Menu
What's New
By:
Filters

Fill up ESI data...

Status
wizzszz said:
Added some more fresh data, and removed 2 erraneous data from post above
Click to expand...

Thanks, and this is enough to have a good approximation in the range of 1-1000 skills:

formula.gif


The formula is as follows:

ChipTT = X * ( 0.001 + 2.56*10**(-6)*X + cos (2*pi*X/600)/5000 )

Here X is skill.

The picture shows auction data (pink squares), approximation by the formula and their difference, that is very close to 0 in the range of 0-1000.
 
BUT: Mathematica finds no integral and my analysis lessons are too far back to let me find one.
Click to expand...

Mathematica can do numeric integration as well as analytical.
It will first try the analytical and (as you said) find no analytical solution.

You can then force it to do a numeric definate integral and specify the bounds,
0 to X.

Paul
 
is this right?


https://www.planetcalypsoforum.com/forums/showpost.php?p=813132&postcount=7

EarlVonKippler said:
several small ones (10 chips at 10 ped each)
http://esi.coolinc.info/?level=1&chip=10
The tt value of your skill (atlevel 1) is 0.00 PEDs, and a chip of 10.00 PEDs tt will raise you to level 1842.3 (a gain of 1842.5 levels).
http://esi.coolinc.info/?level=1842&chip=10
The tt value of your skill (at level 1842) is 10.00 PEDs, and a chip of 10.00 PEDs tt will raise you to level 2571.8 (a gain of 729.8 levels).
http://esi.coolinc.info/?level=2572&chip=10
The tt value of your skill (at level 2572) is 20.00 PEDs, and a chip of 10.00 PEDs tt will raise you to level 3007.9 (a gain of 435.9 levels).
http://esi.coolinc.info/?level=3008&chip=10
The tt value of your skill (at level 3008) is 30.00 PEDs, and a chip of 10.00 PEDs tt will raise you to level 3323.5 (a gain of 315.5 levels).
http://esi.coolinc.info/?level=3324&chip=10
The tt value of your skill (at level 3324) is 40.02 PEDs, and a chip of 10.00 PEDs tt will raise you to level 3637.2 (a gain of 313.2 levels).
http://esi.coolinc.info/?level=3637&chip=10
The tt value of your skill (at level 3637) is 50.02 PEDs, and a chip of 10.00 PEDs tt will raise you to level 3819.2 (a gain of 182.2 levels).
http://esi.coolinc.info/?level=3819&chip=10
The tt value of your skill (at level 3819) is 60.01 PEDs, and a chip of 10.00 PEDs tt will raise you to level 4017.1 (a gain of 198.1 levels).
http://esi.coolinc.info/?level=4017&chip=10
The tt value of your skill (at level 4017) is 70.01 PEDs, and a chip of 10.00 PEDs tt will raise you to level 4195.9 (a gain of 178.9 levels).
http://esi.coolinc.info/?level=4196&chip=10
The tt value of your skill (at level 4196) is 80.01 PEDs, and a chip of 10.00 PEDs tt will raise you to level 4312.1 (a gain of 116.1 levels).
http://esi.coolinc.info/?level=4312&chip=10
The tt value of your skill (at level 4312) is 90.00 PEDs, and a chip of 10.00 PEDs tt will raise you to level 4450.5 (a gain of 138.5 levels).
Click to expand...
 
wizzszz said:
I've done a bit calculating on low TT implants myself as suggested here to help a bit with more precise data.

Its pretty obvious that the displayed TT value of the implants in auctions are rounded downwards!

Hope it helps :)
Click to expand...

Indeed it does! It is clear now why it was so hard to get a good fit for the data we had for very low tt and have it work both directions. I think the new fit i did, even though the error looks large with all data but yours, will be much more accurate. Thanks!

Gingko said:
I have watched this interesting thread for a while and spend some thoughts on this and I think I have a couple of new insights, but also some question marks to add at the end.

First of all, I took a slightly different look at the data, i.e. I looked at the change in TT of the ESI at a certain skill level, called for simplicity dTT/dSkill.
Click to expand...

