Casting speed formula is wrong

Started 8 Jul 2018
by gruenesschaf
in Resolved Issues
My claim is that the current implemented casting speed influence of dexterity is wrong. I am aware that casting speed is determined by breaking points on live, but Phoenix does not use that and has a formula for calculating the cast speed instead.
So the trend line of casting speed tests from Pendragon can be used if breaking points are not considered.

I did not test mastery of the art. The test is done on Phoenix with mota 0, on Pendragon with 0% casting speed bonus.
I made Pendragon and Phoenix cast speed tests during Alpha. The results can be found here:
https://docs.google.com/spreadsheets/d/1BFbHYz_smxP8KPGoytb4SqQnEOoPqUc8SIKmSqQF4BI/edit?usp=sharing

First sheet compares current Phoenix implementation to a formula that was circulating on VN boards before breaking points were found.
The next three sheets summarize the casting speed tests for three different casts made on Pendragon. Casts were done for 3min 30sec each time, so the total number of casts varies from 50 to 243.
The average trendline out of these three different cast tests is very close the the claimed formula from the VN boards. This leads me to the conclusion that the current cast speed implementation:
[casttime]=[delve]* (1 - (([dex]- 60) * 0,15 + ([dex] - 250) * 0,05) * 0,01)

is wrong and should be replaced by this formula from VN boards (mota needs to be added):

[casttime] = [delve] * (1 - ([dex] - 70)/600))
http://www.ign.com/boards/threads/full-cast-speed-test-for-all-delves-click-here-to-see-how-much-dex-you-need-p.250680778/ post #4

You can see that Phoenix uses a formula that has a kink at 250dex, decreasing the slope and hence cast time if 250dex is exceeded. So dex higher than 250 is worth more than dex below 250dex.
The VN board formula, and the live test data provided show a linear behavior without a kink. Every dex is worth the same amount of cast speed decrease.
I added a graph comparing the difference in casting speed between the two formulas.
You can see that the current implementation leads to too high casting speeds beyond ca 290 dex, leading to up to 5% too high casting speed for 375dex.


I am well aware that this is contradicting a grab bag stating:
"Q: Would you please give more detail as to how dex affects a caster? For instance, I understand that when I have my dex maxed I will cast 25% faster. How does this work incrementally? And will a lurikeen be able to cast faster in the end than another race?

A: From a dex of 50 to a dex of 250, the formula lets you cast 1% faster for each ten points. From a dex of 250 to the maximum possible (which as you know depends on your starting total), your speed increases 1% for every twenty points. So, as you can see, a lurikeen may be able to cast A VERY TINY AMOUNT faster than another race, but just barely, and even then depending on the starting stat. "

Firstly this is just wrong. VN board formula that was based on casting speed tests years back, and my data provided just don't show it. Secondly, your current formula does exactly the opposite. Beyond 250dex the casting speed gain per dex is increased instead of decreased as explained in the Grab bag.


Issue Timeline

8 Jul 2018 gruenesschaf created this issue
2 Aug 2018 gruenesschaf triaged this issue
2 Aug 2018 gruenesschaf changed the status of this issue from new to open
22 Aug 2018 gruenesschaf changed the status of this issue from open to soon
8 Sep 2018 benfalk assigned this issue from Nobody to benfalk
9 Sep 2018 benfalk changed the status of this issue from in_progress to resolved
Mon 27 Aug 2018 1:29 PM by Jokerall
Before implementing anything i'd read both of these threads carefully
https://forums.freddyshouse.com/threads/do-you-want-to-know-your-casting-speed.147575/
https://www.uthgard.net/forum/viewtopic.php?f=9&t=38410&sid=7d6336d9e283f48be3308b8c02b950d1&start=30
Specially the uthgard one, at page 3 they seem to have nailed down the correct formula which should be:
( Spell cast time ) * ( 1 - (( Dex - 60 ) / 600 )) * ( 1 - MotA )
They did alot of tests and this one seems to check out
Mon 27 Aug 2018 6:31 PM by keen
I added a comparison to the claimed casting speed formula from Jokerall in an additional sheet "Casting speed regression and comparison".
Since the normalized casting speed varies for each cast time, I took the mean of all three casting times (2s cast, 3scast, 4sec cast) from my Pendragon tests and ploted this against the two formulas:
VN Board 70dex
[casttime] = [delve] * (1 - ([dex] - 70)/600))
Uthgard Board 60dex
[casttime] = [delve] * (1 - ([dex] - 60)/600))

It can be seen that the VN Board formula matches the trend line of the Pendragon test data better.
In fact the formula should be taken from the trend line from the Pendragon tests, but the one from the VN board matches the Pendragon tests fine enough in my opinion.

[casttime] = [delve] * (-1.63E-03*dex+1.11)
Mon 27 Aug 2018 6:31 PM by keen
https://docs.google.com/spreadsheets/d/1BFbHYz_smxP8KPGoytb4SqQnEOoPqUc8SIKmSqQF4BI/edit#gid=366536747
Sun 9 Sep 2018 10:35 PM by Jokerall
Just curious, which formula did you implement?
Wed 19 Sep 2018 6:06 PM by benfalk
`( Spell cast time ) * ( 1 - (( Dex - 60 ) / 600 )) * ( 1 - MotA )`

Return to Resolved Issues or the latest topics