FYI: Skillgain experiment!

[Dawis]

Hi all.

Quite a while ago I asked a question, and it seemed that there is no definite answer for that question. Does skillgain remain the same or does it change over time?

I promised to do some experiments and post some results. It was not easy but here they are.

But first let us get straight and precise once more what we are talking about:

Definitions
Skillgain - is defined as a skill amount (in terms of ESI value) you get by performing certain well defined activity. So we could define skillgain per one shot of opalo, or skillgain per one click on Damper BP, or skillgain per one drop of a bomb with certain equipment.


What we are trying to find out?
is this skillgain constant or does it change depending on some factors. The factor we are interested in this particular experiment is: skill or professional level. We are interested in this due to obvious reason: It seems that we are receiving less skill doing the same activity as our skill level increases.

Experiment Details

We defined skillgain as amount of ESI you gain (chipping out gained skill levels) per 10 000 clicks on standard damper blueprint.

Experimnt was carried out in several stages.
The first stage is fresh avatar. Starting with level 0 in all professional standings. Several runs were carried out, and the results were monitored each time.

Results

Here are the 1. stage results:

ENG     MME     MECH  BPC    ENGesi  MMEesi  MECHesi                    PED
249 	 266	 128	 0	 0.33	  0.36	    0.14	     581.75
399	 417	 234	 0	 0.36	  0.39	    0.16	     492.13
490	 509	 306	 0	 0.32 	  0.34	    0.14	     781.74
569	 596	 369	 0	 0.37	  0.38	    0.16	     492.01
662	 710	 416	 0	 0.32	  0.38	    0.15	     651.95
779	 836	 463	 0	 0.35	  0.38	    0.16	     558.46
876	 924	 500	 0	 0.34	  0.39	    0.14	     566
944	 991	 530	 0	 0.33	  0.42	    0.15	     524.98
1007	1052	563	0	0.33 	 0.38	   0.15	            592.83
1058	1122	594	0	0.4	 0.4	    0.12	     509.37

ENG       - level of Engineering skill level after run
MME      - level of Manufacture mechanical equipment level after run
MECH     - level of Mechanics after run
ENGesi   - value of corresponding skill gained in run
MMEesi  - value of corresponding skill gainer in run
MECHesi - value of corresponding skill gainer in run
PED       - value of TT loot gained in run

Each run TT costs = 700 PED

So as you see experiment total costs are 7000 PED with total return of 5751.22 PED. Experiments were not only costly but lenghty too.
i reserve the next post for result analysis and main conclusions.
Which will follow shorlty.

 

[Dawis]

Conclusions:

1. The skillgain is not constant but variance is small enough.
The reason of this variance is to be determined. And if there exists lootius there could also exist skillius, that is responsilbe for skillgain variations day by day.
2.The skillgain is not affected by your current skill level. Although stage 2. results are no published jet - I must say they look very similar. 2. Stage tests are performed by avatar with engineeging level 4000, so the difference in skill level should be large enough if skill level affects skillgain. The skillgain on 2. stage remains about the same, so it seems that it only seems that skillgain is reduced on higher professional levels.
3. Average ENGeneering skillgain per 10k clicks on standart damper BP appears to be 0.345 ESI after chipping it out.
This result leads me to some interesting thoughts. Assuming a young avatar joins the game and decides to become a top level crafter. ENG is considered one of the main skills, and let us assume that top level crafter needs ENG level 10 000. How long he have to click on dampers to reach that level?
Let us do some calculations
Chipping optimizer suggests that you need to chip in some 2279.84 PED of ESI to get to that level. Let us assume our young crafter is determined to do 10k clicks each day. That means that on day 6608, since joining the EU, he will be a proud top level crafter, that is unless MA will change something in skill system.
So young players get ready.. you need only 18 years and few months of nonstop damper crafting to join the elite crafter society.

In other words it is nearly impossible or rather insane to reach that level with natural skilling - of course standard damper may not be the best skilling BP around, but I would say that it is certainly not the worst and most probably offers best price / performance ratio.

So another conclusion to this conclusion - The crafting skill base grows up very slowly. I would almost like to claim that it shrinks, because chipping out loses 10% of skills, and whenever you see 100 PED chip in auction it means that 10 PED of engi is lost from global EU skillbase and some young crafter will have to work 30 days to gain those 10 PED back into the system.

