Spell resists are too high IMO...this time with proof.

[Nuncio]

5 casts in a row resisted has a 20% chance, because every cast has a 20% chance. :\
You can't count chance to resist a spell based on your previous success. The chance of me flipping heads because I flipped 4 previous tails does not change. It's still 50%.
Unless I'm missing something.

 

[Tempus]

but over a long enough time period you should see close to 50/50 like if you flipped a coin 100 times. The problem is people use to small of a sample to truly get a clear indication of what the resist rates are.

 

[jhessal]

but over a long enough time period you should see close to 50/50 like if you flipped a coin 100 times. The problem is people use to small of a sample to truly get a clear indication of what the resist rates are.

I wouldent consider 194 casts a small sample, really. I havent played a caster since my character was bugged( have re-rolled with my fiance and her sister) but once they get into their mid 40's or so, ill be asking the enchanter to log an evenings worth of xp'ing and ill be going over that and post my results here as well.

Its not just me who thinks the spell resists are a bit on the high side. Theres a difference between fun and challenging gameplay (good ai and encounters) and a lot of frustration (25% resist rate on blue xp mobs).

 

[calaran]

Hmm... yeah, I guess it's a fallacy of sorts... Here's the way I was thinking of it.


Flip a coin 10 times.

Because there are 2 possibilities per flip, and 10 flips, that means there are 2^10 possible permutations of the result set.

That gives us 1024 possible outcome sets. This means that the chance of flipping 10 heads is 1 in 1024. The probability of flipping any other combination is ALSO 1 in 1024.

So, for our example, the chance that any single spell may be resisted is 1 in 5 or 20%. If you've been resisted 4 times thus far, the probability of the 5th casting being resisted is still 20%, and that's what you're arguing. What I'm saying is, when a spell is cast 5 consecutive times, the probability of the result set being 5 resists is MUCH different than the probability that the 5th spell will be resisted, based upon prior resists (since they don't factor in).

At a 50% resist rate, the chance that 5 casts in a row would be resisted is 1 in 32. At a 20% resist rate, the chance of 5 consecutive resists SHOULD be 1 in 3125.

Now, given the fact that the random number generator is not random, this math will not likely stand up to analysis against an experiment using a spell against a mob where the KNOWN resist rate SHOULD be 20%... But, this is the point that was being made... the statistical probability of 5 resists in a row is significantly lower than the statistical probability that any one spell should be resisted.


::edit:: Just for mathematical accuracy, and to preempt someone trying to invalidate the whole post based upon this fact, I will point out something that does not impact the numbers I posted, but should be included to make the argument complete. The 1 in 1024 chance to flip 10 heads is correct. There is a 1 in 1024 chance to flip a heads the first time and then 9 tails. There is also a chance to flip a tails, then a heads, then 8 more tails. In my equations, I treated these as separate result sets though for this discussion, they would be combined... since the question would not be "How likely is it you get resisted on the second casting but not the 1st, 3rd, 4th and 5th" but rather "How likely is it you get resisted 1 out of 5 castings?" Since we are discussing the case where all results are the same, ordering does not matter, and the % chances for absolute resistance (5 of 5 times) or absolute success (0 of 5 times) is still the exact same number I gave originally. There is a significantly higher chance than 1/3125 that you'll be resisted 1 time, 2 times, 3 times or 4 times (each with their own distinct probability). ::edit::

 

[Nuncio]

That gives us 1024 possible outcome sets. This means that the chance of flipping 10 heads is 1 in 1024. The probability of flipping any other combination is ALSO 1 in 1024.

That means flipping 10 heads is as likely as flipping 5 heads and 5 tails. Which means, you are as likely to do 10 heads in a row as anything else. :\

But yes, I know what you mean.

 

[calaran]

That gives us 1024 possible outcome sets. This means that the chance of flipping 10 heads is 1 in 1024. The probability of flipping any other combination is ALSO 1 in 1024.

That means flipping 10 heads is as likely as flipping 5 heads and 5 tails. Which means, you are as likely to do 10 heads in a row as anything else. :\

But yes, I know what you mean.

