Fill up ESI data...

[actam]

ppl, some data I did get from right-clicking on Full ESIs in auction...
(starting level 1, the data will show the TT/skill ratio...)
This data , as you will see, it is greatly different from that one of CareBear's...

btw, feel free to integrate/use/elaborate the data at your wish

TT    Skills
0,01	10
0,02	19
0,03	32
0,05	49
0,06	53
0,07	66
0,09	77
0,10	85
0,12	100
0,15	122
0,18	143
0,19	149
0,21	162
0,30	216
0,32	227
0,33	233
0,37	255
0,40	268
0,47	299
0,49	308
0,51	316
0,53	324
0,56	335
0,57	339
0,60	350
0,63	361
0,65	367
0,73	393
0,77	404
0,80	412
0,82	418
0,83	420
0,89	435
0,94	447
0,99	458
1,00	461
1,02	465
1,12	486
1,13	488
1,36	535
1,40	543
1,45	554
1,54	573
1,76	626
1,80	637
1,94	677
2,04	707
2,29	783
2,45	828
3,04	946
3,11	957
3,44	1003
3,90	1067
4,01	1084
4,28	1128
4,74	1218
6,98	1522
7,36	1558
8,24	1656
8,77	1728
9,22	1790
9,98	1877
9,99	1878
13,09	2092
13,89	2153
15,94	2339
16,53	2380
16,85	2400
16,89	2402
17,15	2417
20,29	2556
36,74	3285
52,18	3642
65,92	3962
70,14	4018
79,91	4171
99,64	4475
102,95	4506
114,89	4627
188,38	5319

Graph:

 

[actam]

ops, forgot to had:
value of skill are implanting skills, not explanted!

So the 1,00TT-->461 skill its the one seen in auction (the fill up from 1 to 461 levels!)

 

[jdegre]

try the following equation (the dots are data gathered from auct, same as you did):

you must go for a much higher order poly than 2, if you're doing min square poly interpolation (highest that excel can go is 6, afaik).

/jdegre.

ps: "x" is skill level, "y" is ESI tt value

 

[Kyl]

yo, i tried to gather the relation between Skill on chip and chip TT value.

First i thought its an exponential plus some kind of trigonometric function (eg. sin).

The prob with the exponential is: it wont go throw the y=0 and x=0.

I stopped @ a polynomial 4th kind atm. But i still don't get a usable function.

Then i have some other theories:

- I tried to gather some info about compression functions but my skill in math isn't high enuf to understand that formulas. Its like calculating the pressure curve if u press a liquid till u cant compress it further. That would be equivalent to filling a 1250 ped chip.

- Or we have to set up a diagram with the esi tt on the y-axis and the skill level on the x-axis. Draw the line and then calculate the surface beneath the curve to get some kind of usable value to go on with...but i am too noob in that one too ..

And there is another thing: I think the 'value' of the skills on the chip has to be calculated an other way then the skill in the avatar...

However, ill add my values this evening here (chipping in).

Kyl

 

[actam]

y = 7E-05x6 - 0,0003x5 - 0,0026x4 + 0,0072x3 - 0,0148x2 + 0,6229x + 6,1467


seems different?

 

[Kyl]

try the following equation (the dots are data gathered from auct, same as you did):

you must go for a much higher order poly than 2, if you're doing min square poly interpolation (highest that excel can go is 6, afaik).

/jdegre.

ah you here too

btw i am pretty sure that a 1250 ped chip can give 8k skills. No proof yet of course.

My excel only does 5th poly and this one dont work with my data. rubbish values..

I think i have to try office 12 soon....


Kyl

 

[actam]

yep, one thing i did not writh:

my graph is log-log!

so it seems line a line...

 

[Kyl]

yep, one thing i did not writh:

my graph is log-log!

so it seems line a line...

but why

 

[actam]

but why

because log-log fits the data better, imo.
But really everyone can one the graph they want, the only thing important is hte skills

 

[Witte]

yep, one thing i did not writh:

my graph is log-log!

so it seems line a line...