This is what i started to do for high levels of skills that were unlikely to ever fit perfectly into the table of known level<->tt: I calculate a slope from a few small chips. I hoped to get enough to fit a function and then integrate, but the sinusoid makes that harder to do accurately. I just got some more data sent me that should help, but i have been trying to approach the problem from below by figuring out the trend in the function above 4k.

Did you evaluate delta y/delta x for all the data points available, or in some more complex way?

So, except for the very low range I got a nice function that I just need to integrate from 0 to Skill to get the TT of the chip, BUT: Mathematica finds no integral and my analysis lessons are too far back to let me find one.

So I took the "cheap" approach and simply calculated dTT/dSkill for all individual skill levels starting at 250 and summed it all up until the skill level I am interested in. With this "manual Excel" integration I end up with the following prediction:

a) Low levels
ESIdata5.gif


b) High levels
ESIdata6.gif
Click to expand...

I found that i could much more easily get a fit to the range 250-1500 using a function of this form. While there is excessive curvature to this function --
Gingko-error.png

-- it does seem to fit inherently better than one of the form exp(skill/tt)+sin(skill)*exp(skill/tt). The thing i'm trying to figure out is what MA does with this function. It's not integrable, i don't think it's easily invertible, and it seems ridiculously complicated to numerically integrate every time the program needs to know the mapping from tt to skill or vice versa.

That's the main reason i was staying with a function that is a sum of sinusoid and exponent: it's possible to differentiate and integrate.

Strange.

I'm considering using this form to cover the range 250-3500 because it does a better job overall, getting about 0.4% less average error. I could create a table and numerically integrate each time the function is called (or store it in code). It just seems ridiculously cumbersome. :p

If we can figure out what the missing bit is that causes leads to excessive curvature, it should be much easier to connect the low level and high level slope data i already have. Still, i tried several different things and haven't found a suitable form yet.

Note, I cannot get a good fit for the very high levels and the medium levels at the same time. At the moment, I suspect that there is one function added on top responsible for the deviation at low skill levels. This function's effect should extend somewhat into the medium skill region as well.

Have no more time to fiddle around with the data and really would love to get some feedback on this (especially the integration) before I try more... any Analysis experts around?
Click to expand...

Kolobok said:
Thanks, and this is enough to have a good approximation in the range of 1-1000 skills:

The formula is as follows:

ChipTT = X * ( 0.001 + 2.56*10**(-6)*X + cos (2*pi*X/600)/5000 )

Here X is skill.

The picture shows auction data (pink squares), approximation by the formula and their difference, that is very close to 0 in the range of 0-1000.
Click to expand...

Your cos should be a sin, and the period is 500, not 600. Other than that it seems to be about as good a fit from 0-1000 as i have.


sob said:
is this right?


https://www.planetcalypsoforum.com/forums/showpost.php?p=813132&postcount=7
Click to expand...

Yes, at least in theory. I don't know if the numbers are absolutely correct but there is a large up and down movement to the advancement so that is certainly possible.
 
I have added 5 more values to my post before here

Added values are blue, to make it easier to find them!

Progress is so slow, because I use ONLY the skills where i can be sure about zero skill gains so far (Dispense decoy, Beauty professions, Wood carving & processing) !

Too bad there is no way to track down the gained skill level to some decimal digits... (except chipping them in, ofc)


/EDIT:

One thought popped up: Is it possible that the larger error at higher TT value comes from summing up TT values up to those levels and thus adding up the rounding errors?

Might be wrong, error seems to rise with almost no deviation, but maybe someone can take a closer look to be sure...
 
Last edited:
Doer said:
Your cos should be a sin, and the period is 500, not 600. Other than that it seems to be about as good a fit from 0-1000 as i have.
Click to expand...

I haven't seen your formula for low TT yet, did I miss it? There can be many approximations at low TT. First I tried with the period of 500, but the existing data show it is 600, and 500 gave larger error. As for the sin being suggested instead of cos, probably you missed the x factor outside the brackets.
 
wizzszz said:
I have added 5 more values to my post before here

Added values are blue, to make it easier to find them!