I agree 100%. That probability is 1 in 1024. This was the quote I was debunking:

5 casts in a row resisted has a 20% chance, because every cast has a 20% chance. :\

5 resist casts in a row has a 1 in 3125 chance or 0.032% chance of happening. It's also true to say that having a resist, a success, then 3 more resists has a 0.032% chance of happening.

The point being (I believe) that much more frequently than once every 3,125 battles, a single spell is resisted 5 times out of 5 times. That means the randomness is not accurate and it seems to be signifcantly flawed.

And I guess I should add that I've not logged/parsed any resists, and they've not seemed out of whack to me at all yet. I'm just defending the mathematics because there are way too many cases where the wrong logic or principle is applied and then used to defend/attack an idea...

 

[Nuncio]

OK. So any combination of resists in a 5 cast scenario has a .032% chance of happening...

stick resist resist resist resist
resist stick resist resist resist
resist resist stick resist resist
resist resist resist stick resist
resist resist resist resist stick

So here's all of the combos with one stick.

The problem, is that these all have the same outcome, an 80% resist rate

So we really only have 6 total outcomes. no resist, 1 resist, 2 resists, 3 resists, 4 resists, or 5 resists.

Sorry, I'm bad at math. Does this change how we look at this, since .032% X 6 is quite a bit less than 100%...

If I'm being dumb, just let me know. I didn't take statistics in school

 

[calaran]

OK. So any combination of resists in a 5 cast scenario has a .032% chance of happening...

stick resist resist resist resist
resist stick resist resist resist
resist resist stick resist resist
resist resist resist stick resist
resist resist resist resist stick

So here's all of the combos with one stick.

The problem, is that these all have the same outcome, an 80% resist rate

So we really only have 6 total outcomes. no resist, 1 resist, 2 resists, 3 resists, 4 resists, or 5 resists.

Sorry, I'm bad at math. Does this change how we look at this, since .032% X 6 is quite a bit less than 100%...

If I'm being dumb, just let me know. I didn't take statistics in school

Heh, you're not being dumb... just incorrect or mislead (sorry if that sounds pompous)... my note a couple posts back, I added a chunk at the end, refuting this argument, but here it is again, since it was sorta unfair for me to go back an edit it.

Just for mathematical accuracy, and to preempt someone trying to invalidate the whole post based upon this fact, I will point out something that does not impact the numbers I posted, but should be included to make the argument complete. The 1 in 1024 chance to flip 10 heads is correct. There is a 1 in 1024 chance to flip a heads the first time and then 9 tails. There is also a chance to flip a tails, then a heads, then 8 more tails. In my equations, I treated these as separate result sets though for this discussion, they would be combined... since the question would not be "How likely is it you get resisted on the second casting but not the 1st, 3rd, 4th and 5th" but rather "How likely is it you get resisted 1 out of 5 castings?" Since we are discussing the case where all results are the same, ordering does not matter, and the % chances for absolute resistance (5 of 5 times) or absolute success (0 of 5 times) is still the exact same number I gave originally. There is a significantly higher chance than 1/3125 that you'll be resisted 1 time, 2 times, 3 times or 4 times (each with their own distinct probability).

If that doesn't make sense, feel free to ask and I'll gladly explain it further. I did happen to take statistics in school, and a math minor, but I don't feel that excludes me from making mistakes... I just believe in this case I am correct.

 

[Birlic]

Probabilities can be counter-intuitive sometimes, Nuncio. Luckily, not much reading is needed to figure out probabilities like the one in the debate above. Developing an intuition about the binomial distribution may be useful in many occasions in life, especially in an age as dominated by computers as today.
If you find the link above not very readable, you can try to start with an introduction to probabilities and reading about the Bernoulli distributions.

 

[Birlic]

but over a long enough time period you should see close to 50/50 like if you flipped a coin 100 times. The problem is people use to small of a sample to truly get a clear indication of what the resist rates are.

