Calculating MPR (in General)


horseguy001 January 21 2007 6:01 PM EST

After scanning the Wiki and Pixels page/CBstuff for quite some time, I can't seem to find any calculations for figuring out how much exp = MPR.

I mean, I did some rough ones myself with some BA I burned today, but I don't know if MPR is based on total experience or trained stats. If it's trained stats instead of total experience then my numbers will be skewed.

Any help will be appreciated, as I want to see how much exp I have to play with on a 2 minion NCB I am considering at any given MPR level.

If I missed an obvious link to this info...sorry :D

th00p January 21 2007 6:03 PM EST

I don't believe anyone has done any research on this before, but I don't think it's on trained stats I believe it's on trained exp. Not that it really matters either way...

horseguy001 January 21 2007 6:21 PM EST

Well on some quick, rough calculations this is what I got:

For the Space Knights, exp/mpr = 42.1815
For Mister Farmer, exp/mpr = 9.7235

So exp/point does have profound effects on MPR, which means I am very confused about how to create a formula...it's been awhile since I did anything like this, and I don't think I have any old calculus text books lying around anymore.

I will come at it from another angle and see if I can get something a little better, but I doubt it.

Flamey January 21 2007 6:28 PM EST

43.1 here.

I think it stays around the 40-ish area.

Get a few more examples, I don't think it's as simple as XP/MPR.

Lochnivar January 21 2007 6:41 PM EST

just plug a bunch of different xp/mpr rates into excel and have it figure out a linear equation.
Specifically xp/mpr rate versus character MPR

It worked beautifully in CB1 for the then non-linear x costs on weapons

AdminNightStrike January 21 2007 6:49 PM EST

It's XP / minion, summed.

AdminNightStrike January 21 2007 6:52 PM EST

Example:

Let's say that the equation is a simple sqrt(x).

One minion with 1,000,000 XP = MPR of 1,000

Two minions, 500,000 each = MPR of 707 + 707 = 1414.

Two minions, 900,000 and 100,000 each = MPR of 948 + 316 = 1264


Obviously, sqrt(x) is not the equation, as the results don't make sense. However, how that f(x) is used and combined on a per-minion basis is as I describe -- I think.

horseguy001 January 21 2007 6:55 PM EST

But what I want to know is how much exp will, for example, two 500 mpr minions have?

Unfortunatly it's been about 5 years since I have used Excel...lets see if its like riding a bike :D

AdminNightStrike January 21 2007 7:01 PM EST

I have no idea whatsoever

Zoglog[T] [big bucks] January 21 2007 7:05 PM EST

I'm really not sure it is that easy to check but to find out how much MPR each of your current minions contribute on your current team just find out the exp spent on each of them and divide by the total exp on the team. Finally, multiply the MPR by the percentage.
This will likely not be exact due to the fact that it doesn't seem 100% linear but it can give a rough estimate.
To try and find a sensible equation you would then need to take at least several points from intervals (a couple of your farms, some mid-level teams and some larger teams) and correlate them to find the relationship between exp and MPR.
Overall though I doubt without proper inside information that we'll be able to get anything exact, just like with most aspects of the game.

smallpau1 - Go Blues [Lower My Fees] January 21 2007 7:09 PM EST

my exp / MPR came out to 24.8

horseguy001 January 21 2007 7:16 PM EST

well based on what I have on excel so far the number of minions doesn't seem to impact the exp/mpr at all, its strictly based on total exp.

AdminJonathan January 21 2007 7:22 PM EST

it's f ( sum ( minion exp ) ), not sum ( f ( minion exp ) )

horseguy001 January 21 2007 7:41 PM EST

I am trying to get a formula that will allow me to calculate roughly the amount of experience I will have based on MPR.

For example, I have 1 mil MPR split evenly between 2 minions, which is 500k mpr each. They will have x amount of experience each to spend on training.

Well I do have a nifty graph on Excel, I have yet to do anything useful with the data :D

horseguy001 January 21 2007 7:46 PM EST

Which makes me realize the thread title should be calculating experience based on MPR :P

Lochnivar January 22 2007 3:03 PM EST

Based on my calculations (from an admittedly small data sample) a 500k MPR would come with about 19mil experience, or about 9.5mil exp/ minion.

If I had more data (specifically total exp trained vs MPR) i'd probably be able to say with more certainty. That said my current formula yields an excellent r value.

Yukk January 22 2007 3:08 PM EST

Does MPR get reduced if you untrain something (at 95%) and
retrain that exp, losing the 5% ?
Or once the exp has been trained, does it count forever ?

