Four Months of Data: CHANGE THE RANKING SYSTEM

RCubed1967

Hi everyone,

Let’s talk about the ranking system and use my four months of my Rubedite-ranked data.

As I have stated many times, the system is broken because it is based on chess. Tales of Tribute is a game of RNG, luck and strategy. We get lucky or unlucky with RNG in numerous ways: if we get the first pick, what is on the table, natural combos, etc. Chess does not have RNG or luck: both players have the same pieces, which are positioned similarly, and there is no randomness in which pieces you have to move. Strategy exists in both games., but it is less noticeable in individual games of Tales of Tribute due to the luck and RNG. For Tales of Tribute, the only way to see strategy is in multiple games so that RNG and luck are statistically removed.

The ranking system needs to be built on the skill of the player. Skilled players have good strategies in a variety of scenarios. We must consider multiple games (a series) played against the same player to see strategies, or rather the results of good strategies. Seeing who is more skilled between two players then provides the stepping stone to a ranking system (and that ranking system gets rid of a secondary issue: exploitation by players).

Each series must be balanced for who gets the first pick due to the huge first-pick advantage (another favourite topic). Consider my July results with ~80% win rate with the first pick but only ~45% with the second pick. If a match against the same person has one person with the first pick all the time, the results will be significantly skewed (e.g., I have played @Meji 6 times in the past 4 months, and I have never gotten the first pick).

I think that topic is relatively well understood, but let's turn our heads to my four months (780 games) of data using the current ranking system. The table below shows the series I have played in Rubedite. At one end of the spectrum, I have played 115 players only once, and I won those 63-52. On the other end of the spectrum, I played one person 19 times, and they beat me 10-9.



In a properly ranked system, the total points per game should show a major score reduction per game on the left and a major score per game gain on the right. The gains per game should be zero for games where we tie (e.g., 3-3). The idea is that the higher the skill differential is between the two players, the bigger the gain should be. A maximum amount of points must be available to each series, which is split between the two players. I get all the points if I win every game in the series. If I lose every game in the series, I get zero. That said, there needs to be a minimum amount of games played between players to be scored this way (and the series must be balanced for the starting pick).

The current system shows the points per game in each series in the following table. The series size of 3 is relatively what I would like to see, but it is still unbalanced. The series with 19 points showed me gaining points, while the 17 games series where I won 11-6 had me lose net points. Those should never happen.



What is also of HUGE interest is how much the single-game series dominate the scoring, especially the losses. A single game, which is dominated by RNG and luck, should NEVER be worth more than a series.

The ranking system must change. It must be based on series, not one-off games. Each series needs a set amount of points to be split, and they must be balanced.

To support the series system, the season should be 4 months long (DLC to DLC); that way, players can get a lot of series done to be properly ranked.

I look forward to the discussion.

Necrotech_Master

this isnt the only game ive played that has a ranking system where the higher you get the less pts you gain on a win and more you lose on a loss

in fact its in a mobile card game (though not a deck builder like tribute) where a player can customize their deck beforehand

4 months would be way too long for a campaign, i mean if this was like cyro pvp nobody would wait that long (and in fact people were going ballistic when 1 campaign duration got extended an extra month because of how they were changing the campaign start times), i certainly wouldnt want to wait on that, and at best it would just give me more time to even bother trying to make it to rubedite tier once to get the achievements

i do think they should fix it so that a win gives some minimum amount of pts so that people dont get 0 pts on a win, as that doesnt seem fair to win and get nothing, but risk losing up to 150 pts on a loss

i personally have very little interest in pvp tribute, i only even bothered doing it at all in order to get card upgrades (about 13 matches in total, and probably had a win rate of 60% at the lower brackets, and yes some of the matches i lost were because i was purely trying to finish the achievements for card upgrades lol)

Personofsecrets

First player RNG also needs to be fixed. Not only does my data show that players may be getting way more 1st player than 2nd player games. It also trends that 1st and 2nd player games come in clumps.

This is the game RNG being manipulated and then the game over compensates.

spartaxoxo

Personofsecrets wrote: »

It's normal for it to be streaky rather than flip flopping constantly. That's not a sign of poor RNG or manipulation..

What is a sign is after enough games it doesn't trend towards 50% .

RCubed1967

I agree fully with @Personofsecrets that the RNG of the first pick has to be addressed.

I also agree with you, @spartaxoxo that the back and forth is normal. Further, the probability of getting at least 29 first picks out of 46 games is only 5.19%. 5% makes it significant...so @Personofsecrets data is very helpful in verifying that the RNG issue of the first pick is problematic.

