resist spell rate bugged - Phoenix Freeshard

the spell resist rate is everything except 10% percent on this server, I dont know what formula you use or is this on purpose to boost stealthers/tanks hegemony but what are the odds to have 4 resist at supposedly "10%" in a row 4-5 times a days in a time span of 2 hours ?

surprized no one is complaining about this non sense.

Fri 8 May 2020 7:57 AM by keen

Well if you want to argue about resist rates like this you need to provide data. Log your casts for the day and compare the result to 10% value.

Fri 8 May 2020 11:44 AM by

gruenesschaf

Post a log. Good luck though. I'd even settle for a log with 2 resists back to back when both spells have a 10% listed resist chance.

Fri 8 May 2020 2:10 PM by Adwaenyth

benny75 wrote: ↑
Fri 8 May 2020 6:45 AM
the spell resist rate is everything except 10% percent on this server, I dont know what formula you use or is this on purpose to boost stealthers/tanks hegemony but what are the odds to have 4 resist at supposedly "10%" in a row 4-5 times a days in a time span of 2 hours ?

surprized no one is complaining about this non sense.

The chance to have 4 resists in a row, with a base resist chance of 10%, at any given moment is 0.01%.

While the odds per se are pretty low, If you do happen to make 10000 casts, the chance of seeing that event at least once is around 66%.

The chance of seeing that event happen twice during that experiment is still around 33%.

The problem with statistics and random events is, that we as human beings are horribly bad at judging them. The only way to reliably test it, is providing data and verifying the resist rate that way.

Fri 8 May 2020 2:16 PM by

gruenesschaf

Adwaenyth wrote: ↑
Fri 8 May 2020 2:10 PM
The chance to have 4 resists in a row, with a base resist chance of 10%, at any given moment is 0.01%.

While the odds per se are pretty low, If you do happen to make 10000 casts, the chance of seeing that event at least once is around 66%.

The chance of seeing that event happen twice during that experiment is still around 33%.

The problem with statistics and random events is, that we as human beings are horribly bad at judging them. The only way to reliably test it, is providing data and verifying the resist rate that way.

Would only apply if there were not streak prevention but there is. It is on this server literally impossible to have 2 resists back to back for a spell with 10% resist chance. Just like it is impossible to have more than 16 resists or less than 10 resists total in any 100 cast sequence at 12.5% resists.

Fri 8 May 2020 2:24 PM by Adwaenyth

gruenesschaf wrote: ↑
Fri 8 May 2020 2:16 PM
Adwaenyth wrote: ↑
Fri 8 May 2020 2:10 PM
The chance to have 4 resists in a row, with a base resist chance of 10%, at any given moment is 0.01%.

While the odds per se are pretty low, If you do happen to make 10000 casts, the chance of seeing that event at least once is around 66%.

The chance of seeing that event happen twice during that experiment is still around 33%.

The problem with statistics and random events is, that we as human beings are horribly bad at judging them. The only way to reliably test it, is providing data and verifying the resist rate that way.

Would only apply if there were not streak prevention but there is. It is on this server literally impossible to have 2 resists back to back for a spell with 10% resist chance. Just like it is impossible to have more than 16 resists or less than 10 resists total in any 100 cast sequence at 12.5% resists.

Now you got me curious. Would be interesting in how you implemented it. Would you be willing to share that part of the code?

Fri 8 May 2020 2:41 PM by Sepplord

i can't find it right now, but gruenes has already explained their RNG systems somewhere here on the forums (long time ago though)
so if you really want to do the digging, at least you know that the answer is out there somewhere

Iirr it has something to do with defined "sets" that the server goes through, depending on your resistrate.

So, for example, a 25% resistrate could be coded to always go through this sequence "success, succes, success, resist" and then repeat. That wouldn't be random though and too predictable. So you could make it better by going "success, succes, success, resist, success, succes, resist, success, succes, resist,success, succes" which would be a bit less noticeable but still very far away from randomness.
Now imagine putting in chains like "success, resist, succes, resist" for bad luck and "success, success, success, success" for good luck.

