A New Ranked System

Please pardon the extensive intro and overall lengthy post. Any rating system implemented takes a lot of research, development, and resources to implement. It is difficult to analyze such a complex system in only a few words. Nevertheless, I will include a TL;DR at the end of the post.

I first played League of Legends before most players did, at the beginning of S1. At this time I played less than a few games a week, due to not really understanding the MOBA genre and therefore not being able to enjoy the game much. Around the first League of Legends Championship (a Dreamhack event), I found out that a lot of guys at my highschool played it. This was when I started playing extensively and my skill level skyrocketed. By the end of S2, when I began to take a break from League, I was averaging 1600 Elo. I played for weeks at a time from that point on, taking months away from League of Legends in between each period, and forgetting my e-mails and log-in info several times. Around S5, my younger brother started to pick up League, with influence from his friends. As a former League of Legends player, I was intrigued whenever he played. While I did not have the time to play myself, I was sometimes a backseat-driver for my brother. During this period, I noticed something seriously disturbing. The new system not only brought new flaws, but did not fix any flaws of the old system.

Issues With The Current Ranked System

Inflation This is the first issue that came to my realization whenever my brother spoke about his friends and their performance. These are friends that I knew from when I did play seriously, and that I still watch over my brother’s shoulder. The realization of this issue further strengthened whenever I briefly spoke about League of Legends with my former highschool classmates, who I played dozens of games with weekly during highschool. Friends who I noticed had not improved much since I quit were now diamond, top 10 percent. They are still playing at a level where, in the old system, they would either be bronze or unranked. To put things in perspective, my Elo garnered me a gold medal, and to get a medal at all, you had to be in the top quarter or so in ranked. While I realize the number of players has now largely increased, and the addition of new tiers shifts the old tiers down, it should not cause the player distribution to shift upwards. Players who were below the top 25 percent should not now be in the top 10 percent.

For those who understand statistics or want the actual data that backs this up, take a look at this graphical analysis:

http://forums.na.leagueoflegends.com/board/attachment.php?attachmentid=882898

Any system that measures performance in an area where people vary in skill level should follow a bell curve. This means the lowest possible distribution has as few people as the highest possible distribution, and the most populous distribution is the average. If either end of the bell is chopped off, whatever was chopped off must be redistributed across the rest of the bell. This also means that if the bell ends abruptly on the high side, there is an artificial performance ceiling that is lower than the actual performance ceiling. Likewise, a bell that ends abruptly on the low side, like this one, has an artificial performance floor that is higher than the actual performance floor. The issues with inflation in a ladder system is that it amplifies discrepancies player placements. The system took the “chopped” tier players and tried to shove them into the rest of the bell, causing a ripple effect that shoves some bronze players into silver, silvers into gold, and so on. The effect is demonstrated by the low magnitude of skill difference between bronze and silver, the tiers closest to the source of the ripple. This goes against the whole concept of a bell curve, where the most drastic change in distribution happens where the bulk of the bell is.

Unrewarding For Many

I am someone who, like a lot of people, do not have much time to play League of Legends. While I happen to have free time right now, normally I can play only a few games a week. However, I believe I have every right to enjoy the game as fully as any other player. More importantly, I believe that it is Riot’s goal to cater to more people than just serious or adolescent gamers. A large problem with the League of Legends ranked systems, old and new, is that they require a very large number of games to be played, to accurately reflect a player’s skill level. This is due to the fact that the game has a very large number of variables. These do not only include rare extremities like trolls, leavers, and smurfs, it also includes common events like being grouped with a player who is simply on their way up or down a tier. However, this effect, the law of large numbers, does not only apply to League of Legends. It applies to any system that has chance variables which you cannot control. Take a rolling a die for example, which obviously has much fewer variables than a game of League. If you take the cumulative average of your rolls, much like your hidden matchmaking rating, it should always be 3.5. However, you will likely find that in the first few rolls, your cumulative average is not within decimals of 3.5, and that it the average will not stabilize on 3.5 until after a couple hundred rolls.

