There were some changes in the works for a while now, and I rolled them out a couple of days ago. This is more of a byproduct of another thing I’ve been working on, but still interesting enough to discuss.
And important note: the data was provided by STRATZ.
There are four new types of immortal ranked reports now:
- Top-100 and DPC Div 1 players, which has statistics for the matches with these players in them
- Top-500, or more specifically Top-10% and DPC Div 1 players
- Top-1000, or Top-20% and DPC players (both divisions)
- Top-2500, or Top-50
Where to find them
These reports can be found here:
- High Rank Meta Trends category includes all four types of them and can be found at the “Featured” tab
- Top-2500 category contains only top-2500 reports (and it’s hidden, like any other category mentioned below)
- Top-1000 category
- Top-500 category
- Top-100 category
All of the ranked reports also have their own custom styles, logos and banners to make it easier to distinguish at a glance which is which. The difference is subtle, but clear enough to make the navigation easier.
Also there are six different versions of the Immortal Rank image remade as SVGs, which I made for these custom styles and banners. Feel free to contact me if you would like to use them as well.
These high rank reports were moved to a separate category to avoide cluttering the Immortal Meta Trends category. For this reason they are going to be kept separate, and it’s also the reason why only “Meta” reports (last week/previous week) will be seen in the main lists (patch reports will be hidden).
I’m considering hiding all versions but Last Week by default since otherwise they will take over the main page. But more likely I’m going to start working on an updated version of the main page to make it easier to navigate featured reports and all the ongoing stuff.
Other important stuff
From this point on Items Stats will be disabled in the patch reports. I mean the data will be there (well, in the main Ranked Patch report at least), but not in any of the Top-X versions.
There are multiple reasons for this, all of them come down to the way the items data is stored and calculated. I have plans to improve it significantly, but for now all this data takes a lot of space, while processing timings and winrates for different heroes and timings starts to take more and more time.
While the plan to solve this issue already exists, there are many things that require complete rework beforehand. And with every day calculating items stats for the patch reports turns into a very tedious and long task, and the data took a lot of space to keep the balance between the calculation speed and the storage.
It wasn’t that big of an issue with one ranked report, but once you get five times more, the routine to update all the reports starts to add up. So for the time being the items data will be disabled for them.
I will still update the items data for every patch once it ends. And there are still Last Week/Previous Week meta reports, they will have items stats no matter what. Meta reports have more than enough data to work with (and they stay pretty much up to date), so I suggest you to use them instead.
In the current state it takes about 8–9 hours for all 15 reports to fully update. I might look into ways to speed things up a little down the road, but not right now.
Important things to note
First of all, these reports include all matches which had at least one player of their specific group. This means that a Top-100 report might have some matches with a big MMR disbalance in place.
Could I filter this kind of matches out? I guess. Will I? Probably not. Based on the numbers of matches played daily, I suspect that the percentage of this kind of matches will be low enough, so their effect won’t be that big. I might revisit this eventually, but not right now.
Second, the groups of players. While the short names have a specific number (500, 1000, 2500), the actual limits are percentages (10%, 20%, 50%). But since two of the regions have roughly 5000 leaderboad players, these percentages should roughly correspond to the same numbers.
The reason why it’s about percentages in the first place is the difference in the numbers of leaderboard players in different regions: 4.2k in SEA, 3.7k in NA, 5k in Europe, 1.5k in China, and the overall number of active players in ranked has a similar distribution.
My idea was that percentages should help to get players of roughly similar MMR from different regions, despite the inlated numbers, and also sort of even out the numbers of matches played. It seems like this idea worked for every region except China, but this might be because of the Lunar New Year happening right now.
In any case, for now I will leave these new reports as is, might adjust the parameters in future but most likely after a month or two.
No related research for now.
There are still other things to sort out for now, so there is no time for research, sadly. But there are enough things to work with, and there are some new additional parameters to work with.
For now I’m focused on my work with Tundra and PGL though. And the gigantic backlog of features of course.