I don't think there are any NZ or Albania only tournaments in the DB ;-)4-man tournaments are a showmatch pretty much too, slightly expanded on the usual premise.
While i can't say that the your model is wrong in anyway or prove to you that the miscallibration between the projections and the actual state of competition is factual, i do have a suggestion that would potentially assist in making it more accurate.
How about throw out all national tournament (Champion of Albania,New Zealand etc.), narrow invitational (<16 players) + ECL, tournaments pre 2016 (they don't really serve a purpose for seeding anymore at least).
What you end with is high calibre tournaments like (NAC, KOTD, HC etc.) + some smaller rank tournaments, and you have an actual ranking of who performs best based on the accessible tournaments.
Although that kinda counter what i said initially about including everything, Being too diligent into including everything is conterproductive as it dilludes the model rather than callibrating it.
Low-tier is either regional tournaments or small prize-pool including 2-3 experts with various other 2k players.
The implementation of the page is fully in the hands of @mungo__ , so my knowledge is quite limited.@Michaerbse I could go through the previous 6 pages, because I think it was answered before, but maybe you know the answer:
Would it be possible to have a similar page/table, but with 2 filters: 1. Time range, 2. tournaments?
So basically you write start and end data and have a checklist with all tournaments included in the database.
Then you could for example for an upcoming Arena tournament choose to include all Arena tournaments from the last 2 years.
That could help a lot with specific seeding.
KotD 3 for example chose 4 tournaments to include in the seeding. That way you could choose whatever you want for your seeding relatively easy.Data is already borderline too thin when including all events, I'm not sure to what extend it would be meaningful to select only the handful of arena-exclusive tournaments that happened in the last few years
Ok yeah, makes sense in that regard. You'll have to mitigate a bit with the DE ladder I guess further down the table, like they did for kotd3.KotD 3 for example chose 4 tournaments to include in the seeding. That way you could choose whatever you want for your seeding relatively easy.
Always depends on how many players you have and how many tournaments you want to include. If it's only a few it's easier to just fill into a table by hand.
But let's say you want to have a tournament on open land-dominated maps, then it could already be 20+ tournaments in 2-3 years.
this might have been true 2 years ago but nowadays there are literally mutiple tournaments every week. I think the top ~50 players is more or less an accurate representation of their current level.There are not enough tournament games to represent the current form of the players.
As one of the 2 people (with @Michaerbse ) updating the site on day by day basis I can try to answer your questions and explain as much as i can.I wrote a lengthy suggestion on the website, but I might as well mention it here:
Including the early rounds of Red Bull is probably a terrible idea. You'll notice that almost every top player that took part in the qualifiers gained elo; you really need a huge underperformance to drop elo there because of all the easy 2-0 of R3. If you still want to include them, there should be a way to better evaluate people who enter your rankings. I reached Round 3 in previous qualifier and expectedly lost 2-0 to Daniel. You put that game as him being 2050 and me entering as a 1900 when I'm 1300 on ladder, so that actually got him decent points even if there was a gigantic mismatch.
That's some insane ladder inflation in the making if a 1300 can enter as a 1900: I was far from the only one, it's just two rounds to pass (in my case I beat a 900 then my 1600 opponent forfeited). I think that I should have entered as 1500 elo max if at all, and that actually RedBull qualifiers should only be included from R4 anyway (when top players start facing each other). Daniel shouldn't get any points for beating me, sorry for wasting his time.
To showcase that, out of the top 50, only 12 players recently lost elo, mostly players who did not take part in Red Bull qualifiers (Hera, Liereyy, Vinchester, Max, Lyx, ...). The overwhelming majority of players in the top 50 recently gained A LOT of elo thanks to small events where players over 2k2 on ladder faced players below 1k8 on ladder but received decent points for that.
That being said, I want to compliment the excellent work. It's fascinating to watch all those events and graphs.
Three suggestions on top of everything:
- including more small tournaments with players between 1800-2000 could give a better estimate of these players' level so that their ranking is more accurate when they face top players and they're not all between 1890 and 1910 with 1-2 games played. Visible Cup should help with that. Of course, they have to enter as 1k9 players (or even 1k8 players) if that's to be included. There are some now, but it's inconsistent. I can help with some input if you lack time. First thing that comes to mind for me is obviously the 1V1s of WDC Sand League because I'm the admin there, but there are probably tons of other events like that that could add granularity in the 1900-2000 range of the website.
