Impact on health of chipping skills

[wizzszz]

Purpose of this thread:

The contribution of the skills to health has been resolved in this thread.
A few rare unlocks remain to be checked (listed below)

If you have one of this unlocks, please take a complete snapshot of your skills and contact either me, Doer or Jimmy B. via PM here on EF.

Contribution to health:

• 148/1600 = 9.2500% Stamina
• 80/1600 = 5.0000% Strength
• 40/1600 = 2.5000% Agility
• 20/1600 = 1.2500% Intelligence
• 20/1600 = 1.2500% Psyche
• 8/1600 = 0.5000% Commando
• 5/1600 = 0.3125% Courage
• 4/1600 = 0.2500% Coolness
• 3/1600 = 0.1875% Athletics
• 3/1600 = 0.1875% Melee Combat
• 2/1600 = 0.1250% Combat Sense
• 2/1600 = 0.1250% Dexterity
• 1/1600 = 0.0625% Combat Reflexes
• 1/1600 = 0.0625% Drilling
• 1/1600 = 0.0625% Handgun
• 1/1600 = 0.0625% Mining
• 1/1600 = 0.0625% Perception
• 1/1600 = 0.0625% Prospecting
• 1/1600 = 0.0625% Rifle
• 1/1600 = 0.0625% Serendipity
• 1/1600 = 0.0625% Surveying
• 1/1600 = 0.0625% Weapons Handling

Contribution to health still unknown:

• Beauty Sense
• Coloring Methodology
• Doctor
• Equipment Methodology
• Force Merge
• Glamor
• Industrialist
• Manufacture Methodology
• Mindforce Harmony
• Quality Sense
• Robotology
• Surface Composition

 

[List Living]

Get it now =)

 

[wizzszz]

Isnt this all already known? i mean jdegres skill calculator works correctly, or i got this wrong

No, not at all, we have estimated numbers, some confirmed, some unconfirmed, and a variety of higher level unlocks that are completely unknown.


This post popped up today:

Weird.

I've just SkillScanned myself and the scanner told me that my HP is at 126.11,
BUT when I loaded all the skills to the Chipping Optimizer as a .csv file it
told me that I have 127.3 HP?

umm, help?


Thanks,
Rus

I was always a bit off for me, too, and a lot of research was going on to finally fix this.
However, since we have the detailed health displayed, nobody made use of the additional digits yet, so here we go.

 

[Doer]

The idea that skills contribute to the amount of 1600/n per hp works so well that i think the only unknowns are in the precise effect of attributes or perhaps a skill or two where n<1. The problem is, each skill that gives HP should allow you to FAP right after you get a gain, so i think they've all been tested. Nevertheless, it would be a start to compile a list of every skill gain after which you can't FAP. We'll need two confirmations of each to verify, due to regenerations.

An easier way to solve this if there were enough complete skill sets in wiki would be to do it the way i did some of the explosives skill contributions, by using a script that tries different permutations to find a fit.

In Ruslan's case it's almost certainly a missing or miss-scanned skill. Deviations from the current predictions that big don't happen at such low skill levels.

 

[wizzszz]

Contribution of Coloring:

=> 47% in Colorer
=> 5% in Material Designer
(No gain in Health)

[br]Click to enlarge[/br]

Stole this from Jimmy B.'s thread

Added to first post.

 

[jdegre]

First of all, great to see this task re-activated; it is about time that we nail down the HP calcs

You may as well use Jdegres skill scanner and post the numbers for Health and the chipped skill here, but i'm not sure if the way the decimals are generated is 100% correct, so screenies are preferred!

you can use screens or the csv's, whatever works better for you; however, i can tell you that the reason for the skill scanner to generate "inconsistent" decimals is that the edges of the green bars are heavily anti-aliased and blended with the background color.

so, what the skill scanner does is to filter the image to sharpen the edges of the bar, and this sharpening is dependent on the brightness of the environment where the screenshot was taken. you can do the same with your "eye", but it is not bullet-proof, either; sometimes it is really hard to say if a certain pixel belongs to the green bar, or not.

