into the outcome of rating gained or lost because unlike DN the W-L ratio on DB is much better with lots of players having twice or three times as many wins as losses.

You can do something like a 2-0 win gives 30 points and the amount gained depends on the ratings of both players. So if a player with 1000 rating 2-0 a player with 1500 rating then that player gains 33 points because he 2-0ed a player with 500 more rating. And if a player with 1500 rating 2-0 a player with 1000 rating then that player gains 27 points because he 2-0ed a player with 500 less rating. And you can set max limit gained from a 2-0 at 35 points so even if a 100 rated player 2-0s a player with 1600 rating he gains 35 points because that's max limit. And you can set a 2-1 win at 20 points and the amount gained depends on the ratings of both duelists. So if a player with 1000 rating win 2-1 over a player with 1500 rating then he gains 23 points because he won 2-1 over someone with higher rating. And if a player with 1500 rating win 2-1 over a player with 1000 rating then that player gains 17 points because he won 2-1 over someone with less rating. And on the flipside if you lose a match 2-0 then you lose 27 points and the amount you lose depends on the ratings of both players. So 1500 2-0ed by 1000 the guy with 1500 loses 30 points. And 1000 2-0ed by 1500 the guy with 1000 loses 24 points. And for 2-1 losses you can set it to 17 points with same formula for 2-0 losses as in the player loses more points if he lost 2-1 to someone with less rating and loses fewer points if he lost 2-1 to someone with higher rating. This type of rating system would drastically increase the ratings on DB which is a good thing, reward more points to players for winning a match and punish not as much for losing. The amount gained or lost depends on the ratings of both players. A player who beats a player with much higher rating say 700-800 more points would gain a lot more points then a player who beats someone with much less rating.

https://imgur.com/AeITUNr

Basically to sum up: 2-0 win standard set at 30 points gained, highest max gained 35, lowest max gained 25
2-1 win standard set at 20 points gained, highest max gained 25, lowest max gained 15
2-0 loss standard set at 27 points loss, highest max lost 32, lowest max lost 22
2-1 loss standard set at 17 points loss, highest max lost 22, lowest max lost 12

Can you guys implement the above system? You could do that and see what happens. You can do that or some type of variation of that that gives or takes points based on the outcome of match (if 2-0 or 2-1) and the rating difference between players. If all goes well then great and if it's not good then you can just revert to the current system you have now. It doesn't hurt to at least try. I see no downsides whatsoever to having a completely different rating system on here that greatly increases the match ratings of all players.

[KRALJ UNlVERZUMA]

Ugh guys much you wrong,there should be example: Win game single/match- 10-15 Lost game singles/match -15 -20 most, -33 is so unfair and really hard to back at

[Yang Xiao Long]

its 4 wins per 1 loss to gain net rating. seems fine to me. it gets annoying when you get to a really high rating and it becomes win 1 get 7 though

[jonjones]

Ugh guys much you wrong,there should be example: Win game single/match- 10-15 Lost game singles/match -15 -20 most, -33 is so unfair and really hard to back at

A match win should award a lot more points than a single win because it's a best of 3 duels. It takes up way more time to play a match than it does a single so the player who wins a match should be rewarded greater with more points than a player who wins a single. On Duelingbook the rating system is the same for matches and singles which is why the ratings are similar for matches and singles. On DN the match ratings were always higher than the single ratings because DN's rating system rewarded match win with more points than single win.

https://prnt.sc/aqr7t4
https://prnt.sc/aqr2fv

Here are links that show the top 12 ratings in match and single from DN. For match you see 3 players with 1900+ rating and the rest with 1870+. For single you see 2 players with 1600+ rating and the rest with 1550+ rating. Notice how all the ratings are close together from 1 to 12.

http://prntscr.com/jk4xow
http://prntscr.com/jk4z0z

Here are links that show the top 12 ratings in match and single from DB. For match you see 1 player with 1800, 1 player with 1700, 3 players with 1600, and the rest with 1540+. For single you see 2 players with 1900 rating, 3 players with 1700, 1 1600, 3 1500, and 3 1400. Notice how the ratings are all over the place especially in singles. This is no coincidence. This is a direct result of DB's rating system which makes it harder for players to increase their ratings because when you have a high rating on DB you have to win a lot more duels than you lose in order to keep increasing rating. Db's system punishes players more heavily than DN for a single match loss and single duel loss because of how little it rewards you for a match win. And this is really bad especially if you D/C a lot since you're losing a ton of points than you work hard to get especially in match.

I really think you guys should make a poll and see if more people want DB's rating system to be like DN's where it is a lot easier to increase your rating. I see no downsides to this whatsoever. All players from the most highly skilled veterans to the newbies who just started playing yugioh will see a much bigger increase to their match ratings if you make the changes. I think this would encourage more players to play match rated and make DB more fun and competitive. No more frustration from winning a hard fought match that took an hour and 30 minutes only to gain 7 or 8 points and no more frustration from losing a match via D/C and having to win 4 straight matches to get the points back.

[KRALJ UNlVERZUMA]

This isnt DN its DB i dont see why we should copy things from DN on DB just change ot a bit

[cherieontop]

wow i had already given up on this thread. i finally checked back on it only to see a whole lot of nonsense suggestions by people who dont know what theyre talking ._. the point was to reward the quality of opponents instead of the set amount of rating change based on expected overall win ratio that is currently in place. this system doesnt promote competition. anyways DB rated is doomed. hopefully the DB tournament scene will be better