There is no belief element on my part. I know for a fact because I know the subject very well.
I’ve linked this literal conversation on a statistics community and they all agree you don’t understand what bias is.
Not trying to bash you, just letting you know.
1 Like
I think you know a bit but your biggest problem (and this was nomuff’s too) is that you don’t actually read what the other person is saying and jump at the first opportunity to virtue signal what you supposedly know. This usually involves finding a mistake that you then bash. The problem is that because you didn’t read you misunderstood and then you look dumb. Then there’s also the problem that you far overstate your understanding and it’s clear that’s the case when you mix up terms. Just a few posts ago you were saying sampling mean when what you meant was population mean. You know, lots of little things just aren’t right when you post and that’s a separate issue from the factually wrong things you post like saying Elo is based on the normal distribution.
You know a bit and I’ll give you credit there.
Its not usually worth the time and effort to go into a detailed explanation (such as proving that the sum of logistic random variables is not logistic, which is out there if you want to look for it), and the mistakes that I have pointed out are critical mistakes that invalidate everything else.
1 Like
The funny thing is that if you knew the subject well it wouldn’t require effort. Have a good night.
1 Like
These forums dont have latex which is a major issue for some stuff. Goodnight.
2 Likes
Is anyone actually falling for this madness? Pvp has the lowest Score on aligulac by far. Its such a poorly designed matchup that blizzard actually put a bandaid on pvp just recently. Ofc protoss Players will look Bad if You substract their pvx score with pvp score.
Once again. Claiming madness without any reasonable proof why it should behave in that way. I could also claim something inherently stupid just to make Terran or zerg seem to be very op.
1 Like
Exacly what Sentry just said above.
I’ve pointed out that before, this chart is not indicative of balance because of how inconsistent PvP is, a rock, paper, scisors game cannot be a parameter of skill.
To add a little more, you are comparing matchups with an inconsistent one, to force an approach of rankings of PvT and PvZ to an inconsistent PvP would only contribute to exacerbate the imbalances for the Protoss side.
1 Like
No it doesn’t. It only does that in the simplified model where:
- mirror match skill level is completely predictive of non-mirror match level
- the average skill distribution by race is the same
Mirror match skill level is not predictive of non-mirror match skill level
This has been told to you multiple times and is the fundamental reason why many players whine about ladder imbalances.
At almost every single ladder rank, your average win rate across the board is about 50%.
This means that if you are doing well against any particular race - even if it is your race, then by definition, you are doing proportionally less well against the other races.
If, for example, my ZvZ win rate is an amazing 80%, the only way for my ranking to remain stable is if my performance against the other two races offsets that 80% and pulls it down. So I might expect that my ZvT performance be a dismal 30% and my ZvP performance 40%, or some other combination of bad.
Why? Because if my performance against the other two were better, then my absolute win rate would increase over 50%, and thus my rank would go up.
This is true at every single place in the ladder except for the absolute peaks, where there aren’t enough peers to create the average balancing of your rank upwards or downwards.
Skill distribution differences influence the significance of mirror match skill level in rankings
Example: Region A is a country where zerg is extremely popular and terran is unpopular. Because of this, zerg is overrepresented in the region compared to the other two races.
To be a top ranking player in Region A, regardless of which race you are, you absolutely need to be very good at fighting zerg because the majority of players you will encounter are zerg. On the other hand, in that same region, to be a top ranking Terran player, you do not need to be that good at fighting Terran, because there aren’t very many high skilled Terrans in that region.
Thus for region A, Zerg will need a much higher mirror rank (ie ZvZ performance) than Terran will (ie TvT performance) to be at more or less the same tier.
In general, in any region where one race is overrepresented, you will need a much higher mirror rank in that race to place high in the rankings. On the other hand, in any region where one race is underrepresented, you can get away with a lower mirror rank in that race to place high in the rankings.
Or, as above, Protoss is so underrepresented that PvP skill level simply is not as important as PvT and PvZ skill level at influencing rank. Which is exactly the opposite of what you’re saying.
TLDR this says nothing again about balance.
2 Likes
Cheesecake mate, from the way he’s arguing and making his points batz sounds like he’s about 16. You’re fighting with a literal child. You’ve started your case and you’re right. It’s clear to see the way he understands this algorithm is flawed. Give it a rest. You’ll never change his mind.
2 Likes
As a forum lurker, I’ve noticed that this account (lllIlIlIlIl-1983 is the full name-ID) seems to always drop a post in batzs threads to 100% back them up and 100% agree with how everyone is attacking batz. Even by using alot of the same words that batz used. AND their profile is hidden too just like batz.
anyone else smell a alt account around here?
3 Likes
Cool, can you plot this overtime, so it’s easier to see where your flaw come from? Appreciate it.
I guess because CS background (and mathematics for that matter is not a mean in itself but an instrument). In order to succeed one has to have both the mastery of the mathematical apparatus and the knowledge of the process at hand being studied.
Just having only the mathematics and making a slightly-wrong assumption leads to major errors.
Hey, Anesthetic, stop upvoting yourself and get a life.
So, Protoss and terran suck at their mirrors and what I said last time is now the conclusion?
https://i.imgur.com/zcB8Bic.png
What you are saying is simply factually incorrect. Mirror elo is the best predictor of skill, period. You can see here that Terran and Protoss mirror predicts their non-mirror with almost a 90% accuracy, and that the relationship between mirror and non-mirror is basically 1:1. Across an average of all players, on a per-race basis, the accuracy is very high. Furthermore, they have virtually identical distributions.
The exception is Zerg. Zergs have much higher mirror ranking than non mirror because the asymmetry of the non mirror gives them a disadvantage and deflates their rank.
That’s incorrect. There is basically a 1:1 ratio of mirror to non mirror Elo for terrans and protoss (on average), meaning it’s a perfect indicator of skill.
True it is an indicator of skill but perfect is a long streatch. It’s like saying that an athlete specialized in 100m is equally capable to reach the same results in 1000m and marathon just because he is athlete and has to do with running.
Precisely for this reason when casters discuss some game between two players (TvZ) they try to see how is the vs. Terran level of the Zerg and vs.Zerg of the Terran.
Adding to my obliteration of your points, here is a direct comparison of the mirror matchups, which proves beyond doubt they measure raw skill with no bias:
https://i.imgur.com/BfYgDVk.png
Nope, perfect isn’t a long stretch. On average, P/T mirror elo is 1 to 0.98 meaning it’s a near perfect indicator (it’s off by 2%). On an individual player basis, the variance means it’s imperfect up to 10-12% but as an average of the group it’s basically perfect.