This allows me to assume, that as crafting skill base shrinks or remains about the same, crafting skill prices could raise once again and even exceed previous levels if the skills will prove to be useful (car manufacturing), cause there will simply be no way to get large amounts of crafting skills fast.

 

[Xen]

Was it 10k clicks, or 10k clicks of materials at the beginning of the run? If it was 10k of mats, the results could vary merely because of variance in the number of partial successes and mats returned.

 

[Dawis]

Was it 10k clicks, or 10k clicks of materials at the beginning of the run? If it was 10k of mats, the results could vary merely because of variance in the number of partial successes and mats returned.

It was 10k of clicks.. all the materials on loot were transfered to different stack - but I can not guarantee it 100% in all cases, because there is no way you can tell EU to put all looted materials on specified stack.
That is why i picked the click count to be large enough to bury small accidents under the heavy data

 

[WoenK]

Isn't the test a bit screwed if you take higher than level 0 BPs ?
Can remember from my early skilling, that I had quite a lot of failures and skill gain seemed also quite low.
Till about level 1 filters did work for skilling a bit, but besides that the SiB BPs were the most cost/skill/time effective.

 

[Dawis]

Isn't the test a bit screwed if you take higher than level 0 BPs ?
Can remember from my early skilling, that I had quite a lot of failures and skill gain seemed also quite low.
Till about level 1 filters did work for skilling a bit, but besides that the SiB BPs were the most cost/skill/time effective.

BP level affects skillgain.. but it is ok as long as we use the same BP..
This experiment was aimed to conclude is the skillgain reduced on higher levels..

If we want to know BP level effect on skillgain we need to do other tests.

 

[Bomb]

Can never hurt to inform the public. Always like to see new information especially since skill gains affect all Entropians. Keep it up.

 

[Dawis]

Can never hurt to inform the public. Always like to see new information especially since skill gains affect all Entropians. Keep it up.

I think i will, but It will not be so fast.. As I said - those tests are lenghty and expensive..
Probably tests with a gun would be more rewarding, but it would be more difficult to assess results, as in hunting probably many more factors than hunting equipment could affect skillgain.

 

[Joat]

Tests are always fun to do (well, almost at least ).
Some thoughts of mine about this:

Sadly tests are hard to compare with others as a precise tool.
Why? Everyone has their unique history, and also current status.
If two with similar skills, do a test, and one is just closing in to new
rank in prof stand, and the other has got the new rank, the result
will not be similar, as if both were closing in to prof stand or if
both had passed it recently.
What has been done before test? If one has skilled zero-cost-sweating
and the other has scanned for +500ped, the result will again vary
some, both in skillvalues and loot. If several join in and shall do same thing,
IMO they should do exact same thing for over a week. But still no
garantuee test will be equal...

So, that said, tests can still show a lot. So keep up the good work.

 

[Quint]

Interesting results, thx and +rep for posting (and for your patience ).

When you post the results of stage 2 (performed by an ava with 4k engi), can you please include the gains on bpc ?

Thanks !

Quint

 

[blubalou]

I like what you did here. +rep for taking the time to do it, post it, and offer your analysis.

What I've always thought is that skill gain is directly related to how much decay you incur (and then the question is 'are crafting losses considered part of decay?' but that's another topic.)

So to break this down even further.... let's assume the numbers you found after 10k clicks are the expected results after n number of clicks. The TT loot return was roughly 82%, so the net expenditure was 1248.78. For that amount you get 34.5 PEC in Engineering skills, or .027627 PEC skills for 1 PEC spent.

As you pointed out, at 7 PEC / click rate, it would take 18 years at 10k clicks / day for a new avatar to reach the 10,000 point level in Engineering. Well we know EU hasn't been around 18 years, but I assume there are some folks out there with 10,000 Engineering points so they must be finding that in order to skill faster you've got to spend more / click.

Which is exactly how the game is designed. As you level up in crafting, hunting, and mining, you are able (and sort of encouraged by the whole SIB thing) to get better gear and hunt, craft, or mine for bigger loots. If you do, I suspect the number of green messages in your chat box will continue to report skill gains at a steady rate. However, if you choose to level off then so do the apparent skill gains. But in reality, at least based on what we see here, the gains remain the same it just doesn't seem that way (probably because we see fewer green messages.)

