Reflect Magic

Is reflect magic nerfed? It hasnt worked all round long. So far I ve reflected like 2 cbs and that it. Anyone else having any luck with RM? I have over 1.5 wpa

ive rmed MS and MV a few times, though ive also had an fb run get through 10 times in a row :(

it's random, I think 20-25% chance of a magic reflect?

Quote:

---

Originally Posted by WhenTheLeveeBreaks

Is reflect magic nerfed? It hasnt worked all round long. So far I ve reflected like 2 cbs

---

Stopped reading there

wrong forum

The number of spells reflected for n attempts is a binomial random variable.

Let X be the number of successes (ie, reflections) and p be the probability of a success, then

P(X=x) = c(n,x)*(p^(x))*((1-p)^(n-x)), where c(n,x) = n!/(x!*(n-x)!)

Get a good sample (say, n=100) and count the number of successes you have (say, N successes).

If P(X<=N) <= 0.05, then you have cause for worries.

Else, you cannot conclude that RM is nerfed at 95% confidence.

well.. first of according to popular belief its not really a binom distribution. (the random feed doesnt seem to be very random they say. well it isnt, question is if its random enough or not) but lets for simplicity assume that it is.. what i dont get is why youre not proposing the use of normal distribution since the bin your talking about will be approximately normal distributed anyway. (Bin(n,p) ~ N (np, V(npq)) if np(1-p) > 10)

binom calcs for n as large as 100 is a true pain in the ass if we gotta do them by hand :)

Or did i already forget all the stats from the stat course i never finished last year? =)

Now does RM work against all spells or just those that get thru your wpa. The question is does RM reflect all spells cast on you or does it just attempt to reflect spells that get thru your wpa. The numbers would make sense(and makes the spell almost useless) if it only tries to reflect successful spells.

>well.. first of according to popular belief its not really a binom distribution. (the random feed doesnt seem to be very random they say. well it isnt, question is if its random enough or not) but lets for simplicity assume that it is.. what i dont get is why youre not proposing the use of normal distribution since the bin your talking about will be approximately normal distributed anyway. (Bin(n,p) ~ N (np, V(npq)) if np(1-p) > 10)<

The test for randomness to see how close the random number generator is to the uniform distribution can be done separately assuming that it is the problem.

However, I was assuming that the random number generator was fine and that the desired test was whether p is below the presumed value.

>Or did i already forget all the stats from the stat course i never finished last year? =)<

No, you're right.

A normal distribution can be used as an approximation.

However, 100 trials is not so large that it can't be computed as a binomial distribution provided that you code the algorithm rather than do all the calculations manually.

I was assuming that I was talking to someone that has some programming brackground, but 0 background in stats so the normal distribution was a no go.

>provided that you code the algorithm rather than do all the calculations manually<

Now that I think about it, 0.05 is not a lot.

My guess is that you wouldn't have that much computation do to to compute the cumulative distribution up to 0.05 and stop.

So a programming background might not even be necessary.

>Now does RM work against all spells or just those that get thru your wpa. The question is does RM reflect all spells cast on you or does it just attempt to reflect spells that get thru your wpa. The numbers would make sense(and makes the spell almost useless) if it only tries to reflect successful spells.<

Thats a good question, but it wouldn't be that hard to test.

Just do the experiment I outlined above twice.

First time, consider all spell attempts as a try.

Second time, consider only the spells that were reflected or that went through as tries.

RM only reflects successful casts. That is why you lose honor when you reflect a spell, as well.

Good answer by Zauper.

Anyway, I think this thread should be made in the Q&As section, not bugs & suggs.

However, 100 trials is not so large that it can't be computed as a binomial distribution provided that you code the algorithm rather than do all the calculations manually.
Ofc not, nor would 10000 be if you let computers do the trick.. but the N approx is there for a reason for large n :)

either way its nice to be able to manually double check your results to see that your alg was right in first place.

>Ofc not, nor would 10000 be if you let computers do the trick.. but the N approx is there for a reason for large n :)<

Actually, my hand calculator won't give me an answer for 10000!

That's why you have the normal approximation.

/me chants "Math Fight!!!"

nor is your hand calculator a computer, is it?

>nor is your hand calculator a computer, is it?<

Yes it is. Just a very weak one.

I tried on the windows calculator and it took about 1 second to give me an answer.

The quantile of 0.05 for binomilal(0.2,10000) is 1934.

Thats roughtly 1934 seconds of waiting which is a fair amount of time (about 30 minutes).

With 100 000!, I gave up after 30 seconds of waiting.

The time factor increases very quickly.

just give it up..

Twisted way to put your computer on the knees, using RM as an excuse.

I believe RM reflects all spell attempts - not just successful casts.

My news page shows both "A spell attempt was reflected..." and " A Meteor Shower was reflected upon the sender!"

If RM only reflected succesful casts, then the the name of the spell casted would always be shown.