 

[Jimmy B]

The idea that skills contribute to the amount of 1600/n per hp works so well that i think the only unknowns are in the precise effect of attributes or perhaps a skill or two where n<1. The problem is, each skill that gives HP should allow you to FAP right after you get a gain, so i think they've all been tested. Nevertheless, it would be a start to compile a list of every skill gain after which you can't FAP. We'll need two confirmations of each to verify, due to regenerations.

An easier way to solve this if there were enough complete skill sets in wiki would be to do it the way i did some of the explosives skill contributions, by using a script that tries different permutations to find a fit.

In Ruslan's case it's almost certainly a missing or miss-scanned skill. Deviations from the current predictions that big don't happen at such low skill levels.

There's three possible sources of error imo:

(i) High level unlocks that no-one has posted info about (MF ones have been talked about quite a bit recently as possibles, but there's others that haven't been checked in detail either)
(ii) Unusual skills that most people have level 1 in. I'm thinking stuff like Animal Taming, Animal Lore, etc - don't know how well the rarer skills have been tested
(iii) Attributes

Anyway, good job starting this off wizz, I'll post my data later - I guess we can just get the skill values from jedgre's scanner rather than using screen shots?

[EDIT: Having read on - what sort of error does the scanner have in the decimals jdegre? I'm guessing its pretty small so OK for these purposes?]

 

[Doer]

There's three possible sources of error imo:

(i) High level unlocks that no-one has posted info about (MF ones have been talked about quite a bit recently as possibles, but there's others that haven't been checked in detail either)
(ii) Unusual skills that most people have level 1 in. I'm thinking stuff like Animal Taming, Animal Lore, etc - don't know how well the rarer skills have been tested
(iii) Attributes

The problem is that there seems to be a lot of variability at all different levels of avatar advancement, but data reliability is a big problem. I have that script that pulls down skills from wiki and crunches it, but it's a perpetual problem to know if the player has uploaded a complete set and how accurate the numbers are. We really need to establish the base HP with some degree of accuracy for new and old -- maybe that's been done. My skills predict a number very close to my actual health, and i have enough of the taming skills that it would skew it -- assuming the baseline is correct.

If people could upload a complete (i.e. all skills over 1.0) and accurate (i.e. using jdegre's scanner and checking that the professions it predicts are correct) skill snapshot to wiki and then post here, i could selectively grab their data and then let the program look for trends and such.

 

[wizzszz]

If people could upload a complete (i.e. all skills over 1.0) and accurate (i.e. using jdegre's scanner and checking that the professions it predicts are correct) skill snapshot to wiki and then post here, i could selectively grab their data and then let the program look for trends and such.

The problem is that you have to change the "Add skills with more than" manually to 0, otherwise it won't update the skills from the scanner data.

Maybe talk to witte to make him remove it, and i.e. add a "source column" that is set to "skill scanner" on upload of a file, and to "manually" if a single skill is changed by hand afterwards.

@ Jimmy B.
I think the error is not much more than 1 or 2 pixels, this won't make much of a difference, because it should be safe that MA used multiples of 10 (well, more like 100) for the calculation of health.

Not exactly sure how many pixels the bar has atm, we can include an error margin for the results until we get enough data (or bigger chipping changes).

 

[Doer]

The problem is that you have to change the "Add skills with more than" manually to 0, otherwise it won't update the skills from the scanner data.

No, it only excludes skills not over 1.0 so that isn't a problem. The problem is people uploading incomplete sets or doing it by hand, which is obviously not going to be as precise.

Not exactly sure how many pixels the bar has atm

124. Each bar (light green, dark green, light green...) is a single pixel in width.

 

[wizzszz]

No, it only excludes skills not over 1.0 so that isn't a problem. The problem is people uploading incomplete sets or doing it by hand, which is obviously not going to be as precise.

Jdegre could include i.e. a MD5 checksum in the skills file, to make sure they are unaltered.
If someone edits the file by hand (i.e. because of scanning errors), it will be marked as changed on upload, and therefore useless for your calculations...

Well, just a thought, i think we'll get it nailed down this time with a shared effort, now that we have the detailed health. If we have the contribution of the skills as exact as possible, it will be no problem to find the contribution of the attributes.