 

[Joat]

I like what you did here. +rep for taking the time to do it, post it, and offer your analysis.

What I've always thought is that skill gain is directly related to how much decay you incur (and then the question is 'are crafting losses considered part of decay?' but that's another topic.)

So to break this down even further.... let's assume the numbers you found after 10k clicks are the expected results after n number of clicks. The TT loot return was roughly 82%, so the net expenditure was 1248.78. For that amount you get 34.5 PEC in Engineering skills, or .027627 PEC skills for 1 PEC spent.

As you pointed out, at 7 PEC / click rate, it would take 18 years at 10k clicks / day for a new avatar to reach the 10,000 point level in Engineering. Well we know EU hasn't been around 18 years, but I assume there are some folks out there with 10,000 Engineering points so they must be finding that in order to skill faster you've got to spend more / click.

Which is exactly how the game is designed. As you level up in crafting, hunting, and mining, you are able (and sort of encouraged by the whole SIB thing) to get better gear and hunt, craft, or mine for bigger loots. If you do, I suspect the number of green messages in your chat box will continue to report skill gains at a steady rate. However, if you choose to level off then so do the apparent skill gains. But in reality, at least based on what we see here, the gains remain the same it just doesn't seem that way (probably because we see fewer green messages.)

Green lines are far from everything. What's interesting is value per gain IMO.
Havn't done a test about value per gain in crafting, so I don't know how
big variation in values are, but in hunting, it can have very different value
per gain.

 

[blubalou]

Green lines are far from everything. What's interesting is value per gain IMO.
Havn't done a test about value per gain in crafting, so I don't know how
big variation in values are, but in hunting, it can have very different value
per gain.

Right. After thinking about it for a few more days, I think this is what I was trying to get to with my post above. The experiment seemed to show a pretty consistent value gain per PEC spent regardless as to whether the Avatar was low level or high level.

But if then if the high level Avatar graduates to a higher level BP that costs more than 7 PEC per / click, would we still see roughly the same .027627 PEC gain in ENG skill / PEC of crafting losses? Or would it be more or less?

 

[Harmony.]

I don't suppose you want to do the same tests on oa-101's too? Would be interested to know how long it would take to reach 10k engineering that way

 

[Dawis]

I don't suppose you want to do the same tests on oa-101's too? Would be interested to know how long it would take to reach 10k engineering that way

gladly.. I would need a few sponsors though..
I don`t suppose it is possible to gain any consistent data per 100 click runs..
And even 100 clicks cost a bit there

 

[kosmos]

good post +rep

 

[Dawis]

I am thinking about another phase of several testing runs to get some more results using this method. However it will be some time before the results actually get here.

 

[WoenK]

Low level SiB BPs would be interesting. Could tell a bit more if it is really advisable to craft on them unmaxed.

 

[Dawis]

Low level SiB BPs would be interesting. Could tell a bit more if it is really advisable to craft on them unmaxed.

I think I will need some time to prepare everything for experiments, when It comes closer will consider I will think about details.

Each experiment session is like 10 hours so it is time consuming - any avatars willing to help out carrying out experiments are welcome - I must warn that such experiments are not always financially rewarding.

 

[kVk]

Hello fellow entropians this is my first post in this forum.
I'm not a native speaker so plz forgive my english,I hope you will understand the following.

I recently chipped out all of my mining skills,I had reached lvl 27 prospector and I devided them into small skill implants so I could sell them faster.I noticed some weird things when I was filling that ESI's.
I will give you examples because it is easier to describe it.
Let's say I had 3500 skills in prospector. I was getting the same tt skill implants for the same amount of skill points I extracted.When I dropped a rank I need more skill points to exctract in order to get the skill implants with the same tt value as I did with the previous rank.So far so good, logical behaviour.
But then at some point as the lvl and ranks were dropping the skill implants I was getting started to be bigger than the previous ones with the same skill points extracted!!!
and after that it became normal again(lower rank-more skill points to get approximately the same tt value of skill implant as I did in the previous rank).
And then I remembered that there were some points in the game while I was mining that the skilling was faster than before.And I assume It was that point that when I chipped out the skills started to give bigger skill implants.So I guess MA made it that way so the player doesn't get very frustrated with the skilling.
Skilling is not a smooth straight line it has ups and downs.
Maybe you would like to take that in mind in your experiment. I hope I helpped a litle.