Lochnivar January 22 2007 3:12 PM EST

mpr drops like a single persons standards the day before valentines.

which is to say yes, it reflects the total to trained exp, and as unlearning lowers EXP it also lowers MPR

horseguy001 January 22 2007 3:30 PM EST

Well what I am going to do is when I start my next NCB I am going to record data on a day to day basis and create a trend from that. It will be 120 data points from 1 to about 1 mil MPR, which should yield an accurate equation.

Lochnivar January 22 2007 3:33 PM EST

i usually just rely on around ten to twenty data points, i like my numbers to be raw (they're more nutritional that way)

Caedmon [Revenge of the Forgers] January 22 2007 3:58 PM EST

Run a log-linear regression...it's worked for everything else so far.

Lochnivar January 22 2007 4:05 PM EST

actually i used a power series for this one but an r = 1 is always an encouraging sign. I'd really like to get some numbers to test it on (ie give me the mpr and see if the exp total is close)

Lochnivar January 22 2007 4:23 PM EST

For those of you wanting to randomly calculate EXP and varying MPR levels:
EXP = MPR*(1.4307*(MPR^0.2501))

That formula with within 100k total EXP when calculating Conundrum's total EXP based on the 1.87mil MPR posted on SNK3R's competition. Well below 1% error.

Have fun and if anyone wants to test it out and let me know how well it works I'd appreciate it.

Xenko January 22 2007 6:24 PM EST

I've been tracking the growth of my character Theseus from ~0 exp until now (~11.5 million exp, having almost 400 data points).

Using the basic trend lines, I get the following equation:
MPR = (4 E-30)*XP^5 - (1 E-22)*XP^4 + (1 E-15)*XP^3 - (8 E-9)*XP^2 + 0.0523*XP + 2013.2

Based on my data, that has an R^2 value of 0.9999. So it should give decent results. I have also noticed that MPR does not depend on the number of minions (although to cost to add minions does depends somewhat on the number of minions you already have, interestingly enough).

Anyways, hope that helps. I'll keep tracking my growth (I also track Max Tattoo, and the cost to hire minions).

I was tracking UC effect based on XP trained into it, but the recent change has basically killed all my data and made it useless I think... and of course I had no internet during the free re-training as I was in the process of moving. Bah!

AdminQBGentlemanLoser [{END}] January 25 2007 7:40 AM EST

mpr: 367,276

xp (unlearn)
2101622 (90) 2335135.556
1049794 (95) 1105046.316
2218403 (95) 2335161.053
489193 (95) 514940
177637 (95) 186986.3158
1455875 (90) 1617638.889
1455875 (90) 1617638.889
1455885 (90) 1617650
1536882 (95) 1617770.526
86 (95) 90.52631579
5914 (95) 6225.263158

total xp: 12954283.33

AdminNightStrike January 25 2007 7:47 AM EST

Use my javascript calculator :)

Xenko January 25 2007 7:57 AM EST

A link to said javascript calculator would be handy.

Lochnivar January 25 2007 10:23 AM EST

Compared to GL's numbers my little formula was 99.98% accurate.
Personally I thing that is pretty good for something drawn from 4 data points.

TheHatchetman January 25 2007 10:26 AM EST

With all the randomness around in this game, 98% accurate is brilliant. Going for 99.98%, I'd say that's as close to the sweet spot as we need :) Good job Lochnivar :)

horseguy001 January 25 2007 12:51 PM EST

Around 1 mil mpr its only off by about 5% as well, so I'd say its a pretty good formula :D

Lochnivar January 25 2007 1:17 PM EST

5%? Are you sure?

That seems a little high, on Conundrum it was 99.96% accurate.
It seems unlikely to have such an increase in error when numbers that are both higher and lower have 1/100th the margin of error.

AdminNightStrike January 25 2007 8:12 PM EST


"A link to said javascript calculator would be handy."

Shade hosts it: http://www.cbstuff.nl/cbtools/cbtools.html

AdminNightStrike January 31 2007 6:20 PM EST

This equation seems to be off by a lot. With an MPR of 264,626, I have 8,599,318 total XP, and the equation yields 132,609 for the estimated MPR. Any suggestions? Perhaps I type it wrong. I used the following javascript code:

var mpr = Math.floor((Math.pow(total,5) * 4 * Math.pow(10,-30)) - (Math.pow(total,4) * 1 * Math.pow(10,-22)) + (Math.pow(total,3) * 1 * Math.pow(10,-15)) - (Math.pow(total,2) * 8 * Math.pow(10,-9)) + (0.0523*total) + 2013.2)

Miandrital January 31 2007 6:25 PM EST

This equation is accurate to within 1%.
EXP = MPR*(1.4307*(MPR^0.2501))
or solving for MPR
MPR = (EXP/1.4307)^(1/1.2501)

Xenko January 31 2007 6:49 PM EST

The equation Miandrital gives seems to work well. I have a lot of data points, I just don't know how/where I can go to try and fit a formula to all the data I have. Excel's trendline is awful.

