Kealah | 2016-03-11 13:47:28 |
5+ level:
10+ level:15+ level: 20+ level:25+ level: 30+ level:
HeadphonesGirl | 2016-03-11 14:05:58 |
lol @ time 30+ 49%
ban this broken class guys
Kealah | 2016-03-11 14:08:24 |
Demonstrated stats above level 15 does not make much sense. Shown exclusively for fans to learn as much information as possible.
HeadphonesGirl | 2016-03-11 14:15:55 |
Demonstrated stats above level 15 does not make much sense. Shown exclusively for fans to learn as much information as possible. I'm not sure how statistics can "make sense" or not.
mamouhaha | 2016-03-11 15:59:51 |
... I'm not sure how statistics can "make sense" or not.
It's because the sapmle is really small.
HeadphonesGirl | 2016-03-11 16:19:51 |
... It's because the sapmle is really small.
It's funny because when lower level statistics are used with a larger sample size, the other argument is made, that low level stats aren't good enough to show the actual balance of the classes. I even made exactly the argument you're making now as to why in the past the stats we've been shown have been lower levels. Now there's 30+ and that's not good enough either. Seems there is no way to win.
Sinist | 2016-03-11 19:16:18 |
lol @ time 30+ 49%
ban this broken class guys better buff, obviously it is way too weak to be competitive .
Now there's 30+ and that's not good enough either. Seems there is no way to win.
Yes, because community unfortunately is too small, so statistics is much less reliable than, lets say, in HS case CyberneticPony | 2016-03-12 15:05:46 |
I actually think Time was hurt a lot by the nerf to Chrono Engine, which is good, because it's where we want it to be for balance now.
Sinist | 2016-03-12 16:38:41 |
well, nerf doesnt show in lower levels and on higher levels stats are too unrepresentative... Imho time 7 is not as dangerous as time 2 or time 4/6
Modified by Sinist on 2016-03-12 16:41:57 CyberneticPony | 2016-03-12 17:25:28 |
well, nerf doesnt show in lower levels and on higher levels stats are too unrepresentative... Imho time 7 is not as dangerous as time 2 or time 4/6 Time 2 is more dangerous, but it's also very easy to signal. From experience, it's signalling that really affects the game at high ranks.
Many nerfs don't show at lower levels because lower level players tend to miss a lot of core concepts in the game. Phoenix might seem strong if we just took a look at data at the lowest levels.
Modified by CyberneticPony on 2016-03-12 17:26:39 Sinist | 2016-03-12 18:06:03 |
Easy to signal, very hard to counter and I dont think that 20+ level players are that "noobish"
GrimJ0ker | 2016-03-14 13:27:37 |
Dominator vs Sorcerer 81.8% r.i.p.
SpectroStat | 2016-03-14 13:41:37 |
Dominator vs Sorcerer 81.8% r.i.p. wrong sorcers)
srbhkshk | 2016-03-14 13:42:31 |
Dominator vs Sorcerer 81.8% r.i.p. To be fair though, that's just 11 games we are talking about. Wavelength | 2016-03-15 17:42:29 |
Time was also hurt by Plynx's disappearance. ;)
Cooler | 2016-03-15 19:54:34 |
Now there's 30+ and that's not good enough either. Seems there is no way to win.
Exactly! :)
ErmelSylphion | 2016-03-17 16:08:30 |
On 30+ Lvl ... interesting Winrate I noticed it's Spirit vs Cleric. 3 matches out of 10 only won by Cleric is quite low. Shouldn't both classes be sort of "mirrors"? :) :) :) Modified by ErmelSylphion on 2016-03-17 16:15:43 Wavelength | 2016-03-20 04:25:08 |
On 30+ Lvl ... interesting Winrate I noticed it's Spirit vs Cleric. 3 matches out of 10 only won by Cleric is quite low. Shouldn't both classes be sort of "mirrors"? :) :) :)
The sample size is 22 games - small enough that this winrate is basically statistical garbage. It could have very easily been a 70% winrate for Cleric instead. Looking at Level 15+ (which has several hundred games as a sample), the winrate in this matchup drops to a very reasonable 54%. MikeBnDe | 2016-03-20 11:18:29 |
I would really be interested if we could find 4 top players (level > 38 or so) who would play lets say 30 games against each other with every class combination (so 6 * 30 =180 games total with every class combination) and then look at the win-rate percentages. I think that would be a good estimate regarding the strength of each class. Anyone willing to volunteer? ;-)
HeadphonesGirl | 2016-03-20 11:37:46 |
I would really be interested if we could find 4 top players (level > 38 or so) who would play lets say 30 games against each other with every class combination (so 6 * 30 =180 games total with every class combination) and then look at the win-rate percentages. I think that would be a good estimate regarding the strength of each class. Anyone willing to volunteer? ;-)
With only 4 players the results would be very heavily skewed by the favored classes of each player. People are better with some classes than others and not always the strongest ones.
