Season 7 win matrix | Forums

[Klingentaenz3r]

Also imo pulling from platinum is probably too low. they probably should only be looking at master and grand master. And even then i'm not entirely sold on using ranked data at all. imo they could have easily just looked at the top 1% instead of 2.5%.

That is what I thought about that at first as well. So Aramusha was not that surprising in that regard as in that skill level bracket people are often not as turtley as the higher ups and he has a good stance against assassins in general and a bad one against defensive ones. You can see that reflected in the table as well. Combined with a low pick rate people are just not used to face/handle aramusha. Also consoles and PC data combined once again. Can't make too much out of it. I can just say Aramusha needs something for turtles and some stamina improvements

Anyhow back to what I wanted to point out: I rewatched the Den and they acknowledged that they had to pull data from this plat including pool since otherwise they would not have enough samples of all characters overall. Meaning some characters, and you can imagine which I am sure, are not picked too often at higher level ranked play and hence not enough data. Note you need the same amount for each and every character otherwise the whole matrix's messed up.

as for the 4v4 data... top 4% is meh but I guess in order to compare developments due to the stance bug fix they wanted to keep that pool size.*shrug*

[Siegfried-Z]

@Siegfried-Z
I'm right there with you about the fear of more 400ms lights. It's almost like if characters had more thoughtout kits with more/better options we wouldn't have to rely on sheer speed alone to beat the human reaction time. If characters had more softfeints and mind games matches could be more interesting with anything other than lightspam at all levels other than top level play, and at top level play with more options turtling would require much more effort and lead to an ultimately more skill based game as opposed to a reaction based natural selection.

Yep that's the point. We want more mix up, more mind game but no more 400ms lights which are too fast for Console.

400ms mooves are just : Frustrating first -> Removing skills -> Makes new players leave the game -> Boring to face as they are to use (where is the fun to be in Super Saiyan mode focus tryharding to block Roch, Nuxia or Tiandi lights ? Where is the fun in the other hand to play a game just for smashing Light button ?) Sorry i respect everyone choice but i would never understand how people can be attracted by lights spammy char ... except musha because at least he has to mix up to enjoy 400ms lights he can't just smash the button.

Just look at Roch rework. i believe all Roch mains would have choose to get the Tozen kick with a follow up than 400ms lights ...
Same for Wulin .. Excuse me but after all 400ms complains, they've released 4 char with 400ms mooves .. Even JJ as a Heavy and Tiandi as a Vanguard got one.. i just don't get it. It's like they don't want to invest some time and make the effort to really think about some nice and healthy solutions to improve offense in the game.

They don't have time for that but they've got some for stuff nobody care as New outfits, new emotes and so on because gess what it makes people spend steel ...

One day every Char gonna have 400ms mooves and this day no one would play FH anymore which is sad for such an amazing and lovely concept.

[NHLGoldenKnight]

No. If the point of the data is purely just "for fun" then yes. Grabbing a wide pool is a good idea for this game because it's interesting to see where things fall despite no one ever really agreeing with each other. If the point of the data is supposed to help the devs in terms of finding problems and using it as a baseline to start looking to balance then in this case no. Because this data doesn't actually reflect the state of each individual hero let alone their match ups with others.

Easy example is aramusha's whole table. None of the people he has over a 50% win ratio against are match ups that are actually in his favor. But even if the devs were doing top 1% anyway the data is flawed because of the stats they base it on. Which is primarily their skill system.

So really any data they could possibly gather wouldn't be super accurate. It would just be MORE accurate the smaller the pool is (aka top 1%)

No data would be super accurate but there is problem in using small sample size, such as 1 or even 2.5% . If this was game played by milions, it would be ok. But overall population is very low if we talk about regular players, not those who play for a few weeks and than leave. So top 1% is how many players exactly? 200, less?

Now, let's say it's 200 just as example. How many of those 200 play with every hero? How many games we have per each hero if we are getting info from only 1%?

I honestly don't think there is enough games per every hero in top 1% and even in top 2.5% to get any useful numbers. It can happend that info for one popular hero is gathered based on thousands of matches played while some unpopular heroes maybe were used only in dozens of matches.