 

[Doer]

Hello fellow entropians this is my first post in this forum.
I'm not a native speaker so plz forgive my english,I hope you will understand the following.

I recently chipped out all of my mining skills,I had reached lvl 27 prospector and I devided them into small skill implants so I could sell them faster.I noticed some weird things when I was filling that ESI's.
I will give you examples because it is easier to describe it.
Let's say I had 3500 skills in prospector. I was getting the same tt skill implants for the same amount of skill points I extracted.When I dropped a rank I need more skill points to exctract in order to get the skill implants with the same tt value as I did with the previous rank.So far so good, logical behaviour.
But then at some point as the lvl and ranks were dropping the skill implants I was getting started to be bigger than the previous ones with the same skill points extracted!!!
and after that it became normal again(lower rank-more skill points to get approximately the same tt value of skill implant as I did in the previous rank).
And then I remembered that there were some points in the game while I was mining that the skilling was faster than before.And I assume It was that point that when I chipped out the skills started to give bigger skill implants.So I guess MA made it that way so the player doesn't get very frustrated with the skilling.
Skilling is not a smooth straight line it has ups and downs.
Maybe you would like to take that in mind in your experiment. I hope I helpped a litle.

It's called a sine wave, and you are very observant. It's a function of level, not time, but in a sense it does slow down at the same place as you get more out of it chipping back out.

Dawis, i think i missed this thread up until now. Nicely done. I think we're pretty convinced that skill gains are not a function of current level except perhaps in a small, cyclic sense. One thing that really would be good to have proof of is whether unlocking a skill increases overall skill volume per activity or just redistributes it. Before and after BPC, for example.

 

[Xen]

It's called a sine wave, and you are very observant. It's a function of level, not time, but in a sense it does slow down at the same place as you get more out of it chipping back out.

Dawis, i think i missed this thread up until now. Nicely done. I think we're pretty convinced that skill gains are not a function of current level except perhaps in a small, cyclic sense. One thing that really would be good to have proof of is whether unlocking a skill increases overall skill volume per activity or just redistributes it. Before and after BPC, for example.

I think what he was trying to say is, it's not a smooth curve. In the past I have taken out (representative numbers) 100 skill points of a skill, then the next 100 skill points I removed resulted in a larger skill implant. This would not be possible if it were a smooth curve.

 

[Doer]

I think what he was trying to say is, it's not a smooth curve. In the past I have taken out (representative numbers) 100 skill points of a skill, then the next 100 skill points I removed resulted in a larger skill implant. This would not be possible if it were a smooth curve.

Chipping out in small quantities is essentially following the derivative of that curve. The slope goes up and down with the 500 period sine so, the smaller the chip, the more likely you are to get in the region where the chip size alternates between bigger and smaller as you go.

 

[Xen]

Chipping out in small quantities is essentially following the derivative of that curve. The slope goes up and down with the 500 period sine so, the smaller the chip, the more likely you are to get in the region where the chip size alternates between bigger and smaller as you go.

Define small? I'm talking about 10-20 ped chips.

 

[Doer]

Define small? I'm talking about 10-20 ped chips.

Say, less than 250 levels. Some examples (predictions from WikiTools):

Level Change: Volume chip required (chipping in, so x0.9 for chipping out)
1000-1100: 0.70
1100-1200: 0.54

5000-5100: 14.89
5100-5200: 11.67
5200-5300: 9.31
5300-5400: 12.86
5400-5500: 19.41

 

[Dawis]

Say, less than 250 levels. Some examples (predictions from WikiTools):

Level Change: Volume chip required (chipping in, so x0.9 for chipping out)
1000-1100: 0.70
1100-1200: 0.54

5000-5100: 14.89
5100-5200: 11.67
5200-5300: 9.31
5300-5400: 12.86
5400-5500: 19.41

This kind of behaviour has been observed in my experiment data.. Hell I have even noticed it without any experiments...
Bu skillgain in terms of ESI value gained, or shall we call it experience remains constant or near constant - fluctuating very little.

