[X] My Assistant

While Tenboro was kind enough to give us various formulas involving our secondary stats, one of the things he has not made available was the formula to determine how many experience points it takes to raise a primary attribute one level. So unless Tenboro comes in here and flat out gives it to us, I'd like to use this thread to gather data in order to figure out a close approximation. Now you may ask why do I want this information? The same reason as the other info we've been given: projections, planning, and general decision-making.

I've been keeping track of how much experience it takes to get a stat to the next level from 11 to 33 (my lowest stat was at 10 and my current highest stat is at 32), which is a pretty small data set. I graphed the data using Ye Olde Spreadsheet and generated a trendline. Two, to be exact. The first is a Power, based on an observation made by Sayo Aisaka in a different thread. This resulted in the equation XPRequired = 0.831*StatLevel^2.055, where XPRequired is the amount of experience to increase the Primary Stat to StatLevel. The second was a third order polynomial wtih a zero intercept (second order line didn't follow the points as closely, and I don't want to go to a higher order without more data), resulting in XPRequired = 0.002*StatLevel^3 + 0.528*StatLevel^2 + 3.721*StatLevel. Botharen't quite accurate, since I have a small data set and I didn't have the foresight to track the info from the start. This is also only assuming that the experience needed is only reliant on the stat level and nothing else.

What I am now asking is if anyone would like to contribute to this endeavour. Post data we can use, do your own analysis, offer whatever insights you have, whatever.

Thanks for reading.

This post has been edited by cmal: Jul 30 2009, 22:16

Specter to my knowledge is doing this and has a shit load of data regarding this as well.

If it'll help, from lvl 199->200 it'll cost 999,999 exp. (IMG:[invalid] style_emoticons/default/tongue.gif)

Also, No trendline in Excel will correspond even remotely to the original data for very long; something to consider.

Another thing, if there is anyone who wants this more than everyone else, it's Sonic for his massive calculator. (IMG:[invalid] style_emoticons/default/rolleyes.gif)

Anyways, I have work in 4 hours, I should take a power nap. (IMG:[invalid] style_emoticons/default/sleep.gif)

This post has been edited by Spectre: Jul 31 2009, 11:06

If it'll help, from lvl 199->200 it'll cost 999,999 exp. (IMG:[invalid] style_emoticons/default/tongue.gif)

Also, No trendline in Excel will correspond even remotely to the original data for very long; something to consider.

Another thing, if there is anyone who wants this more than everyone else, it's Sonic for his massive calculator. (IMG:[invalid] style_emoticons/default/rolleyes.gif)

Anyways, I have work in 4 hours, I should take a power nap. (IMG:[invalid] style_emoticons/default/sleep.gif)

Should probably treat it as a piecewise function and use trends of best fit within the pieces. (Prolly coded something like if (level < 0 && level > 10).. (level < 11 && level > 20), etc.)

CODE

function get_exp_for_prlevel($level) {
    return round(***************************************************);
}

/evil

/evil

you sir are a fucking liar

Nope, it really does look like that. I just x'ed out 51 chars of math. (IMG:[invalid] style_emoticons/default/tongue.gif)

i didn't quote that part (IMG:[invalid] style_emoticons/default/tongue.gif)

/evil

Agreed. (IMG:[invalid] style_emoticons/default/tongue.gif)

Another thing, if there is anyone who wants this more than everyone else, it's Sonic for his massive calculator. (IMG:[invalid] style_emoticons/default/rolleyes.gif)

Well mostly, part of me is just curious to what it is.

This post has been edited by Sonic: Jul 31 2009, 13:32

Nope, it really does look like that. I just x'ed out 51 chars of math. (IMG:[invalid] style_emoticons/default/tongue.gif)

So it's most likely a [en.wikipedia.org] polynomial (IMG:[invalid] style_emoticons/default/biggrin.gif)

It should be possible to create a trend with a [en.wikipedia.org] regression analysis, given enough data points.

stat costs for 1-34 of a stat, not sure if i missed some numbers:
4 8 14 21 41 64 83 101 120 141 164 189 216 245 277 310 346 384 425 468 514 562 614 668 725 786 841 916 987 1061 1138 1220 1305, i've been keeping track of them to see how much i've spent on each stat (IMG:[invalid] style_emoticons/default/tongue.gif)

if you guys really want to do it for me and since I'm not that great with numbers, I could upload my excel sheet with nearly 100 data points, after I update whatever sonic has that I don't. (IMG:[invalid] style_emoticons/default/rolleyes.gif)

or rather, probably whore myself out to google, get a google account and set up google docs

This post has been edited by Spectre: Jul 31 2009, 23:53

Well, now we know:
1) Its not easily trendlined
2) Its got about 51 chars worth of math
3) Its likely a massive polynomial or stupidly complex simple arithmetic

I know abit of MATLAB, so I can try setting something up some stuff there since its a bit more powerful as far as crunching funky numbers goes once I can figure out more about the nature of this formula. If only I paid attention in Statistics and learned SAS properly....

Even if we can't figure out the exact formula, getting close is good enough. There's also a part of me that's curious about how Tenboro is doing it; now I want to know for the sake of knowing.

Ever heard of a program called origin. It's hell of a lot better than excel in terms of statistical analysis. MATLAB is good if you are good at programming the damn thing (I failed at using MATLAB). But if need statistical analysis or any kind of analysis on raw data, origin is the way to go. I still have the software from when I was in undergrad, if some wants to give me the data and tell what kind of info they want from the data, I'll be happy to help. Don't expect something right away though because I'm busy with med school right now.

stat costs for 1-34 of a stat, not sure if i missed some numbers:
4 8 14 21 41 64 83 101 120 141 164 189 216 245 277 310 346 384 425 468 514 562 614 668 725 786 841 916 987 1061 1138 1220 1305, i've been keeping track of them to see how much i've spent on each stat (IMG:[invalid] style_emoticons/default/tongue.gif)

There are two or three missing from the beginning. Unless it was due to a typographical error, we can extrapolate which values are missing based on what we know so far and try to fill them in later.