The number of trials needed to make a claim about the resist rate depends on the underlying distribution and the actual, "true", resist rate which you're trying to measure. For a binomial distribution as the one debated here, you can get a good guess (within 5%) with ~50 trials. At 194 trials (jhessal's experiment) you get close to a "statistical proof" (i.e., 95% confidence interval, which is what most surveys and statistical tests use).
Here's a confidence interval calculator for anyone who wishes to do such experiments in the future.

 

[Grachnist]

I have writen up a little program for my TI-83 Calculator to calculate if you rolled 1-100 5000 times how many would land at 20 or below, also how many times 20 or lower is rolled in a row. Did this to get some hard numbers if the resist rate was infact a solid 20%. So lets say that if the resist rate is 20% then any number 20 or lower in a random roll of 1-100 would count as a resist. Any number 21 or higher would could as a spell that landed.

After 1000 rolls of 1-100:

Total of 214 Resists.
Resisted 1 in a row: 165 Times
Resisted 2 in a row: 41 Times
Resisted 3 in a row: 8 Times
Resisted 4 in a row: 0 Times
Resisted 5 in a row: 0 Times
Resisted 6 in a row: 0 Times
Resisted 7 in a row: 0 Times
Resisted 8 in a row: 0 Times
Resisted 9 in a row: 0 Times
Resisted 10 in a row: 0 Times

After 2000 rolls of 1-100:

Total of 379 Resists.
Resisted 1 in a row: 297 Times
Resisted 2 in a row: 66 Times
Resisted 3 in a row: 15 Times
Resisted 4 in a row: 1 Times
Resisted 5 in a row: 0 Times
Resisted 6 in a row: 0 Times
Resisted 7 in a row: 0 Times
Resisted 8 in a row: 0 Times
Resisted 9 in a row: 0 Times
Resisted 10 in a row: 0 Times

After 3000 rolls of 1-100:

Total of 576 Resists.
Resisted 1 in a row: 453 Times
Resisted 2 in a row: 98 Times
Resisted 3 in a row: 24 Times
Resisted 4 in a row: 1 Times
Resisted 5 in a row: 0 Times
Resisted 6 in a row: 0 Times
Resisted 7 in a row: 0 Times
Resisted 8 in a row: 0 Times
Resisted 9 in a row: 0 Times
Resisted 10 in a row: 0 Times

After 4000 rolls of 1-100:

Total of 776 Resists.
Resisted 1 in a row: 626 Times
Resisted 2 in a row: 121 Times
Resisted 3 in a row: 25 Times
Resisted 4 in a row: 3 Times
Resisted 5 in a row: 1 Times
Resisted 6 in a row: 0 Times
Resisted 7 in a row: 0 Times
Resisted 8 in a row: 0 Times
Resisted 9 in a row: 0 Times
Resisted 10 in a row: 0 Times

After 5000 rolls of 1-100:

Total of 966 Resists.
Resisted 1 in a row: 776 Times
Resisted 2 in a row: 150 Times
Resisted 3 in a row: 36 Times
Resisted 4 in a row: 3 Times
Resisted 5 in a row: 1 Times
Resisted 6 in a row: 0 Times
Resisted 7 in a row: 0 Times
Resisted 8 in a row: 0 Times
Resisted 9 in a row: 0 Times
Resisted 10 in a row: 0 Times


So even after 5000 rolls there were never more then 5 resists in a row, and that only happened once.

Total of 966 Resists = 19.32% of spells were resisted.
Resisted 1 in a row: 776 Times = 15.52% chance to be resisted once then land the next spell.
Resisted 2 in a row: 150 Times = 3% chance to be resisted twice in a row then land the next spell.
Resisted 3 in a row: 36 Times = 0.72% chance to be resisted three times in a row then land the next spell.
Resisted 4 in a row: 3 Times = 0.077% chance to be resisted four times in a row then land the next spell.
Resisted 5 in a row: 1 Times = 0.02% chance to be resisted five times in a row then land the next spell.