Progress is so slow, because I use ONLY the skills where i can be sure about zero skill gains so far (Dispense decoy, Beauty professions, Wood carving & processing) !

Too bad there is no way to track down the gained skill level to some decimal digits... (except chipping them in, ofc)


/EDIT:

One thought popped up: Is it possible that the larger error at higher TT value comes from summing up TT values up to those levels and thus adding up the rounding errors?

Might be wrong, error seems to rise with almost no deviation, but maybe someone can take a closer look to be sure...
Click to expand...

Many of the tt values are taken directly from an implant with higher tt, rather than summing from low values. One 10 PED implant has an error of less than 0.2%, so it's probably only the really low values that came from low tt chips that show this truncation error.

Here's my new fit to the 1-1k data, using your data to create the fit for values below 250 and a function of a form like Gingko's for the rest.

Gingko-fit-1k.png


The periodic errors due to a poor match of the old form of function i have been trying (y = exp(x) + sin(x)*exp(x)) are really pronounced in this range so it seems profitable to use the Gingko fn here in spite of the inconvenience of numerically integrating it:

Gingko-error-1k.png


You can see that the error increases quickly at very low tt as the truncation of value below 0.01 PED becomes a higher and higher fraction of the total value. Your points don't show that trend because they avoid the truncation.

Here's the slope for the same range. This shows why making a fit from the slope rather than from the tt directly is error-prone: the huge amplitude of the sinusoid in the slope tends to increase uncertainty when using the delta tt/ delta level of a chip to create data for the fit.

Gingko-slope-1k.png


Kolobok said:
I haven't seen your formula for low TT yet, did I miss it? There can be many approximations at low TT. First I tried with the period of 500, but the existing data show it is 600, and 500 gave larger error. As for the sin being suggested instead of cos, probably you missed the x factor outside the brackets.
Click to expand...

I'm not sure how multiplying a cosine by x is relevant, but i can assure you that in level-tt space you need a sine, and that the period is 500 throughout (this was arrived at independently by three different people). If you're fitting to slope like Gingko was (dtt/dlevel) you need a cosine (derivative of sine). I tried plotting your function and once i changed the period to 500 and cos to sin it worked pretty well, as i said.

With the addition of wizzszz's discovery and Gingko's fit i'm declaring the low range (< 3k) completely finished. The error is less than 1% through nearly the entire range now, probably even lower for truly accurate data points.

Now, the high range...
 
Doer please clear your private message box, have been trying to contact you.
 
PE|EU WikiTools 0.00001beta

Ok, i got tired of playing with function matching and it seemed like jdegre was having all the fun making the kind of utility i had in mind when i started on this skill chip project, so i spent some time working on enhancing the wiki's scope as i have been intending to do. What i have now is very simple and crude, but it uses data from the wiki to give a value analysis of an avatar's skills. This takes advantage of the work already done on wiki to support several formats of skill capture utilities and hopefully will increase awareness of and collaboration on the wiki, which is a good thing. :)

Anyway, here's a tool for calculating skill values. It doesn't yet take into account the market value of ESIs, but that will be easy to add now that the darn thing functions. :yay:

Please let me know if you get poor results: i haven't tested its output except to see that it looks reasonable.

The fit that it uses to map from skill level to tt looks like this:
tool-fit.png


It currently only will return values for skills up to 8k because there's not enough direct data above that yet to even assume anything about the values. I'll take another look at the slope approach to the upper range Real Soon Now. ;)

zzzzzzz
 
Hey thats me in the link! So my avatar skills is worth 44650 minus ESI (44650-18400=26250). Not too bad for a nondepositer :yay: Anyway, IMO you and jderge should work together instead of compete ;). I helped jderge out to parse the data from wiki more easy. I am not sure how you did it, but you are free to use the same method. Just drop me a PM for details.

Tnx for the great work !
 
