Boards›Gameplay

An Experiment for the Boards related to current Ranked Matchmaking. Any are welcome to participate.

jaymc1130·7/29/2016, 8:03:22 PM·9 votes·1,278 views

So I've had a thought recently, and I've already done this personally myself, but I think it might be a good idea to explain the experiment to the boards at large and to include more data to create a complete picture of how matchmaking is functioning at the moment. Whatever your opinion on matchmaking may be, whatever your opinion of DQ may be, you can participate in this unbiased experiment to help build a dataset that will help Riot determine what issues may be present and require solving. As a preface to explaining the experiment and how you guys can help I would like to say that I did this myself for at least 20 games on 5 different accounts this season ranging from unranked to diamond and each came up with a similar result: the MMR gap between players on a team had increased significantly, in fact by slightly more than 3x (from previous seasons average MMR gap distance of slightly less than 100 MMR to slightly over 300 MMR this season) when compared to the gap in previous seasons. I attributed this as a primary cause for the increase in low quality matches that occurred for this season that make one's ranked experience less enjoyable overall. I think its time we increase that sample size from a few hundred games to several thousand now that Riot has implemented some changes to see if these have affected things for ranked play and improved the quality of matches for players in all skill ranges.

Without further ado, let me breakdown what I would like for each of you who participate to do for the sake of the experiment. This process must be documented so I will be asking for screen shots of your match histories of at least 20 games and sample screen shots of the individual match post game metrics from selected matches within that minimum 20 game sample. What you will need to do is op.gg each member of each team for each of those matches and find the widest skill gap in MMR from the lowest ranked to the highest ranked player on each team. As a sample of how to document this process I direct you to this thread here:

http://boards.na.leagueoflegends.com/en/c/gameplay-balance/upecpYHB-platinum-promos-places-me-with-three-silvers

In this match it was found that the player Miley Dyrus fan was the highest ranked of the players on his team with an MMR of around 1680 when op.gg'd. The lowest ranked member of his team had an MMR of about 1180 for a whopping 500 MMR difference! In addition to his screen shot of the post match metrics I would ask that you include screenshots of the op.gg's of the highest ranked and lowest ranked members of each team (its important you check and include both the enemy team and your own team!) for these matches (you can stick to matches that are outliers from your datamining, its not necessary to include every op.gg for everyone or for every match, a simple textual answer for the distance of highest and lowest for each team is enough for most matches with just the extremely close, or extremely distant gaps being used for screenshots. The textual answers can be easily confirmed by any player who wants to looking up the data on op.gg himself, so again, we don't need screenshots of 4 op.gg's for each match, just stick to the outliers and the overall match history sample size of 20 games.

So, what you will be doing is this:

Poster X - Sample size of 20 matches screenshot Game 1 skill gap for my team: Y MMR. Game 1 skill gap for enemy team: Z MMR. (Screen shots included) Game 2 skill gap for my team: Y MMR. Game 2 skill gap for enemy team: Z MMR. Game 3 skill gap for my team: Y MMR. Game 3 skill gap for enemy team: Z MMR. (Screen shots included) Game 4 skill gap for my team: Y MMR. Game 4 skill gap for enemy team: Z MMR. Game 5 skill gap for my team: Y MMR. Game 5 skill gap for enemy team: Z MMR. Game 6 skill gap for my team: Y MMR. Game 6 skill gap for enemy team: Z MMR. Game 7 skill gap for my team: Y MMR. Game 7 skill gap for enemy team: Z MMR. (Screen shots included) Game 8 skill gap for my team: Y MMR. Game 8 skill gap for enemy team: Z MMR. Game 9 skill gap for my team: Y MMR. Game 9 skill gap for enemy team: Z MMR. Game 10 skill gap for my team: Y MMR. Game 10 skill gap for enemy team: Z MMR. Game 11 skill gap for my team: Y MMR. Game 11 skill gap for enemy team: Z MMR. (Screen shots included) Game 12 skill gap for my team: Y MMR. Game 12 skill gap for enemy team: Z MMR. Game 13 skill gap for my team: Y MMR. Game 13 skill gap for enemy team: Z MMR. Game 14 skill gap for my team: Y MMR. Game 14 skill gap for enemy team: Z MMR. Game 15 skill gap for my team: Y MMR. Game 15 skill gap for enemy team: Z MMR. Game 16 skill gap for my team: Y MMR. Game 16 skill gap for enemy team: Z MMR. (Screen shots included) Game 17 skill gap for my team: Y MMR. Game 17 skill gap for enemy team: Z MMR. Game 18 skill gap for my team: Y MMR. Game 18 skill gap for enemy team: Z MMR. Game 19 skill gap for my team: Y MMR. Game 19 skill gap for enemy team: Z MMR. (Screen shots included) Game 20 skill gap for my team: Y MMR. Game 20 skill gap for enemy team: Z MMR. Average skill gap distance for these games for my team: A, Average skill gap distance for these games for enemy team: B.

If you would like to include the average collective team MMR's that would be beneficial additional information, but it is not required, we know the systems usually pair teams with collective MMR gaps of about 50 MMR, and this has remained consistent for pretty much each season. For the sake of the experiment please also include your rank when you begin this process and your rank at the end of sampling your dataset. This is needed so that we can see how much the tier one plays in affects the skill gap of players allowed on that team, which could have significant meaning, or it could be a completely insignificant factor.

Now, whether or not you are playing as part of a premade or not will have some influence on the data you mine, so its important to include which of these games was played as a solo player, and which were played as part of a premade and how many members on each team were premade. This can be more difficult to asses, but op.gg does have the handy "played recently with" information so you can use that to help you determine if players play together often (we know that on occasion players will get paired with the same individuals more than once WITHOUT being premade, but for the sake of the experiment we will assume any members playing more than 2 games together are part of a group) and that can be included with your documentation as well. Its important to include premade data so that we can see how much, if any, affect playing as a premade has on the skill gap for players within a team. This will be the most unreliable datapoint we mine, but with enough players performing this experiment we can get somewhat accurate results.

