Clearly those were on Islands and Feitoras need a nerf.The only relevant statistic as far as I am concerned.
Clearly those were on Islands and Feitoras need a nerf.The only relevant statistic as far as I am concerned.
I am honestly amazed how you can keep running in circles and ignore everything else but your own voice.
I missed where you explained the statistical test you employed and your p values. Statistical significance is a formal definition (mostly, based on rule of thumb, but probabilities-based) that you cannot throw around without showing your work. Otherwise it looks like someone attempting to yield a stats buzzword poorly.Can you think of a better explanation for why only the 1650+ range saw a statistically significant increase in win % after a pure civilization nerf? I'm using actual data, not conjecture. You go next!
This one's actually hilariously and embarrassingly uninformed to me. Have you ever seen the distribution of elo across all DE players? It's actually a pretty normal distribution. So talking about the raw numbers.... A player ranked #10 on the 1v1 ladder (ad hoc definition of "pro") is in the top 0.02% of all players. A 1700 is around the top 2%. To assert the quantity of data points in terms of games played is even remotely equal here is absolutely laughable. It is literally one of the first things you learn in a stats course, and I know because I've taught several at the college level.Sure it does. High level pro's are the one's playing day in and day out. They in all likelihood make up a greater amount of 1v1 games than joe 1700 since it is their profession.
Hi. Calm down, keep breathing. Does your "brain," or whatever is in there have the simple concept of causality? One random into Inca's doesn't discount that he picked Inca's an entire day to showcase the civ after the nerf. You literally have the stats showing 3x standard deviation from the average picking Inca's. The vod is gone, but you're being dishonest saying a civ represented 3 std dev from the mean is just pure coincidence.JFC Did you even watch the VODs? Do your brain not have the simple concept of causality?
The VODs are where your posted YT videos from. (cant find the Bugum one), proving that he did not pick incas in those games.
I haven't thoroughly read the last 2-3 pages but I'm disappointed that nobody called him out on his poor understanding of statistics.
I missed where you explained the statistical test you employed and your p values. Statistical significance is a formal definition (mostly, based on rule of thumb, but probabilities-based) that you cannot throw around without showing your work. Otherwise it looks like someone attempting to yield a stats buzzword poorly.
This one's actually hilariously and embarrassingly uninformed to me. Have you ever seen the distribution of elo across all DE players? It's actually a pretty normal distribution. So talking about the raw numbers.... A player ranked #10 on the 1v1 ladder (ad hoc definition of "pro") is in the top 0.02% of all players. A 1700 is around the top 2%. To assert the quantity of data points in terms of games played is even remotely equal here is absolutely laughable. It is literally one of the first things you learn in a stats course, and I know because I've taught several at the college level.
Paired with your exclusive reliance on usage rate as a supporting statistic, this whole argument is a generally well-worded but flailing one to support your point.
For one, I would need the source data to evaluate that. And two, yes yes absolutely a score around 2SD vs 3SD is a huge margin in terms of frequencies, likilihoods, interpretation etc. Don't want to go too off topic beyond that.Without any of the condescension -- since you are clearly an expert, does the data indicate that pro players are inordinately playing Inca's greater than other civs, or are you trying to split hairs over 95 vs 99.7%?
For one, I would need the source data to evaluate that. And two, yes yes absolutely a score around 2SD vs 3SD is a huge margin in terms of frequencies, likilihoods, interpretation etc. Don't want to go too off topic beyond that.
wut
You mean that it's not unreasonable to say that it would be artificially increasing their win rate (doesn't matter how much or little, just that it is) because certain pro players are quite clearly playing Inca's more than other civs? Forgive the massive leap in suggesting that it's because they're willingly choosing to do so, to test them out post-nerf -- but none of that seems particularly unreasonable, especially when Hera had an entire pick-Inca civ day post-nerf to showcase them. Not exactly a secret outside of the data set, either!
You may need to sit down before responding. Professor HK's blood pressure is through the roof and we'll have to carefully craft this response and discuss sample size I'm sure to make him feel better about inadvertently spending his time proving my point, less you would prefer to present Professor HK with the probability of all of those events being random, what,someall being 2-3 STD Dev away from the mean.
Please, 95% certainty for one data set -- but that and then some for four? Pray tell thee, what is the standard for most p value tests, .05 no? You'll have to forgive me, I'm not a learned man, but I did find an online article that stated that most authors refer to .05 for statistical significance. I believe this would mean two standard deviations from the mean. I know, this may shock you -- but I have also once consumed an entire beer in less than 2 seconds via a modified paint gun.
As a result, given your stated academic credentials I believe we are equally qualified to continue this discussion.
I'm glad we have another professor here to tell us what's what.
Wait so you saying that out of 1000 games, 14 were Incas? If he'd had an even distribution of games he'd have played every civ around 27 times.... interesting way to go about attempting to skew the statistics...First off, thanks for finally providing some evidence for your theory. Hera indeed does pick Incas occasionally (check at 0:40):
However, I don't think that this has a very big impact on the winrate as Tarsiz already said. On AoENexus, you can download the data from the players, so I did and Hera has played 14 games with Inca's since the third of May when the new patch was introduced. Out of those he won 13, so he had a 92.8 win percentage with the civ. But 13 wins out of almost 1000 (out of which some were randomed as well) doesn't really change the needle for the win percentage, so unless there's more pro's doing the same thing I don't think they're the reason why the win rate is so high.
As a sidenote: Hera also didn't have easy opponents in his games, he played Bugum (2k4), _Davicy (1k6), Liereyy (2k5), MbL (2k4) twice, Yo (2k4) 4 times, TheViper (2k4) twice, Dark (2k1), miguel (2k3) twice and JorDan (2k4).
Hmm, I worded that bad. What I meant was that out of the 1000 games that were played with Inca's in the 1650+ elo group, 13 were from Hera. Hera played 211 games on the ladder on this latest patch, out of which 14 were with Inca's.Wait so you saying that out of 1000 games, 14 were Incas? If he'd had an even distribution of games he'd have played every civ around 27 times.... interesting way to go about attempting to skew the statistics...
All except for Myth...So we are all agreed Incas are in a good spot and the nerf was justified?
not even closeSo we are all agreed Incas are in a good spot
yesand the nerf was justified?