Nice work here. Logarithmics seems to make more sense to me then 5th degree polynomials. I will soon add this to wiki. Tnx for sharing.

 

[actam]

Nice work here. Logarithmics seems to make more sense to me then 5th degree polynomials. I will soon add this to wiki. Tnx for sharing.

np, but look that in the graph I had to use a polinomial on top of the log-log!
Hope now that I've not confused everybody

 

[jdegre]

y = 7E-05x6 - 0,0003x5 - 0,0026x4 + 0,0072x3 - 0,0148x2 + 0,6229x + 6,1467


seems different?

what do you mean? in your graph, the equation is only 2nd degree.

yep, one thing i did not writh:

my graph is log-log!

so it seems line a line...

that's interesting...i tried in the past to draw linear vs log, because i think it is more "logical", but i've not tried log-log... maybe i'll give it a try...

/jdegre.

 

[actam]

what do you mean? in your graph, the equation is only 2nd degree.
/jdegre.

yep 2nd degree on log-log...
the x and y in the graph is x=LN(TT) and y=LOG(Skills)

 

[Kyl]

yep 2nd degree on log-log...
the x and y in the graph is x=LN(TT) and y=LOG(Skills)

head cracking ..... but will try that with my data too of course

 

[jdegre]

Nice work here. Logarithmics seems to make more sense to me then 5th degree polynomials. I will soon add this to wiki. Tnx for sharing.

the polynomials don't have to make any sense it is just a generic interpolation function, using the minimum square method. it can interpolate _any_ function, given enough high order to the polynomial.

i agree though that making one (or both) or the axis logarithmic may make the function more "straight" , making the aproximation simpler.

afaik, it is the same approach used by carebear in his calculator, with the only difference that he used 3-4 polys, from different skill intervals (0-2000, 2000-4000, 4000-8000, and so...)

/jdegre.

 

[Witte]

the polynomials don't have to make any sense it is just a generic interpolation function, using the minimum square method. it can interpolate _any_ function, given enough high order to the polynomial.

i agree though that making one (or both) or the axis logarithmic may make the function more "straight" , making the aproximation simpler.

afaik, it is the same approach used by carebear in his calculator, with the only difference that he used 3-4 polys, from different skill intervals (0-2000, 2000-4000, 4000-8000, and so...)

/jdegre.

Yes i understand, but I even tried 10th degree polynomals and the result was still not very pleasing. And my intuituin (looking at the graph) tells me there is an expontential funtion involved somewhere.


EDIT: The problem with using higherdegree polynomal approximation with a negative coefficient for your higest order (like you have), its that eventually the graph will go down again and become negative.

When I use 10th order, this effect will be allot stronger, making the result even worse. See:

[br]Click to enlarge[/br]

 

[jdegre]

I have tried with the log-log and, honestly, I'm not sure if we gain anything vs. the linear-linear graph. Sure the graph is "straightened" when you apply the logs, so you need a smaller order poly to approximate it, but this does not mean anything special... it is not more accurate or anything. You still can approximate the log-log function with an order-2 poly or order-6 or however.

I guess the proof would be to extrapolate beyond the gathered data, and then check for correctness of the extrapolated data, for instance for very high skill levels.

/jdegre.

PS: Extrapolated data up to 16k skill level:

PS2: Just saw your edit witte. yes, you're right... the higher the order of the poly, the faster it deviates from the data points. that's why i think the best approach is to split the skill levels in chunks, as carebear did. other than that, we could involve the maths heavyweights, i.e., JimmyB

 

[Mathayus]

there's something confusing me...
if i got lets say 1k of a certain skill,and if u look to a chip that doubles your skill,i find chips worth around 10,11 ped tt...i got 1k,and with this chip i would get 2k..
but if i look a skill that i only have 5,6 or 10 for example and if i look for a chip with 1k skill its only worth around4,5 ped or not even that much....?| weird

 

[Doer]

Good work collecting all the data actam.

