THE META REPORT NAME IS TOO LONG, TOO DAMN LONG (n°8)
by Legna
2021/05/24
Data
First week of patch 2.8.
- Sample size: 43808 matches 1
- Of all games, about 5% was not available to collect
Regions’ Frequencies
By number of times a region is chosen
| Region | Play Rate | N |
|---|---|---|
| Bilgewater | 2.5% | 4372 |
| Demacia | 7.82% | 13670 |
| Freljord | 9.96% | 17416 |
| Ionia | 12.43% | 21746 |
| MtTargon | 7.75% | 13547 |
| Noxus | 11.43% | 19992 |
| PnZ | 8.96% | 15669 |
| ShadowIsles | 15.22% | 26629 |
| Shurima | 23.94% | 41866 |
By number of times a card within a region is chosen
| Region | Play Rate | America | Asia | Europe |
|---|---|---|---|---|
| Bilgewater | 2.42% | 2.48% | 3.52% | 1.85% |
| Demacia | 8.15% | 7.15% | 10.39% | 8.29% |
| Freljord | 10.34% | 9.85% | 10.63% | 10.77% |
| Ionia | 15.26% | 15.27% | 14.70% | 15.50% |
| MtTargon | 7.94% | 8.69% | 7.21% | 7.41% |
| Noxus | 10.69% | 10.99% | 10.34% | 10.52% |
| PnZ | 9.41% | 9.85% | 8.20% | 9.44% |
| ShadowIsles | 17.26% | 17.24% | 15.94% | 17.88% |
| Shurima | 18.53% | 18.49% | 19.08% | 18.33% |
Champions’ combinations
In this section I provide the play rate of which combinations of champions (plus the regions) are used in a deck. The champions showed right before a game starts for example. Right now it’s a simple approximation of the archetypes played in the ladder as such information is not restrictive enough.
Date by date
With the previous reports showing that there’s a huge cluster of decks all with a very low win rate the following graph will show the play rate day by day only for the top5 most played decks.
Something that was suggested by the previous graphs and here sort of confirme, it seems that this week was very stale regarding the meta and what was played. I would normally assume that this is because the meta has being solved. But it this really the case? I actually don’t think so, this meta feels much more different from previous “solved” ones like the Go Hard / Fizz TF and Darrowing meta. To be fair those were solved meta also thanks to a tier0 deck which is lacking here.
Win rates
Tie games are excluded
Win rates of the most played combination of champions, against all decks, this week. Cases with less than 1% play rate games are excluded
Underdog(?) win rates
Top Win rates of the least played combination of champions. Min 300 games and a play rate of less than 2% play rate (in this sample)
Most played combination
The win rates on the grid are among the 10 most played champion combination. Match-ups with less than 300 games are not included
Match-up grid
While this is one of the most interesting data for many the results are still heavily affected by the small sample size. They may be the 10 most played combination of champions but it’s still a 100 cell grid. Also, while my approximation for archetypes usually works, this weeks more than usual, two decks are not well reported: Dragons and Azir/Noxus. In these two cases the inclusion of J4 / Garen / Zoe and Draven / Darius create different values for in the end it’s the same deck with just a couple of different card. I already have an idea and part of the code how to solve this but will use it for next week. I could use Dr.LoR approach with bayesian statistics but I want to be consistent with my metholody in order to have a better comparison for my results. Of course I can still make changes in the future if the quality vastly improve with them.
Match-up table
No filter this times because of the small sample size
The meta and you
Thou are the meta and the meta are thou. Following the section where I wrote I don’t think the meta il solved I wanted to try a little “experiment”. The idea was of estimating the “Indirect Standardized Win Rate”. For those who don’t know it, the indirect standardization usually refer to using an estimation using a “reference population” (at least for mortality rate), so answering the question: what would be the WR if the play rate had values provided by me and not the real ones. Would the best decks remain the sames? In that case I would say that the current top decks are indeed the strongest, at least among the ones I selected.
If I use an equal distributed population (and limit myself to the top15 combinations) the results are the following:
This is mostly just playing a bit with the data, but at a first glance it seems that certain decks remain about the rest, like the current top with Draven/Ez, TLC, and so on…
Dragons on the other hand seems to be good thanks to the current play rates, they may easily drop if people start to play other “top decks”
At the opposite side Thrall and Shurima Overwhelm may actually be decks that people should be aware of? Shurima overwhelm is already a decent choice but Thrall decks would see a massive improve in WR with a change in the meta.
As said, this is just playing around with simple hyphotesis. I could try different cases of “reference population”, other decks and so on. If it seems interesting I may continue to work on this.
Deck Structure of the week
This week I choose to show the deck with the highest WR in the grid: Ashe / LeBlanc.
How to read the table:
Play rate: How often a card is included in this class of decks / the table is order by this column.
3/2/1 is the relative and absolute frequency of the number of copies in the decks that plays them
Frequencies from 50% to 100% are colored from shades of green to white to identify more easily the highest values
Ashe / LeBlanc
Cards’ presence
Most played cards
Top 3 most used cards within each Region
Legal bla bla
This Meta Report was created under Riot Games’ “Legal Jibber Jabber” policy using assets owned by Riot Games. Riot Games does not endorse or sponsor this project.
Ranked games / Patch 2.8 / Master players
Last update: 2021-05-27 04:52 (UTC)
EU Master players in the ladder: 404 while the number of Master players I recovered from matches is: 475 so missing: -71
NA Master players in the ladder: 413 while the number of Master players I recovered from matches is: 480 so missing: -67
ASIA Master players in the ladder: 149 while the number of Master players I recovered from matches is: 168 so missing: -19