Ever heard of a program called origin. It's hell of a lot better than excel in terms of statistical analysis. MATLAB is good if you are good at programming the damn thing (I failed at using MATLAB). But if need statistical analysis or any kind of analysis on raw data, origin is the way to go. I still have the software from when I was in undergrad, if some wants to give me the data and tell what kind of info they want from the data, I'll be happy to help. Don't expect something right away though because I'm busy with med school right now.

Everything is better at statistical analysis than Excel; just a matter of by how much. I've only worked with Mathcad, Maple, and SAS through my undergrad math classes and MATLAB through my controls classes and linear algebra. I was an engineer, so my exposure was pretty brief to the first three and I'm more competent with the last. I'll see if I can get a hold of one of my old math department buddies and see if he can lend us a hand, since he's probably just playing WoW instead of doing his homework.

CODE

function get_exp_for_prlevel($level) {
    return round(***************************************************);
}

I just remembered something...is that in standard format or in [en.wikipedia.org] RPN? It just seems like you would throw in a curveball at some point to confuse people (IMG:[invalid] style_emoticons/default/tongue.gif)

I'm going going to rage so hard if its in RPN. Maybe even fling some poop.

Right now, I've got the equation set up as a 6th order polynomial, but the first two terms are ridiculously small and what appears to be a 0 (respectively). Its off by a few points here and there as far as predicting the values go, but its pretty close. I still have a pretty small data set that's missing a few values in the beginning and also a few rounding errors, but that's to be expected.

here's some more numbers, values from 1 of a stat to 61 of a stat (never figured out which values were missing people can compare), 4 8 14 21 41 64 83 101 120 141 164 189 216 245 277 310 346 384 425 468 514 562 614 668 725 786 841 916 987 1061 1138 1220 1305 1395 1489 1587 1690 1798 1910 2028 2150 2279 2412 2552 2698 2849 3008 3173 3344 3524 3710 3904 4106 4316 4535 4762 4999 5245 5501

I also have this messy scrap that I use to estimate total costs of training a stat:
4 1pt
4+8= 12 2pt
4+8+14= 26 3pt
47 4 pts
88 5 pts
152 6 pts
235 7 pts
4+8+14+21+41+64+83+101= 336 8 pts
456 9 pts
597 10 pts
761 11 pts
950 12 pts
4+8+14+21+41+64+83+101+120+141+164+189+216= 1166 13 pts
4+8+14+21+41+64+83+101+120+141+164+189+216+245= 1411 14th pt
14411+277= 1688 15th pt
1411+277+310= 1998 16th
1411+277+310+346= 2344 17th
1411+277+310+346+384= 2728 18th
1411+277+310+346+384+425= 3153 19th
1411+277+310+346+384+425+468= 3621 20th
1411+277+310+346+384+425+468+514= 4135 21st
1411+277+310+346+384+425+468+514+562= 4697 22nd
1411+277+310+346+384+425+468+514+562+614= 5311 23rd
1411+277+310+346+384+425+468+514+562+614+668= 5979 24th
1411+277+310+346+384+425+468+514+562+614+668+725= 6704 25th
1411+277+310+346+384+425+468+514+562+614+668+725+786= 7490 26th
1411+277+310+346+384+425+468+514+562+614+668+725+786+841= 8331 27th
1411+277+310+346+384+425+468+514+562+614+668+725+786+841+916= 9247 28th pt
+986= 10233 29th pt
+1061= 11294 30th
+1138= 12432 31st
+1220= 13652 32nd
+1305= 14957 33rd
+1395= 16352 34th
+1395?= 17747 35th?
+1495?= 19242 36th?
+1489?= 20731 37th
+1587?= 22318 38th
+1690?= 24008 39th
+1798?= 25806 40th
+1910?= 27716 41st
+2028?= 29744 42nd
+2150?= 31894 43rd
+2279?= 34173 44th
+2412?= 36585 45
+2552?= 39137 46
+2698?= 41835 47
+2849?= 44684 48
+3008?= 47692 49
+3173?= 50865 50
+3344?= 54209 51
+3524?= 57733 52
+3710?= 61443 53
+3904?= 65347 54
+4106?= 69453 55
+4316?= 73769 56
+4535?= 78304 57
+4762?= 83066 58
+4999?= 88065 59
+5245?= 93310 60
+5501?= 98811 61

The ?= means not exact value since I didn't have the value available at the time so I just added a previous value to get to an estimate (although some values are missing at start so it wouldn't be exact anyway, but it's pretty close)

The reason I have all this is because I keep the cost of my dex 75% of str costs (but then i add 1 to it since seem to miss a bit lately), agi 1/3 of dex costs (but then i add 1 to it since seem to miss a bit lately), end same as dex, wis 1/4th of agi (value not cost total e.g. 36 agi 9 wis), then int 1.5x value of wis (e.g. 12 int 8 wis) (IMG:[invalid] style_emoticons/default/tongue.gif)

current base stats at lev 53, 60 str, 56 dex, 40 agi, 55 end, 13 int, 9 wis

This post has been edited by XMike: Aug 25 2009, 04:34

I bet tenboro just randomly made up the numbers for each level and then claimed there's a forumla just to mess with us.

No, there's definitely a formula there, I can somewhat see it based on the data, I just can't say exactly what it is. That said, it really is fairly complex just looking at the numbers~