My 10 in a row (<0.1%)and 11 (<0.05%) in a row are insanely out of whack. The fact they happened in the same month is about once in 875 years, given 200 games a month. Yes - 875 years...

Thank you both!!!

Amottica

RCubed1967 wrote: »

In a properly ranked system, the total points per game should show a major score reduction per game on the left and a major score per game gain on the right.

This would be the case if the two players were similarly ranked. However, in a properly ranked system, the total points lost or gained per game would vary depending on the difference in rank between the two players.

A top-ranked player who defeats a lower-level player should lose more points than if the player was similarly ranked and the total points lost would scale depending on the difference between their ranking. That lower-ranked player should also have a greater gain depending on the difference in ranking.

I point this out since players that play during peak hours are probably more likely to be paired with similarly ranked players assuming that the ranking system is used properly for pairing, and it should be. However, some players may choose to play during off-peak hours and higher-ranked players would receive to much benefit from defeating a notably lower-level player, and visa versa if they were to lose to the lower-level player.

ESO_player123

RCubed1967 wrote: »

The ranking system needs to be built on the skill of the player. Skilled players have good strategies in a variety of scenarios. We must consider multiple games (a series) played against the same player to see strategies, or rather the results of good strategies. Seeing who is more skilled between two players then provides the stepping stone to a ranking system (and that ranking system gets rid of a secondary issue: exploitation by players).

Wouldn't it require storing a lot more information? The statistics of each player playing each other player that participates in the series? It will be a lot even if the duplicate pairs AB and BA are reduced to one pair AB.

RCubed1967

@ESO_player123 - yes, more storage may be required to hold the data, but not significant. It should be just a simple database matrix storing wins, losses and start position. The logic is pretty easy to work out.

What is more troublesome is to figure out what is wrong in the programming of the coin flips. That may require a serious overhaul. I have heard rumblings that the luck RNG spills over to other things like farming drops. I would rather they just focus on ToT, but it might be game-wide modifications.

ESO_player123

RCubed1967 wrote: »

@ESO_player123 - yes, more storage may be required to hold the data, but not significant. It should be just a simple database matrix storing wins, losses and start position. The logic is pretty easy to work out.

What is more troublesome is to figure out what is wrong in the programming of the coin flips. That may require a serious overhaul. I have heard rumblings that the luck RNG spills over to other things like farming drops. I would rather they just focus on ToT, but it might be game-wide modifications.

Considering the reason for introducing AwA was cutting back on storage, what you propose might not be feasible.
Though I have to admit that I was puzzled when I occasionally was getting 0 for a win when I was playing ranked. Now, I only play at the start of the season to get the rewards for reaching rubedite rank, then I switch to casual.

maxxiestackhouse

Amottica wrote: »

RCubed1967 wrote: »

In a properly ranked system, the total points per game should show a major score reduction per game on the left and a major score per game gain on the right.

This would be the case if the two players were similarly ranked. However, in a properly ranked system, the total points lost or gained per game would vary depending on the difference in rank between the two players.

A top-ranked player who defeats a lower-level player should lose more points than if the player was similarly ranked and the total points lost would scale depending on the difference between their ranking. That lower-ranked player should also have a greater gain depending on the difference in ranking.

I point this out since players that play during peak hours are probably more likely to be paired with similarly ranked players assuming that the ranking system is used properly for pairing, and it should be. However, some players may choose to play during off-peak hours and higher-ranked players would receive to much benefit from defeating a notably lower-level player, and visa versa if they were to lose to the lower-level player.

Come live in Australia then mate. I rarely see top ranked players unless I’m on a day off so there is almost ZERO reason for me to rank to Rubidite as then I’ll always lose 150 points if RNG screws me and only every get 0-25 points for a win. I’d rather not get matched at all if there are no other Rubidite ranked players Q’d than the BS scoring system.

Seraphayel

Honestly, what the f* is wrong with the ranking system? I was at 26, then I lost three matches (two of which I went second) because the opponents either were stacking up on Crow cards or I couldn’t buy anything from the tavern - and now I’m at rank 123. Like… what is going on?

SilverBride

I still think it should be based purely on the number of wins compared to the number of losses.

If a player wins a game they move up the ladder by one rung. If they lose a game they move back down the ladder by one rung. The players winning more will move above those winning less.

The skill level of the opponent shouldn't be a factor at all, especially since there is no way for it to be accurate with such a flawed system.