124. Each bar (light green, dark green, light green...) is a single pixel in width.

Is that the full length of the bar? It seems to start off with a few pixels already. But iirc thats exactly the length of the "old style" progress bars.

If you assume +/- 2.5 pixels, that would be pretty exactly +/- 0.02 health points, and for the skills, the error is neglectable when the chipping resulted in a change > 500 points.

 

[Doer]

Jdegre could include i.e. a MD5 checksum in the skills file, to make sure they are unaltered.
If someone edits the file by hand (i.e. because of scanning errors), it will be marked as changed on upload, and therefore useless for your calculations...

I'm talking about people changing their skills in wiki by hand, as some people do. Not maliciously, just to avoid the hassle of figuring out a skill scanner. That and missing a page when scanning the skills are both very plausible reasons for bad data on wiki.

Another source of trouble in this context is that you can actually deselect the box that updates the health value when you upload skills.

Is that the full length of the bar? It seems to start off with a few pixels already. But iirc thats exactly the length of the "old style" progress bars.

There is an empty pixel (black+faded background color) on either end inside the box, so the total length is 126 pixels, of which only 124 reflect actual active values. The error of missing a single dark bar is fairly negligible -- it's the missing skills that are the problem.

 

[wizzszz]

I'm talking about people changing their skills in wiki by hand, as some people do. Not maliciously, just to avoid the hassle of figuring out a skill scanner. That and missing a page when scanning the skills are both very plausible reasons for bad data on wiki.

I've suggested a "source" column in entropedia (that'll change to "manually" on single changes) in my last posts.

But there is a much more simple solution:
The skill scanner could ask for a direct upload to i.e. jdegre.net directly after scanning.

Or gather some people who are willing to provide the unchanged .csv files for that purpose. I keep backups of mine, although i do not verify the scanned values other than by estimated total skills value.

However, let's see what we can gather here, i.e. the 533.3 for athletics and melee combat listed on entropedia always seemed a bit odd to me, imo a result of neglecting the decimals of health before & after chipping.
(not sure if jdegre uses them in his formula though)

 

[Doer]

However, let's see what we can gather here, i.e. the 533.3 for athletics and melee combat listed on entropedia always seemed a bit odd to me, imo a result of neglecting the decimals of health before & after chipping.
(not sure if jdegre uses them in his formula though)

1600/2 = 800
1600/3 = 533.33...
1600/4 = 400
1600/5 = 320
1600/8 = 200

The fact that the 1600-based numbers work so well is why i think it's only in attributes/modifiers and/or base values that we're off.

It's true that the rounding could introduce some error, but that's only on the order of 0.0012 HP/10k skills.

 

[wizzszz]

1600/2 = 800
1600/3 = 533.33...
1600/4 = 400
1600/5 = 320
1600/8 = 200

The fact that the 1600-based numbers work so well is why i think it's only in attributes/modifiers and/or base values that we're off.

It's true that the rounding could introduce some error, but that's only on the order of 0.0012 HP/10k skills.

Well, indeed, never saw 533.33 as 1600/3, as the others are such a nice multiple of 20 Stupid me

The rounding was more meant towards full healthpoints, as we had no decimals before.

Well, still it hasn't been possible to nail it completely down so far (and some of the new skills and several unlocks are missing), do you at least get good results with your method when you check only avatars with a stamina of 1?

I played a lot with skill sets i made my on own during the last months (partially taken exactly on a "you've gained additional health" message), and still wasn't able to figure out some good numbers for the attributes, infact it worked a lot better when i changed some of the skill contributions.

You may see it as unnecessary to gather the data again, but additional confirmation with more precision should reveal what went wrong up to now, at least IMO.

 

[Doer]

Well, indeed, never saw 533.33 as 1600/3, as the others are such a nice multiple of 20 Stupid me

That came from MG Mighty's thread on health, which was the first rational explanation for the skill contributions to health that i saw. Attributes fit in there as well:
1600/40=40

do you at least get good results with your method when you check only avatars with a stamina of 1?