This is VERY simplified, the "chains" are MUCH longer, and by taking a random starting point and/or changing it by different triggers there is no way to find out the chain and predict what will happen next. So you get really close to randomness at least in playerfeeling and can at the same time make sure that something like "resist, resist" doesn't exist anywhere in the chain for a certain resist-%

But again, this is overly simplified and is just a short overview into the general method. If it interest you you should dig up gruenes posts about it, i remember that it was a quite intresting read.

Fri 8 May 2020 2:45 PM by Adwaenyth

Sepplord wrote: ↑
Fri 8 May 2020 2:41 PM
But again, this is overly simplified and is just a short overview into the general method. If it interest you you should dig up gruenes posts about it, i remember that it was a quite intresting read.

Might do that. It'd be quite an interesting read in how to prevent predictable results while also lessening the effects true randomness could have. Thanks for the info.

Fri 8 May 2020 3:14 PM by

gruenesschaf

Adwaenyth wrote: ↑
Fri 8 May 2020 2:24 PM
Now you got me curious. Would be interesting in how you implemented it. Would you be willing to share that part of the code?

We have different types of rng, spell resists and defensive combat results use a large list of results that is perfectly deterministic but large enough that it feels random while being curated to reduce streaks (or avoid them entirely for low chances). The list usually contains 1000 or 10000 results, depending on the required resolution, and is filled with numbers from 1 - 1000/10000 and each rng check is just next++ <= chance basically, that means over 1000 (or 10000) events the listed chance is 100% accurate, e. g. at 10% exactly 10% of the attempts the event will have happened. There are then some analyzers to ensure certain properties are met for any given chance like occurences within 50 or 100 attempts being within some bounds or streaks etc.

Basically all positive things like mp crafting, procs, crits and multi hits (svg, cd, dw) use prng

Things that don't have anything to do with combat, with the exception of sm pet intercepts, use normal rng.

Ah right, group loot is 100% fair round robin per item type, e. g. trash loot like opal or animal skins would be one type, collection task items are one type, rog weapons are one type, rog armor is one, rog accessories is one, df seals would be 1 for each kind of seal.

Fri 8 May 2020 5:17 PM by Adwaenyth

Thanks for the info. So each character gets a random entry point assigned for each ability type on login I guess, or are they rotated while playing too?

Fri 8 May 2020 5:24 PM by

gruenesschaf

Adwaenyth wrote: ↑
Fri 8 May 2020 5:17 PM
Thanks for the info. So each character gets a random entry point assigned for each ability type on login I guess, or are they rotated while playing too?

Depends but it's usually not per ability but per character / weapon hand / item and some things may or may not reset on release or zoning. For example spell resist is per character, crafting failure streak per recipe and proc failure streak per proc (casted procs like cleric heal proc) or item slot (weapon procs, armor procs).

Thu 14 May 2020 3:38 PM by benny75

gruenesschaf wrote: ↑
Fri 8 May 2020 2:16 PM
Adwaenyth wrote: ↑
Fri 8 May 2020 2:10 PM
The chance to have 4 resists in a row, with a base resist chance of 10%, at any given moment is 0.01%.

While the odds per se are pretty low, If you do happen to make 10000 casts, the chance of seeing that event at least once is around 66%.

The chance of seeing that event happen twice during that experiment is still around 33%.

The problem with statistics and random events is, that we as human beings are horribly bad at judging them. The only way to reliably test it, is providing data and verifying the resist rate that way.

Would only apply if there were not streak prevention but there is. It is on this server literally impossible to have 2 resists back to back for a spell with 10% resist chance. Just like it is impossible to have more than 16 resists or less than 10 resists total in any 100 cast sequence at 12.5% resists.

do you mind sharing the code formula related to the resist ? because I assure you I had multiple resist on a spell at 10 percent on the same target several time

Fri 15 May 2020 6:39 AM by Sepplord

benny75 wrote: ↑
Thu 14 May 2020 3:38 PM
do you mind sharing the code formula related to the resist ? because I assure you I had multiple resist on a spell at 10 percent on the same target several time

you know what would be REALLY assuring...a screenshot

the technique to make it impossible was already described here, what else do you need? If it is bugged, screenshot it and they will look into it