MikeBnDe | 2016-03-20 20:31:57 |
@HPG I knew that this counter argument would come. It is so obvious. Of course there are some assumptions/requirements included with my suggestion. One of them is that a real top player has mastered all classes to such a high degree that the phenomenon of favorite classes is negligible. Even if thats not totally the case, there are more additional arguments, but i leave it to yourself to find them. Modified by MikeBnDe on 2016-03-20 20:37:00 HeadphonesGirl | 2016-03-20 21:29:40 |
@HPG I knew that this counter argument would come. It is so obvious. Of course there are some assumptions/requirements included with my suggestion. One of them is that a real top player has mastered all classes to such a high degree that the phenomenon of favorite classes is negligible. Even if thats not totally the case, there are more additional arguments, but i leave it to yourself to find them. Whether the argument is obvious or not doesn't have any bearing on whether it's true. The only situation where personal style and ability would not skew the results is if all 4 players had mastered the game to the extent that it was "solved," ie, they would always make the mathematically best move possible under any circumstance. Given that no one has yet even succeeded in solving chess, this will never happen, as it never happens with any competitive game worth playing. You could get perfectly objective results using this method to test tic tac toe, but that's about it. I'm also going to also counter your strategy by saying that there are many additional arguments as to why I'm right, and it is officially your homework assignment to now figure them all out.
Modified by HeadphonesGirl on 2016-03-20 21:30:45 SpectroStat | 2016-03-20 22:45:32 |
Anyone willing to volunteer? ;-)
I can be one of them
MikeBnDe | 2016-03-20 22:51:43 |
Cool spectrostat :) But before you volunteer, please consider that highly intellectually stimulating post of HPG. Read it carefully. Its sheer amount of insights is breathtaking. So if you have the same definition of what constitutes a 'strong class' in a game played by humans, you might come to the conclusion its not worth the effort.
HeadphonesGirl | 2016-03-21 01:38:09 |
Cool spectrostat :) But before you volunteer, please consider that highly intellectually stimulating post of HPG. Read it carefully. Its sheer amount of insights is breathtaking. So if you have the same definition of what constitutes a 'strong class' in a game played by humans, you might come to the conclusion its not worth the effort.
I don't really know what your problem is man. You seem to have taken my first post personally and gotten snarky but it wasn't personal, I was just making a point.
Wavelength | 2016-03-24 04:12:18 |
I agree mostly with HPG. While such a format would be a highly interesting experiment (and I'd be keen to see the results), I don't think we could really derive anything conclusive or even close to it from such a small sample size.
I remember a while ago, before winrates were widely available, discussing with FinalSlayer about Holy vs. Time - I thought Holy was the stronger class and definitely the better one to take in heads-up, whereas he thought Time was the strongest class and had no really weak matchups. We agreed that we could never determine this for sure but that we should at least back up our words with a long series of duels, agreeing to 20 matches over 2 days (me always on Holy, him always on Time). He took the first set 6-4 and said that a 60% winrate was pretty big, admitting that yeah it was a small sample. I took the second set 7-3 and reminded him that a 70% winrate clears a lot more error than 60% does, admitting I had only proven his theory wrong and that 11-9 is a far cry from proving my own right. In the end we had fun and both realized that most likely we had formed too-strong beliefs that we could only partially back up.