Surprisingly, after all the research Riot games put into their ranked systems, they still ignore this law. I would also like to emphasize that this applies both ways, to climbers and fallers.

Long Term Placement, Short Term Evaluation

In a ladder where your placement reflects your skill level, which is something that does not simply spike after a month, it is counter intuitive to expect a player to win almost consecutively in order to rise. This again, makes it unnecessarily difficult only for players who cannot play multiple games in one sitting, or even in the span of a couple days. In addition to the exaggerated obstacle, in the very probable event of a bad series, the player must prove themselves time and time again against the same level of opponents. While a player should be able to advance a tier eventually if they truly deserve it, a ladder system should not have such an ambiguous process.

Proposal For A New System

While my ideas are not remotely refined, I believe this proposal can be a start to developing a system that fixes very obvious flaws.

Points System

This would be the basis of the system. Your points are an explicit measurement of the value of your wins or losses throughout the season. They are persistent from the start to the end of the season. They do not reset when you enter a new tier. Once you receive your starting points for the season, your opponents are determined by your number of points, using an algorithm. Your MMR at the end of the season is determined by the number of points you have.

This gives players the transparency they deserve and leaves less room for any discrepancy between matchmaking and your “skill level”.

Initiation, Re-Adjustment, and System Re-Adjustment

Players playing ranked for the first time in their League of Legends career will play around a dozen placement games. At the end of these placement games, they receive however many points the average player in their tier and division has. Their tier and division is determined by their MMR at the end of the placement games. For the most part, this works just like the current system. At the start of a new season, players who have already completed placement games will play only a few placement games. Their starting MMR is based on their finishing MMR the previous season, and their division after the placement games is determined by their MMR relative to the average MMR’s of each division in the previous season. The number of points they start the season with depends on their division. The amount of points allocated to each division is determined by Riot.

This allows Riot to experiment, calibrate, and fine-tune the system after every season to, for example, restore the bell curve, accommodate for influxes of players, or find the sweet spot for the average number of points.

Progression

A player will advance a tier or division when the number of points they have surpasses the minimum requirement to stay in the next division, beyond a certain threshold. Players who drop below the minimum requirement for a division, beyond a certain threshold will fall to the lower division.

The threshold reflects the fact that players may have fluctuations in the outcome of their games. This also allows a true climber or faller, who would be steadily approaching the mean points of their real division, to climb or fall unrestricted.

Bonus Pool

Players will begin their season with the same number of bonus points. The number of starting bonus points will decay over the course of the season. When losing, points are first deducted from the bonus pool. When winning, points are drawn from the bonus pool to multiply your gains. The starting amount of bonus points should be generous, enough to play dozens of game. The bonus pool will allow players who cannot play hundreds of games in a single season will have the same opportunity as players who do play hundreds of games.

I understand that some questions may arise regarding this concept, so I will answer the ones that I think people are guaranteed to ask, and I will continue to respond as people post their feedback.

Won’t this cause inflation?

No, because everyone receives the same number of bonus points at the beginning and throughout the season. If you gain 100 extra points, so will everyone a division below you and above you, and therefore the distribution remains intact.

Won’t this give players who join the season late an advantage?

This is first assuming that the placement system that is adopted from the current system fails. The decaying bonus pool helps to eliminate this advantage by emulating the usage of this pool for these players. Furthermore, any potential gain is minimal, as reaching higher than your true level will garner excessive losses, which will eat up a player’s pool quickly.

TL;DR: Ranked is plagued by inflation, slow adjustments, and is terrible for busy people. Implementing a points and bonus points system will fix those issues.

Thank for for reading, please leave feedback! I hope players will challenge this, ask questions, and maybe Riot could shed some light too.

P.S. For those interested, I do not personally play much ranked due to the issue of not having much time to beat the law of large numbers. I have well below 50 games played. I normally play with my younger brother, some friends, and my girlfriend for fun, and you'll see some of the same players in my match history. You also won't see the same players when I'm on a ranked win streak.