I'm not sure there's any reason to use log-log plots here: all we need is a good approximation of the curve for the sake of calculations, not to reverse-engineer the original software formula used. Since everyone will be working in normal (vs log) space a polynomial fit in normal space is simpler.

 

[Kyl]

new list with my data

TT        	Skills    	Avatar                   
0,01      	10        	Actam                    
0,02      	19        	Actam                    
0,03      	32        	Actam                    
0,05      	49        	Actam                    
0,06      	53        	Actam                    
0,07      	66        	Actam                    
0,09      	77        	Actam                    
0,1       	85        	Actam                    
0,12      	100       	Actam                    
0,15      	122       	Actam                    
0,18      	143       	Actam                    
0,19      	149       	Actam                    
0,21      	162       	Actam                    
0,3       	216       	Actam                    
0,32      	227       	Actam                    
0,33      	233       	Actam                    
0,37      	255       	Actam                    
0,39      	264       	Kyl                      
0,4       	268       	Actam                    
0,45      	291       	Kyl                      
0,46      	295       	Kyl                      
0,47      	299       	Actam                    
0,49      	308       	Actam                    
0,51      	316       	Actam                    
0,53      	324       	Actam                    
0,56      	335       	Actam                    
0,57      	339       	Actam                    
0,6       	350       	Actam                    
0,63      	361       	Actam                    
0,65      	367       	Actam                    
0,73      	393       	Actam                    
0,77      	404       	Actam                    
0,8       	412       	Actam                    
0,82      	418       	Actam                    
0,83      	420       	Actam                    
0,89      	435       	Actam                    
0,94      	447       	Actam                    
0,95      	450       	Kyl                      
0,99      	458       	Actam                    
1         	461       	Actam                    
1,02      	465       	Actam                    
1,12      	486       	Actam                    
1,13      	488       	Actam                    
1,34      	531       	Kyl                      
1,36      	535       	Actam                    
1,4       	543       	Actam                    
1,45      	554       	Actam                    
1,54      	573       	Actam                    
1,76      	626       	Actam                    
1,8       	637       	Actam                    
1,94      	677       	Actam                    
1,96      	683       	Kyl                      
2,04      	707       	Actam                    
2,29      	783       	Actam                    
2,44      	826       	Kyl                      
2,45      	828       	Actam                    
2,68      	882       	Kyl                      
2,74      	894       	Kyl                      
3,01      	941       	Kyl                      
3,03      	945       	Kyl                      
3,04      	946       	Actam                    
3,11      	957       	Actam                    
3,44      	1003      	Actam                    
3,59      	1024      	Kyl                      
3,9       	1067      	Actam                    
4,01      	1084      	Actam                    
4,28      	1128      	Actam                    
4,43      	1156      	Kyl                      
4,74      	1218      	Actam                    
5,06      	1284      	Kyl                      
5,78      	1401      	Kyl                      
5,84      	1408      	Kyl                      
6,03      	1431      	Kyl                      
6,98      	1522      	Actam                    
7,36      	1558      	Actam                    
7,95      	1620      	Kyl                      
8,24      	1656      	Actam                    
8,43      	1681      	Kyl                      
8,77      	1728      	Actam                    
8,92      	1749      	Kyl                      
9,07      	1770      	Kyl                      
9,22      	1790      	Actam                    
9,26      	1795      	Kyl                      
9,68      	1846      	Kyl                      
9,76      	1854      	Kyl                      
9,98      	1877      	Actam                    
9,99      	1878      	Actam                    
10,15     	1892      	Kyl                      
10,6      	1929      	Kyl                      
10,88     	1949      	Kyl                      
10,89     	1950      	Kyl                      
11,19     	1970      	Kyl                      
11,22     	1972      	Kyl                      
11,33     	1979      	Kyl                      
11,75     	2005      	Kyl                      
11,91     	2015      	Kyl                      
11,93     	2016      	Kyl                      
13,02     	2087      	Kyl                      
13,09     	2092      	Actam                    
13,31     	2109      	Kyl                      
13,89     	2153      	Actam                    
14,95     	2253      	Kyl                      
15,94     	2339      	Actam                    
16,47     	2376      	Kyl                      
16,53     	2380      	Actam                    
16,85     	2400      	Actam                    
16,89     	2402      	Actam                    
16,98     	2406      	Kyl                      
17,15     	2417      	Actam                    
20,29     	2556      	Actam                    
21,66     	2622      	Kyl                      
23,12     	2709      	Kyl                      
25,33     	2841      	Kyl                      
27,97     	2947      	Kyl                      
36,74     	3285      	Actam                    
41,65     	3416      	Kyl                      
46,05     	3509      	Kyl                      
47,29     	3533      	Kyl                      
49,25     	3573      	Kyl                      
52,18     	3642      	Actam                    
55,03     	3723      	Kyl                      
62,57     	3909      	Kyl                      
65,59     	3957      	Kyl                      
65,92     	3962      	Actam                    
70,14     	4018      	Actam                    
74,7      	4082      	Kyl                      
79,66     	4167      	Kyl                      
79,91     	4171      	Actam                    
88,94     	4346      	Kyl                      
99,64     	4475      	Actam                    
101,07    	4488      	Kyl                      
102,87    	4505      	Kyl                      
102,95    	4506      	Actam                    
103,14    	4508      	Kyl                      
114,89    	4627      	Actam                    
119,66    	4689      	Kyl                      
157,91    	5049      	Kyl                      
188,38    	5319      	Actam                    
190,91    	5341      	Kyl                      
201,37    	5414      	Kyl                      
226,18    	5535      	Kyl                      
261,17    	5750      	Kyl