Diferent skill levels however require various amount of experience. While in general experience required to get next skill level increases exponentially, the amount of experience needed for next skill level tends to increase berofe getting to next level and decrease after getting the next level..

Although the level start and end points are not the only deviations from normal exponential curve.

 

[Dawis]

It's called a sine wave, and you are very observant. It's a function of level, not time, but in a sense it does slow down at the same place as you get more out of it chipping back out.

Dawis, i think i missed this thread up until now. Nicely done. I think we're pretty convinced that skill gains are not a function of current level except perhaps in a small, cyclic sense. One thing that really would be good to have proof of is whether unlocking a skill increases overall skill volume per activity or just redistributes it. Before and after BPC, for example.

Although I have not published that data, the skillgain experiment performed on avatar that had BPC seemed to show the same results - thus unlocking the BPC does not decrease other skill skillgain - it stays the same.

In other words - unlocking new skills increases your total skillgain per activity.

 

[Doer]

This kind of behaviour has been observed in my experiment data.. Hell I have even noticed it without any experiments...
Bu skillgain in terms of ESI value gained, or shall we call it experience remains constant or near constant - fluctuating very little.

Yes, that's what i said back in my first post. We just got a little sidetracked after that.

Diferent skill levels however require various amount of experience. While in general experience required to get next skill level increases exponentially, the amount of experience needed for next skill level tends to increase berofe getting to next level and decrease after getting the next level..

Although the level start and end points are not the only deviations from normal exponential curve.

Are you talking about individual skill levels or skill ranks? How can you even tell if there's a cycle within an individual level at level 5000? Isn't that like 3-4 green lines, at best?

Good to hear that your post-BPC data supports the working hypothesis. Skilling towards unlocks seems less aimless when it works that way. Is there data before and after the unlock for the same avatar, or are you comparing between two different ones?

Finally, a bit of a methodology check: what are you using to calculate the skill volume change? None of the 3rd-party tools are perfect. Some make more effort to follow the exact skill curve than others, but you have to consider that anomalous results could be "instrumental". You can reduce this possibility by cross-checking. FYI i believe WikiTools is the closest (because i've got more OCD than jdegre and Carebear modeled the Entropiatools one after WikiTools).

 

[Dawis]

Yes, that's what i said back in my first post. We just got a little sidetracked after that.



Are you talking about individual skill levels or skill ranks? How can you even tell if there's a cycle within an individual level at level 5000? Isn't that like 3-4 green lines, at best?

Good to hear that your post-BPC data supports the working hypothesis. Skilling towards unlocks seems less aimless when it works that way. Is there data before and after the unlock for the same avatar, or are you comparing between two different ones?

Finally, a bit of a methodology check: what are you using to calculate the skill volume change? None of the 3rd-party tools are perfect. Some make more effort to follow the exact skill curve than others, but you have to consider that anomalous results could be "instrumental". You can reduce this possibility by cross-checking. FYI i believe WikiTools is the closest (because i've got more OCD than jdegre and Carebear modeled the Entropiatools one after WikiTools).

Yes I got a bit fuzzy with my terminology.. I didn`t use the right word for ranks. What I meant to say is that skill levels becomes harder to attain before the change to next skill rank, and easier to attain after you gained a new rank. At least that is my observation - this means experience needed for each next skill level generaly increases after exponential curve, but deviates from it creating various spots where skilling seem easier and harder than normal. that way although in general higher the skill harder to skill it up, on certain intervals skilling up may seem easier than a moment before.
Does it make sense now?
I think you implied that by saying that the skilling seems to reflect sinus curve, and you experience fast and slow skilling periods.

Yes my post BPC data supports the hypothesis, but not with the same avatar. I will look forward to make tests with same avatar.

And I use chip optimizer (or something like that) for determining the gained ESI value after each test. However your comment gave me an idea. I know they are unprecise, so after the whole testing is done, I will chip out the ESI's in exaclty the oposite ordder as I got skills, so I will get the real values.

That way we will also be able to evaluate the precision of skill tools.

Of course all data here will only be inside the precision scope of one skill level.
which means that testing on very highly skilled avatars tends to become more vague and not precise.