That concludes what needs to be done for the sake of the experiment. If you would like to play 20 fresh games, go for it. If you would like to simply use the last 20 in your match history, go for it. It won't make a difference unless the dates for play are significantly far apart (say about a month long gap, which is the baseline for most decay patterns which could affect the results of your datamining. If you have a gap like this please note it for the sake of experiment completeness, or choose to play enough games that all your sample size of 20 occur within the last month) from one another. Thanks in advance to everyone who chooses to participate and help provide data that can improve the overall state of league!

9 Comments

doiminat0r36007/29/2016, 9:05:07 PM4 votes

sees post Oh an experiment! Let's take part!

scrolls down wall of text with no TLDR

Welp, maybe the next one.

jaymc11307/29/2016, 8:45:59 PM1 votes

I'll be including a personal example for this here in a little bit guys, I've got some things to take care of at work before I can free up some time to provide it. Hang tight!

I Want Sweets7/29/2016, 10:30:27 PM1 votes

Last game, I was 5/1 mid and I had taken two of their turrets in mid lane. Then everything went to hell. I thanked bot lane but there were 4 enemies so we backed. Then jungle goes 1v4 dies and support follows. We go b then group mid. One shot from jhin did 300 dmg and more cause they fed him. I checked my team's score and all lanes fed, EVERY SINGLE ONE OF THEM. I don't say that I am too good for the help and other bs like that, but why am I in the same team with players who don't know how to play? That is unfair for me to lose LP since the game was practically a 1v5. Of course they also flamed me for being useless. Hate this matchmaking and I want it to be improved. I still remember that Nasus I was with who was 2/30 something and lost more than 20 games in a row.

jaymc11307/29/2016, 10:52:30 PM1 votes

Thanks to Sheep Shearer for being our first volunteer participant! Lets get down to business!

His last 20 ranked matches: https://i.gyazo.com/9d806ccf4d097dad5d244b2147e5b966.png https://i.gyazo.com/f3a76c034db66aeca1f341b7fda7dbf9.png https://i.gyazo.com/adce894c517939ef61dee4eacba7f845.png https://i.gyazo.com/3b43bcbd7ae1607f6bb1c9fba7f7edb9.png

The record for those matches is 6-14 so its likely his ranking has changed somewhat in that time frame as he has slipped a little in the last month. Op.gg has his current MMR at just shy of 1300 (1298) and lolking says his MMR has dropped about 100 points in that time frame from about 1400ish. He is currently silver 3 and was likely silver 1 or 2 at the start of that time frame. Using lolking in tandem with op.gg we can backtrack players MMRs to the dates the games were played.

Game 1 Friendly team MMR gap: 325 (estimated using both sites, 300-350, 1600ish-1300ish). Game 1 Enemy team MMR gap: 275 (estimated using both sites 250-300 1575ish-1300ish). Game 2 Friendly team MMR gap: . Game 2 Enemy team MMR gap: . Game 3 Friendly team MMR gap: . Game 3 Enemy team MMR gap: . Game 4 Friendly team MMR gap: . Game 4 Enemy team MMR gap: . Game 5 Friendly team MMR gap: . Game 5 Enemy team MMR gap: . Game 6 Friendly team MMR gap: . Game 6 Enemy team MMR gap: . Game 7 Friendly team MMR gap: . Game 7 Enemy team MMR gap: . Game 8 Friendly team MMR gap: . Game 8 Enemy team MMR gap: . Game 9 Friendly team MMR gap: . Game 9 Enemy team MMR gap: . Game 10 Friendly team MMR gap: . Game 10 Enemy team MMR gap: . Game 11 Friendly team MMR gap: . Game 11 Enemy team MMR gap: . Game 12 Friendly team MMR gap: . Game 12 Enemy team MMR gap: . Game 13 Friendly team MMR gap: . Game 13 Enemy team MMR gap: . Game 14 Friendly team MMR gap: . Game 14 Enemy team MMR gap: . Game 15 Friendly team MMR gap: . Game 15 Enemy team MMR gap: . Game 16 Friendly team MMR gap: . Game 16 Enemy team MMR gap: . Game 17 Friendly team MMR gap: . Game 17 Enemy team MMR gap: . Game 18 Friendly team MMR gap: . Game 18 Enemy team MMR gap: . Game 19 Friendly team MMR gap: . Game 19 Enemy team MMR gap: . Game 20 Friendly team MMR gap: . Game 20 Enemy team MMR gap: .

Well all I had time to investigate before leaving work was that one game. I'll do the others sometime tonight after I get home! One game is not enough of a sample size to form any sort of a trend, but we can see from this one game that Sheep Shearer's team had both the worst and best ranked players (at the time the game occurred) on his team. It is also worth noting that in this one game the opposing team had a 2 time previously platinum player who was the only player who had ever previously ranked outside of gold. Again, too small a sample size for it to mean anything, lets get the other 19 games in before we try to identify any trends over Sheep Shearer's last 20 ranked matches. Until I get home folks and have some time to complete the investigation for this player feel free to begin your own investigations and post results. Feel free to update things as you have the time.

legendarytiger7/30/2016, 12:18:39 AM1 votes

this seems interesting. i dont play much ranked due to the current state of league (dynamic disaster) so i wont be helping much, but i DO want to keep tabs on this, as i have a theory that large skill gaps between TEAMMATES is highly likely to create more toxic players, or simply a more toxic environment.