User Pass
Home Sign Up Contact Log In
Forum > Discuss GLB Issues With Catch22 > NOT A BUG - CLOSED > Incorrect display of training value? - NOT A BUG - CLOSED
Stray Doug
offline
Link
 
Training a skill at 81.5 with 2 enhancements:

L - 2
N - 4
I - 6
M2 - 6.3
M3 - 7.2
M4 - 7.8

Where base = the "old normal"... that means:

M2 - 6.3 = base * 1.2 (enhancments) * 1.05 (Multi-2) * 1.1 (intense)... base = 4.54
M3 - 7.2 = base * 1.2 (enhancments) * 1.20 (Multi-3) * 1.1 (intense)... base = 4.54
M4 - 7.8 = base * 1.2 (enhancments) * 1.30 (Multi-4) * 1.1 (intense)... base = 4.54

At least the answers are consistent! That's a good sign that we've done the math correctly so far.

Now OK, so that should mean that in reality:

L - 4.54 (base) * 1.2 (enhancments) * 0.40 (light) = 2.18
N - 4.54 (base) * 1.2 (enhancments) * 0.85 (normal) = 4.64
I - 4.54 (base) * 1.2 (enhancments) * 1.10 (intense) = 6

This is fine except that "normal" displays as 4%, but it seems it should clearly be rounding up from 4.64 and displaying as 5% instead.
Edited by Catch22 on Aug 20, 2010 13:46:49
Edited by VolBrian on Aug 20, 2010 09:54:35 (I hate math)
 
Catch22
offline
Link
 
the displays are correct, you're just doing math with two decimal places instead of the many decimal places that is actually done in the actual calcs. For example 6.3 could actually be 6.2508834 and N = 4 could actually be N = 3.55302.
 
tragula
title
offline
Link
 
Originally posted by Catch22
the displays are correct, you're just doing math with two decimal places instead of the many decimal places that is actually done in the actual calcs. For example 6.3 could actually be 6.2508834 and N = 4 could actually be N = 3.55302.


Is that a confirmed fact ?
 
Stray Doug
offline
Link
 
I am doing the math with more than 2 but was just displaying as such for brevity.

In order for normal training ("N") to display as 4%, that means N < 4.499999 (so that it's rounding down).

Therefore, the base ("old normal") is < 4.411764... (where base * 0.85 * 1.2 < 4.499999)

Any base value greater than 4.411764 would yield 4.5 or more... which would mean N = 4 was inaccurate.

So using this maximum base value of 4.411764 and multiplying by 1.2 for enhancements, then by 1.1 for intense, and then by the appropriate multi-train bonus would yield:

M2 < 6.114705
M3 < 6.988234
M4 < 7.570587

However the values for M2/3/4 that are displayed are all greater than this (even with rounding), and therefore they cannot possibly be derived for any value of N < 4.499999. So the N value and the multi-values appear to be inconsistent. Because the multi-values are displayed to a greater level of precision that the normal training values, I was inclined to believe that they were more accurate and the normal value was inaccurate.

The only other possibility is that the formulas do not work as previously understood (with all modifiers being independently multiplicative).
Edited by Stray Doug on Aug 20, 2010 09:56:47
 
VolBrian
Rocky Top
offline
Link
 
Math.

Marking admin response
 
Stray Doug
offline
Link
 
One other way of looking at this is to say that intense training is 129.412% of normal training (1.1/0.85).

Multi-4-way is 130% of intense... or 168.235% of normal.

Therefore, if normal is 4.5, multi-4-way is 7.5706, which would round to 7.6.

However, since multi-4-way is 7.8 (and if we assume that value to be correct), then normal must therefore be greater than 4.5, and thus should not round to 4.

This method removes training enhancements from the calculation to keep things simpler, as well as the need to use the "old normal" base value as a translator, but yields the same result.
Edited by Stray Doug on Aug 20, 2010 10:06:24
Edited by Stray Doug on Aug 20, 2010 10:05:06
 
Catch22
offline
Link
 
It would also round up Doug - 3.5000001 would equal N of 4. I've seen the code that Bort uses for these calcs and am confident that the calcs are being done correctly (it's using all the correct values for light, normal, intense, multi, and enhanced training). The calcs also do rounding more then once fwiw.
Edited by Catch22 on Aug 20, 2010 10:10:59
Edited by Catch22 on Aug 20, 2010 10:09:58
 
Stray Doug
offline
Link
 
Oh... yeah if you're doing the calculation as multiple steps with rounding along the way, then that could be what's throwing things off. I understand that 3.5000001 would round up to 4... my point was that for the given multi-train values that N would have to be greater than 4.5000001... thus no scenario where it could round to 4 (up or down).

I'll see if I can figure out where the intermediate rounding is occurring for this set of values to make sense.
 
Stray Doug
offline
Link
 
OK, here's a method that works...

Base ("old normal") * training type (0.4/0.85/1.1) = X

Round X to nearest integer

Then multiple X by enhancements (1.2) and round again.

Then multiply by any multi-train bonuses... no rounding.

Why there would be intermediate rounding is beyond me (and is a bit deceptive IMO compared to how it was described to have worked), but this works in the case above for a base value of 4.1 (meaning the "old normal" for a skill was 4.1%, which is certainly possible... 81 was around the level where normal training used to drop from 4% to 3%, but maybe I'm just before that threshold, and maybe the 4% was really 4.1% before we just couldn't see it.
 
Stray Doug
offline
Link
 
Just to make a point here... the intermediate rounding is, IMO, a very poor choice.

Just in theory, if my base value is 4 and I had five "10%" bonuses, and was told the bonuses were multiplicative, I would expect to get 4 * 1.1 * 1.1 * 1.1 * 1.1 * 1.1 = ~6.44

Instead under this rounding scheme I'd get 4 * 1.1 = 4 * 1.1 = 4 * 1.1 = 4 * 1.1 = 4 * 1.1 = 4. That's absurd.

Furthermore, I'd point out that using this intermediate rounding scheme, for INTENSE training (including multi)...

- Level 1 enhancements provide NO benefit when training a skill over 82.

- Level 2 enhancements provide the SAME benefit as Level 1 enhancements when training a skill between ~66-82.

- Level 3 enhancements provide the SAME benefit as Level 2 enhancements when training a skill ~62-66 or over 82.


Also, it should be noted that on LIGHT training Level 1 enhancements provide NO benefit when training a skill over 50.


If all of these outcomes are intended, which I find hard to believe, it has been very poorly communicated to the users imo. I think people expect that any time they spend BTs on a training enhancement, that it will always provide SOME benefit (~10% per level in fact!) than if they hadn't purchased the enhancement.
 
Catch22
offline
Link
 
I think you're reading a bit too much into this Doug and the gains have always been approximate gains, not exact gains. I'm going to lock this thread, but I'll give a bit more of an explanation on the calculations before I do.

It's rounded twice. Once to determine base training value (the one that is usually displayed without any enhancements or multi training), and once after all calculations have been done (if there are enhancements or multi training). So if base value is .038994 it's rounded to .04 then that value is multiplied by enhancements (let's say 1.2) and multi training bonus (let's say 1.3) .04*1.2*1.3= .0624 - this value is then rounded (probably to .0625).

 


You are not logged in. Please log in if you want to post a reply.