 

[jdegre]

there's something confusing me...
if i got lets say 1k of a certain skill,and if u look to a chip that doubles your skill,i find chips worth around 10,11 ped tt...i got 1k,and with this chip i would get 2k..
but if i look a skill that i only have 5,6 or 10 for example and if i look for a chip with 1k skill its only worth around4,5 ped or not even that much....?| weird

No, it's not weird; what you have observed is exactly what those graphics try to show.

Basically, the "chip tt value" represents your skilling experience. It takes an equivalent skilling effort to fill an ESI from 0 to 10 ped, than from 10 ped to 20 ped.

However, the skill levels don't grow linearly. You go very fast from 0 to 2k, but it takes much much MUCH longer to go from 2k to 4k. This up-hill, is exactly what the graphics represent.

Therefore, from 0 to 1k you need tt=3.5ped (as you observed) but from 1k to 2k you need ~8.2 ped.

/jdegre.


PS: Kyl, I bet your last point is wrong:

618,53 6894 Kyl

Could you please double check it? I think it should be around 570 tt

 

[Kyl]

PS: Kyl, I bet your last point is wrong:

618,53 6894 Kyl

Could you please double check it? I think it should be around 570 tt

i guess you got me there

But honestly i am not sure about this one too but maybe you can help

Here is how i 'found' it:

I chipped out all my rifle in 4 steps with that results:

ESI TT used   	Level     	Skill lost	Skill remain   	ESI with Skill 	ESI back  	total back	Re Chip   	Lost TT   	Lost %    	ESI Needed
18,2      	7187      	52        	7135           	16,16          	0,24      	16,4      	7182      	1,8       	10,02%    	17,96     
337,38    	7135      	1068      	6067           	303,47         	0,18      	303,65    	7056      	33,73     	10,00%       	337,2     
314,27    	6067      	3615      	2452           	282,79         	0,04      	282,83    	5959      	31,44     	10,01%       	314,23    
71,42     	2452      	2451      	1              	16,11          	53,51     	69,62     	2351      	1,8       	10,05%       	17,91

all i could imagine was to calc the difference between start skill and re chip value, sum that up and subtract it from my starting skill value.

like that:

Level     	Re Chip   	Diff
7187      	7182      	5
7135      	7056      	79
6067      	5959      	108
2452      	2351      	101

7187 - 293 = 6894

Maybe this was way too nooob

 

[jdegre]

You cannot do that ("7187 - 293 = 6894"), since "skill levels" is a non-additive magnitude. However, "chip tt value" is additive (you can add, substract, take 10%, etc..)