- including the option to compare more than 2 players (unless I missed it). I think you could compare up to 4 players at once and still have it really readable with the colours. Ideally up to 6-8.
- another cool feature would be the option to zoom in on parts of the graph to make things more readable.
The last two would allow to compare the DE top 4 to see that they had almost exactly the same rating in late 2020 and to see how it shifted towards that all over 2020 then shifted away from it again since then.
Well, qualified players will face ones who were invited in the main event, resulting in them losing elo again (also losing a lot of points if they were "overrated" because of the qualifiers).Every qualifier should not be counted, it's not the tournament, it's a way to get in there.
Players who don't play the qualifiers will always be "penalized".
Every qualifier should not be counted, it's not the tournament, it's a way to get in there.
Players who don't play the qualifiers will always be "penalized".
Including the early rounds of Red Bull is probably a terrible idea. You'll notice that almost every top player that took part in the qualifiers gained elo; you really need a huge underperformance to drop elo there because of all the easy 2-0 of R3. If you still want to include them, there should be a way to better evaluate people who enter your rankings. I reached Round 3 in previous qualifier and expectedly lost 2-0 to Daniel. You put that game as him being 2050 and me entering as a 1900 when I'm 1300 on ladder, so that actually got him decent points even if there was a gigantic mismatch.
That's some insane ladder inflation in the making if a 1300 can enter as a 1900: I was far from the only one, it's just two rounds to pass (in my case I beat a 900 then my 1600 opponent forfeited). I think that I should have entered as 1500 elo max if at all, and that actually RedBull qualifiers should only be included from R4 anyway (when top players start facing each other). Daniel shouldn't get any points for beating me, sorry for wasting his time.
To showcase that, out of the top 50, only 12 players recently lost elo, mostly players who did not take part in Red Bull qualifiers (Hera, Liereyy, Vinchester, Max, Lyx, ...). The overwhelming majority of players in the top 50 recently gained A LOT of elo thanks to small events where players over 2k2 on ladder faced players below 1k8 on ladder but received decent points for that.
As hallogallo mentioned our rough guideline is to include tournaments starting from rounds where 2k players start matching each other. In the current RBW qualifiers, there were roughly 70 to 80 2k players signed up so we thought of Ro128 to start with.@hallogallo I think you miss my point: I don't really care that players like me have a provisional elo and I know I'd need more games to appear in the ladder, and that I'll lose more and more points; I'm more worried by the fact I fed 4 points to Daniel because this was considered a 70% chance of win for him every game when it was actually 100%, because he was 2060 and your system considered I entered at 1900 which is only a 160 elo difference even if on ladder there is a 1000 points difference between us. I don't care that it's just 4 points and that Daniel might lose them to someone else; it's just the sign of something problematic.
Disagree here. The value of the data comes through quantity. Every match with a competitive meaning helps the system better judge the player levels. And RBW/KotD/HC qualifiers seem to be way more competitive for most players than Tier B tourneys, even if they only have a very small chance of qualifying. If we only include main events, we will only have data on the top 25 players (optimistic guess) and I doubt that it is even possible to get an accurate reflection of top 15 players simply by the lack of matches between the "low rates" within that range.Every qualifier should not be counted, it's not the tournament, it's a way to get in there.
Players who don't play the qualifiers will always be "penalized".
As mentioned we tend to draw the line at 2k, but this usually includes players between 1800 and 2k as well as they are "filling up" the bracket. Visible Cup is on our list and data collection has already started. We also still have WDC Sand League on our list and did not come to a final decision yet on whether / from when to include that. :-)- including more small tournaments with players between 1800-2000 could give a better estimate of these players' level so that their ranking is more accurate when they face top players and they're not all between 1890 and 1910 with 1-2 games played. Visible Cup should help with that. Of course, they have to enter as 1k9 players (or even 1k8 players) if that's to be included. There are some now, but it's inconsistent. I can help with some input if you lack time. First thing that comes to mind for me is obviously the 1V1s of WDC Sand League because I'm the admin there, but there are probably tons of other events like that that could add granularity in the 1900-2000 range of the website.
I don't know tbh. You can download the data in the "About" section, maybe that helps. It doesn't contain any decimals so I think it happens at least after every match.@hallogallo @Michaerbse do you happen to know where in the calculations the rounding takes place and how often the values are rounded (after every game? every series?) Some of the serious seem like there has been a lot of rounding, slightly influencing the ratings.
Working fine for me but you are not the first one to mention that.Is it just me or is the site down?