Stamina 9 Avatars (Health, deviation from predicted health)
91 0.66
92 0.22
92 -0.18
95 -0.22
95 0.78
97 -0.65
98 -2.37
99 0.19
100 0.64
102 0.34
102 0.61
102 -0.50
103 -0.14
103 -0.19
104 0.55
105 -0.05
107 0.26
108 -1.04
108 -0.09
110 -0.40
110 -0.25
111 -0.77
111 -0.30
111 0.32
114 0.27
116 -0.28
117 0.54
117 0.54
117 -0.40
121 -0.15
121 -0.31
123 1.12
125 0.35
125 0.54
126 0.41
128 -0.20
129 0.28
130 0.42
131 0.06
133 -0.62
134 -0.64
135 -0.22
136 -0.24
139 0.26
142 -0.22
145 -0.20
149 -0.60
155 0.86
162 0.48
172 0.24
175 -0.24
176 -0.54
176 1.05

Stamina 1 Avatars
89 1.37
90 0.50
92 0.95
97 0.08
100 0.16
105 -0.42
144 0.73


One additional problem is that some of these were probably uploaded before the health was shown to a decimal value, and therefore are off by +-1 anyway.

I played a lot with skill sets i made my on own during the last months (partially taken exactly on a "you've gained additional health" message), and still wasn't able to figure out some good numbers for the attributes, infact it worked a lot better when i changed some of the skill contributions.

My Avatar
143.0 0.60
143.0 0.98
143.0 1.16
143.0 1.28
144.0 0.76
144.0 0.92
144.0 1.04
144.0 1.23
144.0 1.41
145.0 0.47
145.0 0.76
145.0 0.93
145.0 0.99
145.0 1.16
145.0 1.35
145.0 1.46
146.0 0.52
146.0 0.64
146.0 0.76
146.0 0.88
146.0 0.97
147.0 0.96
147.0 1.36
148.0 0.47
148.0 0.61
148.0 0.79
148.0 0.98
148.0 1.07
148.0 1.22
149.0 0.48
149.0 0.54
149.0 0.71
149.0 0.94
149.0 1.13
149.9 0.50
150.2 0.49
150.3 0.49
150.4 0.49
150.5 0.51
150.6 0.51
150.7 0.51
150.8 0.51
150.9 0.51
151.0 0.51
151.1 0.50
151.2 0.50
151.3 0.50
151.4 0.51
151.5 0.52
151.6 0.52
151.7 0.52
151.8 0.52
151.9 0.52
152.0 0.51
152.3 0.51
152.6 0.51
152.7 0.52
155.7 0.54
156.0 0.53
156.1 0.53

I tried playing around with different values back when the 1600-trend was pointed out, both with my skills and some very high avatar skills, and nothing else came close to being as accurate at predicting all the values. As you can see, there's a very gradually increasing error visible since the decimal values of health became available. That becomes more pronounced for very highly skilled avatars. The difference is very subtle, so it's either due to a very slowly increasing skill (ie. not one of the known contributors) or some modifier effect (from an attribute, perhaps -- maybe health itself ). All the known contributors that i have would make too big of a difference if they were off by even a few points.

You may see it as unnecessary to gather the data again, but additional confirmation with more precision should reveal what went wrong up to now, at least IMO.

So it's not that i don't see it as necessary, just as effort better spent eliminating the skills currently not considered to give health and creating a large body of very precise data so that baseline health and more subtle modifiers can be teased out.

Edit: I forgot to reset the base health to the currently used value before i printed out my health values, which is why they're off more than in the previous example i gave.

 

[Jimmy B]

Well, was going to go to CP for the afternoon but the malls let me down with only 2 out of 7 parts of Serum found.

So did my skills instead. Scanner reads my health as 154.45, Optimiser calculates it as 154.28.

Full list of skills: http://www.entropiawiki.com/Info.aspx?chart=Avatar&id=2786

 

[wizzszz]

I've read MG Mighty's thread, i've read almost every post ever made on that topic.

My data are a bit different, as my skills rise, the error (0.25 points less than predicted by the chipping optimizer back then) get smaller as my skills rise.