AdminNightStrike January 31 2007 6:52 PM EST

Using that method, I get 264,666. That's pretty accurate.

AdminNightStrike January 31 2007 6:53 PM EST

Kultur: for starters, don't use a polynomial regression line. Those are highly accurate for the data points you have, and horrible for data points you try to extrapolate.

Lochnivar January 31 2007 7:29 PM EST

"This equation is accurate to within 1%.
EXP = MPR*(1.4307*(MPR^0.2501))
or solving for MPR
MPR = (EXP/1.4307)^(1/1.2501)

--Miandrital, 6:25 PM EST"

Thanks for checking it... I was starting to wonder if my calculations were way off an nobody wanted to hurt my feelings.
And to think, I've failed at least 2 calculas courses in my life.
(mainly a result of not going to class or doing homework I admit)

Xenko January 31 2007 7:34 PM EST

"for starters, don't use a polynomial regression line. Those are highly accurate for the data points you have, and horrible for data points you try to extrapolate."

I believe I said that already. I want to know how to get a good regression for a non-linear data set.

AdminNightStrike January 31 2007 7:38 PM EST

"I believe I said that already. I want to know how to get a good regression for a non-linear data set"

Well, to be more accurate, what I meant was that polynomial regression equations of non-linear data give you very good results for interpolated data. It's when you try to extrapolate that you run into issues. So, your equations should work well for points falling between your collected points.

As for how to get logarithmic regression equations, try Mathcad, Mathmatica, or Maple V

Lochnivar January 31 2007 7:41 PM EST

Kultur I usually just try a couple of the ones in excel and see which has the best r value.
It only took 4 or 5 data points to get an r = 1 on the power series line so that's what I went with.

Incidentally most things in the game can be estimated this way. I find it particularly useful on weapon PTH and Armour + although the type of regression line does differ.

Caedmon [Revenge of the Forgers] January 31 2007 8:27 PM EST

Kultur,

Jonathan likes using power functions. Regress the x-values against (log y). Then, (log y)=m x + b gives y = B x^m.

Miandrital January 31 2007 9:13 PM EST

Oh yeah, I forgot to credit lochnivar for finding that formula, that was one of the "Great Unknowns" in cb.

Sorry loch, and mad props. :)

AdminNightStrike February 1 2007 1:51 AM EST

I added this to the calculator. It'll be up as soon as Shade uploads it.

Vicious Cat February 6 2007 1:42 PM EST

Just tried using the calculator, and it's not doing anything for me :-(
Using FF with Veri's latest script if that makes any difference

AdminNightStrike February 6 2007 2:06 PM EST

FF 2.0+ required, as the script uses JS 1.7.

AdminNightStrike February 6 2007 2:24 PM EST

Link: http://www.cbstuff.nl/cbtools/cbtools.html

Vicious Cat February 6 2007 4:45 PM EST

ah cheers.
I thought FF was supposed to upgrade itself automatically. Ne' mind - it's done now thanks :-)

AdminNightStrike February 6 2007 5:33 PM EST

No problem :)

All suggestions are welcome. I'm working on a way to display a box when you hover over an attribute giving information like possible effect for various things and the unlearn amount.

QBsutekh137 February 6 2007 6:03 PM EST

I like it! VERY spiffy!!!!

One question: I am confused by the XP entry at the bottom... I understand the final tally (it's the best part of the screen), but that part to the left is confusing...is that just a little scratch area to check a single stat? What does it do differently from the rest of the screen? Maybe it just needs it's own box/caption: "Scratchpad" type of thing.

Or, as is more likely, I am missing what it is for... *smile*

AdminNightStrike February 6 2007 8:41 PM EST

I thought I described it in the help text. Basically, it's a calculator to say I have this, I want this, and it will cost this. For instance, say you have 15m XP to spend, and your stat is level 100k. Put in 100k for Start, add 15m to the XP of 100k and put it in End, and the total will be calculated. Or, say you want to know the XP it will take to go from level 100k to level 200k. Put those values in, and you'll see the XP difference. So yes, it's a scratchpad area for figuring out what you can do.

Ultimately, I'd like to be able to incorporate it better with your current character. Like, click a button to submit some number to some area of that box. Something like that.
This thread is closed to new posts. However, you are welcome to reference it from a new thread; link this with the html <a href="/bboard/q-and-a-fetch-msg.tcl?msg_id=0020fT">Calculating MPR</a>