Witte said:
Hey thats me in the link! So my avatar skills is worth 44650 minus ESI (44650-18400=26250). Not too bad for a nondepositer :yay: Anyway, IMO you and jderge should work together instead of compete ;). I helped jderge out to parse the data from wiki more easy. I am not sure how you did it, but you are free to use the same method. Just drop me a PM for details.

Tnx for the great work !
Click to expand...

uh?? compete?? lol what do yo mean by "compete"?? :laugh:

doer an I have been working together in that function for some time now (with the help of many other people too).
it just happens that, for the purposes I had in mind, the approximations that we had were good enough (imo), so I started to work on one of the real applications of this whole thing (the chipping optimizer); the other application is the skill calculator that doer has done, which if awesome btw... so, where's the competition? :)

great work, doer.

/jdegre.
 
jdegre said:
uh?? compete?? lol what do yo mean by "compete"?? :laugh:

doer an I have been working together in that function for some time now (with the help of many other people too).
it just happens that, for the purposes I had in mind, the approximations that we had were good enough (imo), so I started to work on one of the real applications of this whole thing (the chipping optimizer); the other application is the skill calculator that doer has done, which if awesome btw... so, where's the competition? :)

great work, doer.

/jdegre.
Click to expand...


Well it was more of a joke ;). But now there are two separate sites with similar functionality, it could easely be integrated into one.
 
Witte said:
Well it was more of a joke ;). But now there are two separate sites with similar functionality, it could easely be integrated into one.
Click to expand...

Where's the fun in that? :D Well, i needed an excuse to learn python! OTOH i'm not as excited about learning the JS and such needed for AJAXish niceties everyone expects these days.

Actually, the site is down this morning. I'd like to claim it was slash-dotted but obviously that's not the case. ;) Hopefully, whatever is wrong with the host will be fixed soon: all i can get to is the usage stats right now.
 
Solaris said:
Cool work done here :D
reps on the way :silly2:
Click to expand...

Thanks :)

I went ahead and added a function segment to cover up to 16k skills. It's based on the rather sparse slope data i have for 8k+.

tool-slopefit.png


It predicts somewhat less than 1250 PEDs (one full ESI) for 8k and is off from 2500 PEDs (two ESIs) for 10k by about the same amount. I'm not sure if that means it's wrong by that much -- we have no reason to believe that 8k is an ESI now and 10k, two, except that it makes some sense historically.

Next, ESI price will be included.
 
New and improved, it calculates how much ESI you need, what it will cost, and what the net value is considering the ESI cost.

More to come (another day).
 
In the old days (before esi change), the curve would become linear after aproximately 7k skills.

I remember some tests done where highly skilled players tested their stats with in/out chipping. Results was that above 7k skills, you would fill one old esi of 10 ped for every 40 skillpoints.

Thus you needed about 250 peds chip to go from 7000 to 8000 skills.

I have not verified this by search of old data - just rambling as I remember it vaguely. :)
 
Solaris said:
In the old days (before esi change), the curve would become linear after aproximately 7k skills.

I remember some tests done where highly skilled players tested their stats with in/out chipping. Results was that above 7k skills, you would fill one old esi of 10 ped for every 40 skillpoints.

Thus you needed about 250 peds chip to go from 7000 to 8000 skills.

I have not verified this by search of old data - just rambling as I remember it vaguely. :)
Click to expand...

Yes they were good old days

Currently you'll need arround 480 PED chip to go from 7000 to 8000 skills :mad:
 
MG Mighty said:
Yes they were good old days

Currently you'll need arround 480 PED chip to go from 7000 to 8000 skills :mad:
Click to expand...

Yep that sounds pretty accurate.. checked auc a few mins ago and with a 200 PED lwt chip i would go from 6500 to 6970 or something like that.. it's crazy..

This is from the week when the new system was introduced:
(1124 ped hg chip... takes me from 2678 to 9986 lvls)

 
Last edited:
Fjolla said:
Yep that sounds pretty accurate.. checked auc a few mins ago and with a 200 PED lwt chip i would go from 6500 to 6970 or something like that.. it's crazy..