Ubi should release numbers on which matrix is based off , if they haven't already. What was the sample size, how many matches per hero etc. Without those numbers, matrix is useless in statistical sense.

[NHLGoldenKnight]

Not universally true, especially relating to feedback about balance amassed from individuals with varying levels of experience, competency and knowledge. Players below a certain threshold are not suited to be a source for feedback pertaining to the game's balance and state of heroes, and PvP games' balance in general should always be based mostly on what the top players think - they often know the game way better than the developers themselves, they can use, in this case, the heroes to their full potential and they can assess how easy or difficult it is to counter other heroes at their full potential, thus making the most accurate and unbiased conclusion on who is trash, bad, mediocre, good, competitively viable, etc.

Anyway, Conqueror has lower win % than I thought he'd have. Aramusha being top 4 already speaks ill of the increased data breadth. And Centurion is pretty much solidifed as the 4th or 5th worst duelist in the game.

Gonna be a long wait...

I agree that there should be certain threshold but in caxe of For Honor, 1 or even 2.5% can't be that because of overall small sample size. We are talking about few hundred of players of which most of them stick to playing mostly few certain heroes. In post above, I expkain what is potential issue but in short-there is just not enough data per hero to make matrix any useful.

I would however disagree that game should be balanced on what only top players think. Firs of all, they can be biased like all of uus. If majority of top players are assassins, guess what? They will complain about turtle meta and suddenly we get 400ms lights. Also, Ubi is often testing balance by having top player play the game in perfect, almost laboratory like conditions. No latency, best equipment they can get and on top of that those top guys know each other very well. It makes a huge difference. Why not pull data for average latency and balance attack speed based on that? Because those are real conditions for majority of players.

Just as example, I am playing as goalie on NHL games and I was among top 10 goalies in the world for last season. But guess what? I don't think goalies should be balanced on what I have to say because playing constantly at the very top is easier in a way that you know opponents and their tendencies. So if EA approached me to ask about the balance I would say how everything is great, it should be harder for the goalies even. But in reality that simply isn't true and for a long time, goalies were simply broken. But just because it works for me, doesn't mean it works for vast majority of playerbase and game depends of that huge mid tier chunk of players. It doesn't depend on top percentage or the botton one. Mid tier and upper mid tier especially is what's count for game to be balanced, well received and for game to eventually survive.

[Knight_Raime]

That is what I thought about that at first as well. So Aramusha was not that surprising in that regard as in that skill level bracket people are often not as turtley as the higher ups and he has a good stance against assassins in general and a bad one against defensive ones. You can see that reflected in the table as well. Combined with a low pick rate people are just not used to face/handle aramusha. Also consoles and PC data combined once again. Can't make too much out of it. I can just say Aramusha needs something for turtles and some stamina improvements

Anyhow back to what I wanted to point out: I rewatched the Den and they acknowledged that they had to pull data from this plat including pool since otherwise they would not have enough samples of all characters overall. Meaning some characters, and you can imagine which I am sure, are not picked too often at higher level ranked play and hence not enough data. Note you need the same amount for each and every character otherwise the whole matrix's messed up.

as for the 4v4 data... top 4% is meh but I guess in order to compare developments due to the stance bug fix they wanted to keep that pool size.*shrug*

See I didn't watch the stream. So if they had to pull from plat in order to have a reasonable size of players then they shouldn't have bothered pulling from ranked to begin with.

No data would be super accurate but there is problem in using small sample size, such as 1 or even 2.5% . If this was game played by milions, it would be ok. But overall population is very low if we talk about regular players, not those who play for a few weeks and than leave. So top 1% is how many players exactly? 200, less?

Now, let's say it's 200 just as example. How many of those 200 play with every hero? How many games we have per each hero if we are getting info from only 1%?

I honestly don't think there is enough games per every hero in top 1% and even in top 2.5% to get any useful numbers. It can happend that info for one popular hero is gathered based on thousands of matches played while some unpopular heroes maybe were used only in dozens of matches.

Ubi should release numbers on which matrix is based off , if they haven't already. What was the sample size, how many matches per hero etc. Without those numbers, matrix is useless in statistical sense.

Yes the current data in any form is useless to us in the community right now. We need more info or have them explain it better.