From your data, we can get the following points:

2452 -> 17.91
6067 -> 332.14
7135 -> 669.34
7187 -> 687.30

Cheers,
/jdegre.

 

[Doer]

I busted out Matlab and played around with the data: i changed my mind about using log-log plots to fit it.

A 4th degree polynomial fit of a log-log plot is the lowest degree with relatively small residuals (slightly better than quadratic); it doesn't get any better with higher degrees. However, using such a fit equation to extrapolate shows a lot of divergence even by the last data points. On the other hand, trying to fit the entire data with up to ten degree polynomials without using logs is, as witte demonstrated, rather unsatisfactory.

It may be the breaking it up into segments and optimizing a polynomial fit for those segments would get better results.

I'm not sure that extrapolating above 8k skills is useful, in any case, because it's possible that the curve straightens out at some point. We simply don't know without some large data points (like a full tt ESI for example).

Edit: in short, nothing new to contribute -- just wanted to see for myself.

 

[Witte]

I got abit bored with my study (again) so thought lets solve this thing. In matlab, when you do:

plot (b, log(a./b)) (where a=ped and b= skill, so a./b = cost per skillpoint)

you get an almost linear relation for >1000. The coefficients are:

0.00054208(b)-6.2295. When you substract this from the plot above, you get all the leftover, which looks like:

[br]Click to enlarge[/br]

These are pretty much the slowdowns and the speedups. I havnt realy analised these yet, but its seems the selected value is not correct, as it doesnt fit in the series (actam?).

Well back to work 4 me

 

[actam]

The data is got from chips, and is some case same tt did give me different skill-gain (starting always from 1):

#1 tt 1.00 skills from 1 to 1000 (just making the numbers for an example)
#2 tt 1.00 skills from 1 to 998

I did get the first one and max the similar TTs

but problem aroused:

1) to not know if the chips where max full TT filled
2) do not know the 3rd TT number... (tt 1.002 or tt 1.000), can't use sweat-method because I nataurally was in auction...
3) I do have data and double data of many but not on all TT values

I will insert other data when other chips will be disponible in auction hoping to get a better precision

 

[Witte]

The data is got from chips, and is some case same tt did give me different skill-gain (starting always from 1):

#1 tt 1.00 skills from 1 to 1000 (just making the numbers for an example)
#2 tt 1.00 skills from 1 to 998

I did get the first one and max the similar TTs

but problem aroused:

1) to not know if the chips where max full TT filled
2) do not know the 3rd TT number... (tt 1.002 or tt 1.000), can't use sweat-method because I nataurally was in auction...
3) I do have data and double data of many but not on all TT values

I will insert other data when other chips will be disponible in auction hoping to get a better precision

The small variations arnt realy a problem, but the point I am talking about, 36,74 3285 seems to be about 2 ped off. When I make it 38.74, it seems to fit the series allot better.

 

[actam]

The small variations arnt realy a problem, but the point I am talking about, 36,74 3285 seems to be about 2 ped off. When I make it 38.74, it seems to fit the series allot better.

oh! might be a transcription error!

I did right down the TT and skill values as integers (3674/3285 for 36.74 -> 3285) and some times I do invert letters (and word)
So 3285 could have been 3825...
I did double check but some times sh*t happens

 

[Witte]

oh! might be a transcription error!

I did right down the TT and skill values as integers (3674/3285 for 36.74 -> 3285) and some times I do invert letters (and word)
So 3285 could have been 3825...
I did double check but some times sh*t happens

aah ok, so it could also be 37.64 ped. That also seem to fit better , but more data around that point is needed to know for sure.

 

[The187]

Good thread thx