The 1600 thingie always sounded very plausible, but there is still an error left.
With the old data, we can't be sure if it's not maybe 1610 or 1590.
(The 1600 is a nice number for humans, computers would prefer i.e. 1536 = 0x600)
And, we can't be sure if there aren't other skills that contribute like 1600/0.5 = 1 HP per 3200 or even less, too.
Such small contributions could very well slip through the detection by fapping.

We should generally separate old avatars with a stamina of 9 and newer avatars with a stamina of 1, it's very likely MA was forced to add/take some points of the base health to not change the HPs of older avatars too much on the last skill nerf.

This can only be done if we have the best possible data available, to minimize the error, and i'm pretty confident this thread will help revealing some things we haven't thought of before.

And even if not, it's a good start to gather people working on the health formula again. The effort to maintain this thread is minimal, most of the data will come from people chipping out anyway.

We still have no real good formula, and everytime i get stuck, i return to the start and re-validate the basics, and that's the idea behind this thread.
And i'm sure you'll agree that research on health is kinda "stuck" atm.

 

[wizzszz]

Well, was going to go to CP for the afternoon but the malls let me down with only 2 out of 7 parts of Serum found.

So did my skills instead. Scanner reads my health as 154.45, Optimiser calculates it as 154.28.

Full list of skills: http://www.entropiawiki.com/Info.aspx?chart=Avatar&id=2786

Did you just reveal your ingame identity?

100 ped and i'll delete the quote

 

[Jimmy B]

Did you just reveal your ingame identity?

100 ped and i'll delete the quote

No, its be known since I started posting screens from WoF globals.

The 1600 thingie always sounded very plausible, but there is still an error left.
With the old data, we can't be sure if it's not maybe 1610 or 1590.
(The 1600 is a nice number for humans, computers would prefer i.e. 1536 = 0x600)
And, we can't be sure if there aren't other skills that contribute like 1600/0.5 = 1 HP per 3200 or even less, too.
Such small contributions could very well slip through the detection by fapping.

I think the 1600 thing is pretty safe. MG tested that pretty much exactly for a few skills (see start of his thread). But if people are chipping its very easy now to check how much HP is gained from a chip so thats one way to verify them.

I think the errors are likely to come from attributes, from skills we have missed from the list, and possibly from an error in assumptions about starting health.

But we'll see. What we need first is plenty of data basically.

 

[Oleg]

Did you just reveal your ingame identity?

100 ped and i'll delete the quote

Those of us in the England WoF squad discovered Jimmy's secret identity a while ago

PS Jimmy your crafting skills suck, you noob

 

[Jimmy B]

Ah, I should have read your first post properly. I see you're after chipping data as your priority.

That's certainly useful and worth doing to verify/disprove the 1600 thing and check other skills, but from past experience it will take a long time. I think there's merit in data collecting (in the form that I've posted) too, since with reliable data we may be able to make some useful deductions much quicker. We still need a reasonable number of reliable data sets though.

PS Jimmy your crafting skills suck, you noob

Well, that's why I put 'Hunter' as my profession and not 'Crafter' My mining skills suck too.

But as I have no more unlocks due this decade, I may skill crafting and mining up a bit sometime to get a few more low level unlocks.

 

[Oleg]

Well, that's why I put 'Hunter' as my profession and not 'Crafter' My mining skills suck too.

But as I have no more unlocks due this decade, I may skill crafting and mining up a bit sometime to get a few more low level unlocks.

Yeah, you should be able to get BPC pretty easily from where you are now and I guess the first mining ones too but I don't know much about that digging stuff.

On the plus side, I am impressed by your taming skills

Anyway, enough of this off-topic ribbing, carry on with the HP thing guys

 

[Jimmy B]

On the plus side, I am impressed by your taming skills

Considering that I've never tamed a mob I afaik I think they're pretty good too

 

[Doer]

Here's an interesting bit of trivia:
Contribution of Health to Health, r^2 for 67 avatars from Entropedia
2000.0 0.329
1000.0 0.327
750.0 0.326
500.0 0.323
250.0 0.315
125.0 0.307
100.0 0.305
90.0 0.305
80.0 0.306
70.0 0.308
60.0 0.314
50.0 0.326
40.0 0.354
30.0 0.427

There are plenty of ways to explain this, but there's definitely a minimum around 80-100 (ie. 100 health gives 1 additional health).

