I agree slightly. JR is a fucking awesome Akuma, he’d beat me senseless any day. However, his non-vortex play-style might not play as well against Cammy, know what I mean? This isn’t even a criticism on his play because he’s great at what he does, but I don’t think I’ve seen him tatsu-sweep ONCE, lol.
The error you have committed is that when you decrease your sample size you increase your error substantially. Also you can not just throw out data points as you have without statistical justification. So yes under your calculations (and for the moment let’s just go by your numbers) it is 58%, the problem is a group of 17 has a rather high error range associated with it.
Let me give you an example. Let’s flip a coin 100 times. Let’s say the first 7 flips are all heads. According to the first 7 flips we have a 100% chance of obtaining heads right? The problem with this assumption is that we have a small sample size of 7 as a result the expected error associated with that sample size is expected to be high because by random chance it is possible to get 7 heads in a row.
Where as out of 100 throws we expect a more likely average of 50% with a deviation of +/- 5. Applying the 95% confidence intervals we can see that we would expect the true population to be within +/-15 or in other words we would expect to get between 35 to 65 heads out of a group of 100, 95% of the time.
So where as I am trying to increase my match up instances (and there by decreasing effect of random variables such as luck,nerves, human factors, and yes even online factors) you are decreasing your instances and therefore increasing the effect of random variables.
I’ll discuss this more later when I don’t have a take home final to deal with.
This is why I suggested to remove online play as it has a significant statistical variance vs offline play. Collect enough offline data to reduce the probability of variance. As you can clearly see there is a pronounced difference between online and offline results in your data pool.
You haven’t proven anything. You statistically had no justification to remove that many data points other than your own opinion. In fact you haven’t looked at variance at all of online vs offline which is what you really need to prove two groups are different. You just calculated the ratio of two groups and called it good. If you tried that in any major science you would be laughed at and have your data discarded.
As in my coin example, 7 heads out of 7 throws doesn’t mean the 7 throws aren’t part of the larger population of a 50% average. You can a 50/50 distribution and still have streaks of 7 heads in a row.
If you look at stats with a cherry picking mentality you will always get the wrong answer. The more you narrow your data the less you can conclude.
Well I said that they were good, strong players but compared to other known Akumas I wouldnt consider them to be at the top of their game. I just dont feel they represent high level Akuma play.
And you had no reason to include them other than your own opinion they were no different. Works both ways.
Hypocrisy doesn’t get you far. You don’t get to do whatever you like and then turn around a chastise somebody else who gives a different perspective on what data should be used.
I already gave the solution. Collect and calculate online and offline separately. It’s not like we don’t know for a fact that online is different than offline. If you don’t want to do that it’s fine, but don’t try to pretend there isn’t a difference.
And that is where you are wrong. Unlike you I actually did the math. Both values were within each other’s confidence interval of the t-test and the F-Test concluded there was no difference. So in other words, my inclusion of them was not based off opinion it was based off the fact that there was no statistical proof that says I couldn’t.
I’VE STATED THIS FACT ABOUT 20 TIMES NOW AND YOU CONTINUE TO IGNORE IT
I even re-ran the tests using your numbers and it showed there wasn’t enough of a difference to separate the two.
Please serious go learn some stats before you come back here again. Because right now you really look ignorant because so far you clearly have shown you have no understanding of statistics and are just arguing for the sake of arguing you point.
If anything, the representation of the characters (mentioned earlier in this thread) is much more important than the actual differences between online/offline. Also have to agree with LoyalSol that Japanese online play is pretty much live.
At the end of the day what matters most is your matchup experience. If you just know that much more about Akuma vs Cammy than your opponent, it doesn’t matter which side you’re playing. (i.e. I think anyone giving a matchup ratio vs Gen who has not played say, Amiyu doesn’t actually know anything)
I know my point is obvious, but just want to post my thoughts on the “uselessness” of matchup ratios.
You never calculated offline tourney results separate from online tourney results. If you had you would have noticed that giant 20% discrepancy between the two and we wouldn’t be having this discussion.
Seeing the huge variance between results in your own test pool is common fucking sense. You don’t need a degree in statistics to see the obvious.
Offline tourney = 14 matches with a 57% or 71% win ratio
Online tourney = 17 matches with a 41% win ratio
There is an obvious discrepancy. So the question is why? Well one is online one is not. So now the question is does that make a difference? Well the data pool is too small to be sure, so more data would need to be collected. Once enough data of the two separate categories is collected then we can determine if there is or is not a difference.
This is ignoring things like player caliber which matchup charts are based on.
Thank you for once again proving you don’t have a clue about how statistics work. You are so hung up on averages that you don’t realize those don’t mean jack in the overall picture. When proving two things are different you have to calculate the ERROR terms. Which you clearly don’t understand.
Now I now why Super, Alou, and practically everyone else has given up arguing with you.
You know what dude I respect you skills at SF, but I don’t agree with much of what you have said. I am going to leave it at that.
I’m not talking about your math, I’m talking about an obvious difference between offline results and online. It’s there, it exists and it’s a fact.
I’m not looking at averages, I’m looking at the factual results.
Offline and online aren’t required to be proven as different because it is a known fact they are different. You aren’t calculating the statistics of a group data that is assumed/believed to be the exact same.
If you really are taking science or statistics you know what I say to be true. The data needs to be collected and calculated separately. I don’t see why you would be against this. If you really believe there are no or very little differences then a separate analysis would only prove that even further.