This is from the week when the new system was introduced:
(1124 ped hg chip... takes me from 2678 to 9986 lvls)

Click to expand...


Yes at that high levels it comes evident the change

Currently a 1124.8 PED chip of Marksmanship would take somebody from
level 2340 to level 8078

big difference I would say when comparing that old increase of
level 2678 to level 9986
 
MG Mighty said:
Yes at that high levels it comes evident the change

Currently a 1124.8 PED chip of Marksmanship would take somebody from
level 2340 to level 8078

big difference I would say when comparing that old increase of
level 2678 to level 9986
Click to expand...

yes there is a very big difference.
8078 is nowhere near 9986... it will for sure be another couple of thousands to chip it there.
 
Fjolla said:
Yep that sounds pretty accurate.. checked auc a few mins ago and with a 200 PED lwt chip i would go from 6500 to 6970 or something like that.. it's crazy..

This is from the week when the new system was introduced:
(1124 ped hg chip... takes me from 2678 to 9986 lvls)

Click to expand...

We need a 1124 PED chip to go through auction now so we can get exact data for such high levels. Well i guess the good news about that is that it shows that before the skill VU, 10k skill wasn't 1250 PEDs (one full ESI) as was sometimes thought. It was apparently more like ~1150.

8k skill is now around 1117 PEDs, so 8k is indeed the new 10k.

I fixed a bug jdegre caught in my skill worth utility so that the net worth would sum incorrectly. Thanks jdegre for testing it.

Try it, people. Break it, and tell me what's wrong. :)
 
Great tool! I was using it all of yesterday, and hoped to today, but now it's throwing errors at me, for any avatar i enter.

Also, would it be possible to make the output columns sortable?

I'll definately keep using the tool, great work!
 
mythicpunk said:
Great tool! I was using it all of yesterday, and hoped to today, but now it's throwing errors at me, for any avatar i enter.

Also, would it be possible to make the output columns sortable?

I'll definately keep using the tool, great work!
Click to expand...

:ahh:

a) Use a development branch
b) Stop programming before going to sleep

Working again. :)

MGMighty has data that suggests the prediction is off by 3% by the time it's just over 8k level (it's much better up to 7.5k). Because the fit works better with the data i have than with that one point i'll leave it as is for now, but when i get more data i'll tweak the fit. Just be aware (as is mentioned on the page, i think) that over 8k there is a lot more uncertainty in the values.

Edit: Sorting the columns is something i want to allow (as well as entering custom markup values), but it required some code refactoring and caching of avatar data, which i have been working on behind the scenes. Hopefully it will come soon.
 
Last edited:
Wow Doer,

wonderful tool made in no time... respect! Works just fine for me and the upload from the wiki is cool stuff :)

If I had more time on my hands atm I would have responded a bit more but I have not :( - at least not now.

Now we need jdegre's chipping optimizer added to this :D
 
And one final post in this thread about the new utility: i think the skill value calculator portion is done -- unless someone needs something more than what it can do. It refuses to guess for level 16k and up because we really have no idea what those values would be. From 7.5k to 16k the error will increase a lot, probably reaching 10% or more by 16k. Below 7.5k the error should be less than 1% almost everywhere.

I moved the script to be the default page handler, so go to http://esi.coolinc.info/ to access it now.

The original purpose of this thread -- to find the ESI tt needed to extract for any level of skill -- is still unfinished. If you have skills over 7k and find a very big chip on auction for those skills, post here with your beginning level, the tt value of the chip, and what the final level would be as described on the chip. You can also check to see how the calculator does so far, like this:
http://esi.coolinc.info/?level=7500&chip=449.99
Or you can go to the tools page and enter the values in the calculator yourself.

This kind of information lets us add another black diamond to this plot:
tool-fit.png

And that lets us find a better way to guess at values that aren't known.

Also useful is information about the slope of the skill gain at levels above 7.5k. In this case, giving the same information for a few chips that give between 15-30 levels will be helpful. In this case, the data added goes on this plot:
tool-slopefit.png
 
Status
Back
Top