Of course these numbers are only viable if the resist rate is a constant 20%, which it is not. But yet these numbers show what resist rates would be IF the resist rate was a solid 20%. So in other words the SoD resist rate it much much greater the 20% if you are getting 6,7,8 or even 18 resists in a row.

 

[calaran]

I'd be curious to see you modify the program, tweaking the resist % until you get your first 18 resist in a row result on a set of 5000 runs. I wonder what resist % it would take to make something like 5-6 resists be rather normal/likely, and 18 casts possible while still generally considered a statistical anomaly.

 

[Birlic]

There are two problems with your experiment:
1. it uses a different random number generator - unless the SoD server is hooked up to a TI calculator to get rand() results, which I doubt
2. it uses a uniform distribution
The outcome of a resist check in SoD is probably the result of a lot of filtering (by level, skill, specialization points) and that may change the distribution if not done carefully. I would suspect that's the cause for people getting many resists in a row. If it were simply the mean/expectation/"resist rate", we'd be getting much higher resist rates in experiments like jhessal's.

 

[calaran]

I think I heard that SoD runs on a TI-8700 or something like that...

 

[Birlic]

I'd be curious to see you modify the program, tweaking the resist % until you get your first 18 resist in a row result on a set of 5000 runs. I wonder what resist % it would take to make something like 5-6 resists be rather normal/likely, and 18 casts possible while still generally considered a statistical anomaly.

Don't need an experiment to find this out.
P(18 resists in a row) = 1/5000 implies P(resist) = (1/5000)^(1/18) = 0.62302
Kinda high, which is why I suspect the distribution assumption or the random number generator are at fault (see above).

P(resist) = 0.62302 implies P(5 resists in a row) = 0.62302 ^ 5 = 0.09386
P(resist) = 0.62302 implies P(6 resists in a row) = 0.62302 ^ 6 = 0.05848

 

[calaran]

I am so glad you started posting in this thread. It's been so long since I was in college, I have a horrible memory, and never use math in any of my recent/current jobs, so I have gotten rusty and was doing hella research to try to back up my claims...

 

[Grachnist]

Of course SoD doesnt run on a TI calculator, nor does it use a /random 100 to calculate spell resists.

But what my post above does prove is that, just as a percentage, resists on SoD even against Blue mobs is well above 20%.

Edit: Also its not that say 20% of spells are resisted, its that 20% less damage is being done and 20% of your mana gone to waste on average.

So if SoD has a higher resist rate then 20% thats just that much more Dmg not being done, that much more mana being wasted, that much more medding between fights, and that much more time.

 

[Suul]

There are two problems with your experiment:
1. it uses a different random number generator - unless the SoD server is hooked up to a TI calculator to get rand() results, which I doubt
2. it uses a uniform distribution
The outcome of a resist check in SoD is probably the result of a lot of filtering (by level, skill, specialization points) and that may change the distribution if not done carefully. I would suspect that's the cause for people getting many resists in a row. If it were simply the mean/expectation/"resist rate", we'd be getting much higher resist rates in experiments like jhessal's.

SO, If i'm to understand this correctly; maybe 20% resists is the best you could do with CHA maxed, and on a mob that isn't especially resistant. So for a level 30 with 120cha, you're looking at more like 40%, which honestly seems about what it is.

 

[Xeldan]

Posting one more time in this thread to comment that jhessal had a legitimate concern here. My review of Heartland resists is done and should hopefully be in next patch. Some mobs resists were lowered by up to 33% to be more inline with mobs their level in other zones. It's important to note though that this primarily impacted the forest area - the goblins and constructs in the caves saw much lower decreases(or in some cases, none at all - especially the rare named).

 

[jhessal]

Posting one more time in this thread to comment that jhessal had a legitimate concern here. My review of Heartland resists is done and should hopefully be in next patch. Some mobs resists were lowered by up to 33% to be more inline with mobs their level in other zones. It's important to note though that this primarily impacted the forest area - the goblins and constructs in the caves saw much lower decreases(or in some cases, none at all - especially the rare named).

Thanks again Xeldan, much appreciated