Alioune came in here and accused me of things I never did and had no idea what he was talking about. Superllolo is a fucking retard who is mad at everybody everywhere.
So if those guys have given up on “arguing” with me (I never actually had an argument with Alioune) then whoopie shit.
You do what you like, but I have pointed out what needs to be done in a logical and concise manner. One doesn’t need to be a statistics major to understand there is a difference between online and offline play. And if there is no difference, the stats will prove it. Whatever the case nothing I said is wrong or illogical.
Once again, thank you for proving you don’t understand anything about statistics and the scientific process. On four separate occasions my analytical chemistry professor, my probability professor, and everyone else in the scientific and math community would have slapped you in the side of the head.
And BTW I am receiving my degree in Math and Chemistry this Saturday. What’s your degree in? So please, do me a favor and don’t talk about the scientific model because you have already proven you don’t understand it.
I will also say you say there is a clear difference between online/offline. In the US/Canada I would agree full heartedly; however, I have played on Japanese internet and it is a whole different world. The average lag is less than a frame.
Perhaps you should drop your ego and actually be interested in the facts.
This little statistical project of yours has been shown as nothing more than a way for you to support some agenda rather than being a search for the facts.
If you were interested in the facts you would have had no problem collecting and calculating the data separately. Instead we get this farce.
Well you clearly aren’t getting one in common sense or honesty.
Your previous comments about having an honest discussion are absolutely ridiculous considering.
The scientific method is certainly not about taking two different sets of data and lumping them together and proclaiming they are the same, unless you wanted to influence the results.
If you actually believe in the scientific method, you haven’t shown it.
SSF4 has a minimum 2 frame delay. Yes the internet in Japan is awesome, but it certainly isn’t instantaneous.
Guys, please try to get back on topic.
Casuals don’t mean jack shit. If I beat someone in tourney and they have an ego, they will challenge me to some casuals after the tourney and I will try about 10% to beat them. The motivation to win just isn’t there in casuals.
Y’all still at it? That’s a damn shame.
Strange that is what I have been telling you this whole time. How about you drop your ego and actually listen when I tell you that your math is bad math. You are currently the equivalent to a scrub who comes onto SRK to tell everyone that full screen demons work.
And maybe if you actually understood what you were talking about then maybe you would realize that the scientific method isn’t what you are twisting it to be.
Seriously all it takes is one google search on Null Hypothesis to show that I am following exactly the same method that researchers use. You assume two populations are statistically the same and try to prove otherwise. Well guess what,
the F-Test
Stats: F-Test
and the T-Test
Student’s t-Tests
Both say that as far as match up data goes the populations are not statistically different enough to conclude that there is any difference. This is because they are within each other’s 95% confidence interval.
Here since you are so anal about this
This is from your data so that way you can’t whine about that.
Mean online: 38%
Mean offline: 57%
F-Test:
We take the two calculated variances of the populations in question (Online vs Offline) and take the ratio of the variances. We then compare the two variances to a critical F value. If the critical F value is less than the ratio then the two populations are not distributed in the same way which would indicate that they do not represent the same data
Variance = 10p(1-p)
This variance comes from a binomial distribution.
Variance offline (14 matches/ 8 wins) = 2.44898
Variance online(21 matches/8 wins) = 2.358227
Ratio: = 1.04
Critical F from Table=2.20
Therefore the Critical F is greater than the Ratio which means there is a 95% chance the two populations are distributed in the same way. Thus test one is conclusive that there is a reasonable chance that the populations are no different.
T Test:
t-Test for Two Independent Samples
This test determines if the two averages could reasonably be within the same population AKA are the same populations.
STDev offline= 1.564922
STDev online=1.535668
Mean offline +/- 1STDev = 4.149 to 7.279
Mean online +/- 1STDev = 2.273 to 5.345
1 STDev contains about 50% of the population. Two STDev for a binomial contains 90%. Thus
Mean offline +/- 1STDev = 2.584 to 8.844
Mean online +/- 1STDev = 0.738 to 6.880
Thus we are 90% confident that the two can be of the same population since both lie within each other’s interval of confidence.
So by the F-Test and t-test (LIKE I TOLD YOU) the two populations show no statistical difference.
And if you want to bitch and complain about this then you are giving the middle finger to the scientific community.
Now as I said, shut up you don’t know what you are talking about. I did these tests before and I told you that I had done them yet you went on with your ignorant rants.
question…
i’m still wondering at the rose matchup.
1/2 of the time watching tokido/infiltration match none of his rose opponent barely slide at them.
but when i played against my opponent rose. their plan is as follows.
slide backdash
slide throw
slide ex drill
slide cmp, drill
i tried to footsie out of slide range but i feel that’s asking a bit too much. or should i just stay close to the slide range and sweep before she slides?
Dude stop it. Your dishonesty is sickening. Anybody can see there is a difference in Akuma’s win ratio online vs offline. We know online and offline is different. It’s obvious the data should be collected separate.
Your project was never done in good faith and thus is a huge waste of time.
I just try to stay away from her slide sweet spot. I’m not sure if Tokido has any anti slide strats.
Hopeless. You didn’t understand a single thing I said and the best part was I actually quoted credible colleges and scientific methods.
What you fail to understand is that differences in venue to not necessarily translate into a change in the match up. It could be 4-6 online and offline.
I guess you don’t want to listen.