Reality of AI dodge, complete statistics

So then if u have the time maybe u can calculate the confidence interval on this data being plus or minus 5% given the sample size.

And that will show how good the data is.

What is with the need more data, maybe 1000, maybe two weeks. If u can calculate the confidence interval u know exactly how many attacks to get a 95% confidence interval. I don't remember how to do that, and I am not going to look it up.

Why should there be any reason to believe in AI advantage devoid of any data that suggests it? When is it good enough? U should be able to say exactly.

And what is wrong with the structure of the data I collected?

When you do analysis like this, you can really only use data from teams with an equivalent power level (apples-to-apples). I know "power" doesn't matter as much as synergy but in-terms of a tangible statistic, power is all we have to gauge the strength of a team + / - 5% or less. In-relation to GW, that's really only the last 5 nodes maybe. You would definitely need to throw out data from the first 6 nodes.

Even within the 12 GW's you did or the 20 arena I suggested, the data is probably not statistically significant and here's why. There's seemingly good evidence to suggest that EA is not calculating Dodge on a simple 2 out of 10 attacks; if they did your Evasion squad stats would be much closer to 15%. If we're honest with ourselves, we've all had arena matches where most of our attacks get dodges, whereas we've also experienced the opposite. These variations reflect a huge standard deviation away from a mean of 15% for Evasion, meaning there is a pretty big range. We would need considerably more simulations to try and get a smaller S and get closer to the mean.

OK, nevermind. I do not doubt u have a masters degree. But i dont think it pertains to hard science or math. Thanks, anyway. Ur exactly the Internet PhD I could care less to hear yapping. Where in this pretentious pontification is any useful insight or concrete suggesting that amounts to more than:
1. lvls might have something to do with it
2. a vague amount of more data would be better
3. It isn't exactly 2 dodges per 10 attacks. As we have all seen, streaks happen. Perhaps there's something called a bell curve to describe when events occur with a static statistical probability. Enough such events are observed, and whadya know. You end up with more or less than the exact amount predicted in a sequential segment of these events. Who'da thunk? When I flip a coin, it doesn't come up heads every other flip?!?!? WhoaAAAAA! AmAAAzing!
4. You never know, because... well you never know. If we get enough data it might be enough to suggest that we might suggest a suggestion. But well, you never know.

U could have just said good job and thank you.

Btw, I just observed it is the TM part of mind tricks that counts as a dodgeable attack. If u miss TM, there is a dodge. If TM lands, there is no dodge proc, even if both negative status effects are resisted. But since O down sticks (in my battles) 80 percent of the time and apparently applies after and overrides dooku O up, i didnt notice this. Tm has to be "dodged" and O up has to be resisted to see this against a dooku L. This is curious since TM from teebo is resisted, not dodged. Maybe in the coding one of the effects has to act like the "attack."

Poe taunt also reduces enemy TM. This is probably why Poe taunt is an attack and can be dodged. If not expose also counting.

I enjoyed this thread. Keep up the good work.