I will try it with a bigger set of data, and then with stamina=1 vs. stamina=9.

 

[Jimmy B]

Here's an interesting bit of trivia:
Contribution of Health to Health, r^2 for 67 avatars from Entropedia
2000.0 0.329
1000.0 0.327
750.0 0.326
500.0 0.323
250.0 0.315
125.0 0.307
100.0 0.305
90.0 0.305
80.0 0.306
70.0 0.308
60.0 0.314
50.0 0.326
40.0 0.354
30.0 0.427

There are plenty of ways to explain this, but there's definitely a minimum around 80-100 (ie. 100 health gives 1 additional health).

I will try it with a bigger set of data, and then with stamina=1 vs. stamina=9.

Shit, that made me burn my toast.

But anyway, Health contributing to Health is meaningless no?

If Health = A.Skill1 + B.Skill2 + ... + Y.Skill25 + Z.Health

Then rearranging gives:

Health = A.Skill1/(1-Z) + B.Skill2/(1-Z) + ... + Y.Skill25/(1-Z)

So that would kinda suggest we just have the contributions all out by a factor of 1-Z.

 

[Doer]

Shit, that made me burn my toast.

But anyway, Health contributing to Health is meaningless no?

If Health = A.Skill1 + B.Skill2 + ... + Y.Skill25 + Z.Health

Then rearranging gives:

Health = A.Skill1/(1-Z) + B.Skill2/(1-Z) + ... + Y.Skill25/(1-Z)

So that would kinda suggest we just have the contributions all out by a factor of 1-Z.

Sorry about your toast.

Yeah, that's true. Still, if it turned out that making Health contribute to health at the rate of 1/40 like the other attributes explained the discrepancies, well what would be much nicer than if all our skill contributions are off by x^-3.1415. I'm just trying to figure out a way to get a very small cumulative contribution to explain the gradual drift.

It looks like a dead end, anyway. At least it seems we can put a nail in that coffin. That leaves missing individual contributors and some modifier effect.

New avatars (Stamina=1; n=22)
Health contr. to health, Base health, R^2
2000.0 87.9 0.16
1000.0 87.8 0.16
500.0 87.7 0.16
250.0 87.5 0.18
150.0 87.2 0.21
120.0 87.1 0.24
100.0 86.9 0.27
90.0 86.8 0.29
80.0 86.6 0.32
70.0 86.4 0.37
60.0 86.2 0.43
50.0 85.8 0.54
40.0 85.3 0.74
30.0 84.5 1.15


Older avatars (stamina=9, n=323)
Health contr. to health, Base health, R^2
2000.0 88.3 0.21
1000.0 88.3 0.21
500.0 88.1 0.21
250.0 87.9 0.21
150.0 87.6 0.23
120.0 87.4 0.24
100.0 87.2 0.26
90.0 87.0 0.27
80.0 86.9 0.29
70.0 86.7 0.33
60.0 86.4 0.37
50.0 86.0 0.45
40.0 85.4 0.60
30.0 84.4 0.93

Weird that the first 150 entries on wiki gave that trend.

At least i was able to get a more accurate base health:
87.95 for new avatars, 88.39 for old avatars

 

[jdegre]

guys, remember that the base HP used in the chipping optimizer is not the same as the one suggested in the HP thread. it is off by 0.5 HPs, and the reason (as explained in other thread) is that it simply produced better results for a vast majority of avatars (9-stamina avas without extremely high attribute values).

when i implemented this function in the chipping optimizer (1 year ago), this type of avatars were the most abundant; these days, the 1-stamina avas are getting more and more popular, thou.

my point is that this value should not interfere with the theoretical calculations; we know it is wrong, and it is simply a hack that tries to improve a bit the result with the knowledge we have today.

 

[Jimmy B]

I don't think we know the base HP for sure at all do we? (ie. the one in the HP thread is also just a guess).

Does your stuff there prove those base HP numbers you've posted Doer?

 

[levithanikos]

Nice thread Wizzszz

Here are some Senior EFDs