Modified by Wavelength on 2016-03-24 04:12:57 MikeBnDe | 2016-03-24 12:38:05 |
@Wave: Please consider that you played 20 matches total, in this case it would be 180 (even though between different players). Of course a much higher number of players and a much higher number of matches per class combination would be better, i only said 4 and 180 (perfectly aware that this is far from perfect) because realistically, i cant imagine we can find a lot more players of level > 38 (dont ask me why i chose 38) who would be willing to play 1000 games or more per class combination. And please note that it wasn't only about sample size, it was implicitly also about what your definition/understanding of a 'strong class' is and - depending of your definition - whether skewing due to favorite classes etc. is a problem or not. If you define it mathematically (in terms of perfect moves which probably is the way that comes to mind first), you would have to use a 'perfect algorithm' to find the strongest classes. But even though this outcome might be of theoretical interest for some, humans will not come even close to copy that so such a definition is not particularily interesting to me because it might yield totally different results compared to human matchups. Moreover, we dont have such an algorithm, maybe we should try to recruit AlphaGo :)
Besides, no matter what your conclusions of the results are (maybe none), for me, personally, i would be - just like you - very interested in this experiment, besides for me it would be fun so see top players fighning for the overall victory in such a marathon So any other top players up for it? Tendou? Grim Joker? I would track the results the players mail me.
Modified by MikeBnDe on 2016-03-24 13:15:56 SpectroStat | 2016-03-24 13:01:38 |
By the way about the marathon. Prospectro now as time passes like this, where each participant plays every 10 games.
https://goo.gl/N9ZV1gModified by SpectroStat on 2016-03-24 16:23:30 MikeBnDe | 2016-03-24 13:04:30 |
SpectroStat | 2016-03-24 16:08:15 |
Try it now https://goo.gl/N9ZV1gThere is a table undergoing the championship that involves 15 players and each player plays every 10 games. There are serious limitations that reduce the choice of schools, with prohibitions and embezzlement for victories. Players with 40+ but in the tournament only four (Grim, Knife, Krugo and me aka TosheR)When the tournament will be played will be quite a lot of information on classes. What a win, what people don't like to play by yourself which class you prefer to block the opponent.Modified by SpectroStat on 2016-03-24 16:36:15 MikeBnDe | 2016-03-29 08:42:43 |
This tables could be interesting to me. Unfortunately, i dont speak russian, so i can't read a lot of things there :( Are there only russian players taking part? Or do you have an english version as well?
SpectroStat | 2016-03-29 12:57:44 |
At this point in my tournament play participants from Russia, Belarus, Kazakhstan and Italy.What exactly you want to translate? The rules and tournament format, description columns of results, grades used, shaded colors?Modified by SpectroStat on 2016-03-29 14:17:18 MikeBnDe | 2016-03-30 23:20:26 |
At this point in my tournament play participants from Russia, Belarus, Kazakhstan and Italy. What exactly you want to translate? The rules and tournament format, description columns of results, grades used, shaded colors? Well if you have the time, explaining the rules of the tournament/tournament format would be nice. Will there be another one of these tournaments? Could i take part next time? Or is my ranking not high enough?
Theodorus | 2016-03-31 14:37:07 |
Wanna talk about 47%ish sorcery? Mana burn is so cost/efficient high to be a void into repertorie of spells. SpectroStat | 2016-03-31 15:21:14 |
This is the second tournament of this format, the first was a few years ago he won the Kroma. Registration in reaching the tournament is not yet complete, but if you're willing to participate, you will need to play very many games. Level is not a problem, you have enough. The problem may be an opportunity to negotiate about the games due to the language barrier and the fact that negotiations usually take place in the theme on the portal Prospectro. GrimJ0ker advertises their capabilities through other players, with whom - Kealah and Krugopryad.
Modified by SpectroStat on 2016-03-31 15:26:52 SpectroStat | 2016-03-31 15:24:21 |
Wanna talk about 47%ish sorcery? Mana burn is so cost/efficient high to be a void into repertorie of spells. Yes, in the next topic was announced winrate the third slot of class Sorcery. And when it falls the spell, the spell is more often lost than when there sorcery 5
i have asked repeatedly for a small buff in mana burn: make it deal 3 damage to the opponent (equal to the amount of mana drained). what sort of MANA BURN is it if it doesn't cause pain at all?
|