Lucas Kell
19-02-13, 11:56
Yesterday I decided that it would be fun to play with the barter skill, to figure out how it is calculated. This is all in an efforts to give me a better understanding of the skill to determine how much is needed to be effective. I've seen a lot of number being thrown around the help channel about potential hard caps and effectiveness of the skill, but nothing solid. So I figured I'd take a look.
Now getting an initial (very rough) calculation proved to be easier than I thought. First you need to get the base cost of an item. For this I'm going to pick "Big Pack of construction Grease". Now I know the base price for this item is 12000. Getting the base price for an item may not be easy as its not written anywhere, but if you get a spy with no barter and 5 INT, look at an item and multiply its cost by 1.405747548, you'll get a close enough estimate.
So the calculation roughly is: =BASE * (0.710655 - (BRT * 0.165820) - (INT * 0.014210))
Nice and flat. So I thought, bit of polish to smooth off the rough edges and I'm done! Yeah right...
I decided to validate my formula with some helpful guys in the help channel. This taught me 2 valuable things:
Firstly, your INT may not be a solid number. Decimals are used for calculations, so that pesky implant taking away 1.22 INT really takes away 1.22 INT. Chucking the number 74.78 into INT instead of the flat 74 made the formula continue to work as intended so I jotted that down and carried on testing.
The second thing I discovered was more horrific. My formula didn't work on high levels of barter. In fact, with 182 barter It was 6% off, which is pretty big. So back to the drawing board to work the numbers.
After some considerable data gathering, I put together this graph of item price per level of barter:
http://i45.tinypic.com/2dikmc5.png
Interesting little kink. I decided it was probably best to plot the difference between each level, this would then allow me to see more easily how higher levels affect the barter differently. Here's where it all started to become clear:
http://i50.tinypic.com/2vv9ikj.png
As you can see from this the first 75 levels are completely flat. You have the same % saving for every single point. Once it hits 75 however it spikes, then tapers off. This is exactly the same for INT. When INT hits 75, it spikes, then tapers off. One interesting thing about these is that INT and barter don't affect each other. They appear to be applied as independent price reductions.
After seeing the above spike, I thought, let's generate something useful from all this data. Now points in barter are not 1:1. from 0 to 50 they are, but then the cost of skillpoints goes up, to 2:1, then 3:1, then 5:1. So lets chuck that into the mix and get ourselves an effectiveness chart. (This is the one you barter guys want as it shows you how effective each point is going forward).
http://i48.tinypic.com/4ux47n.png
So there's that. Effectiveness is drastically lower per raw skill point in the later numbers.
Any of you guys from the old NC1 days, I swear there was either a post, or some patchnotes, maybe even as far back as beta that explains why we would see this behaviour. They may not have been talking about barter, maybe about another skill, but my memory is telling me they made a change to the calculations to make this behaviour instead of a raw curve. If anyone can remember what patch or post it was, linky linky please ^_^.
There's still a lot more that I need to look at on this to generate an accurate method to predict barter savings, but I figured I'd share the work so far.
Enjoy!
Now getting an initial (very rough) calculation proved to be easier than I thought. First you need to get the base cost of an item. For this I'm going to pick "Big Pack of construction Grease". Now I know the base price for this item is 12000. Getting the base price for an item may not be easy as its not written anywhere, but if you get a spy with no barter and 5 INT, look at an item and multiply its cost by 1.405747548, you'll get a close enough estimate.
So the calculation roughly is: =BASE * (0.710655 - (BRT * 0.165820) - (INT * 0.014210))
Nice and flat. So I thought, bit of polish to smooth off the rough edges and I'm done! Yeah right...
I decided to validate my formula with some helpful guys in the help channel. This taught me 2 valuable things:
Firstly, your INT may not be a solid number. Decimals are used for calculations, so that pesky implant taking away 1.22 INT really takes away 1.22 INT. Chucking the number 74.78 into INT instead of the flat 74 made the formula continue to work as intended so I jotted that down and carried on testing.
The second thing I discovered was more horrific. My formula didn't work on high levels of barter. In fact, with 182 barter It was 6% off, which is pretty big. So back to the drawing board to work the numbers.
After some considerable data gathering, I put together this graph of item price per level of barter:
http://i45.tinypic.com/2dikmc5.png
Interesting little kink. I decided it was probably best to plot the difference between each level, this would then allow me to see more easily how higher levels affect the barter differently. Here's where it all started to become clear:
http://i50.tinypic.com/2vv9ikj.png
As you can see from this the first 75 levels are completely flat. You have the same % saving for every single point. Once it hits 75 however it spikes, then tapers off. This is exactly the same for INT. When INT hits 75, it spikes, then tapers off. One interesting thing about these is that INT and barter don't affect each other. They appear to be applied as independent price reductions.
After seeing the above spike, I thought, let's generate something useful from all this data. Now points in barter are not 1:1. from 0 to 50 they are, but then the cost of skillpoints goes up, to 2:1, then 3:1, then 5:1. So lets chuck that into the mix and get ourselves an effectiveness chart. (This is the one you barter guys want as it shows you how effective each point is going forward).
http://i48.tinypic.com/4ux47n.png
So there's that. Effectiveness is drastically lower per raw skill point in the later numbers.
Any of you guys from the old NC1 days, I swear there was either a post, or some patchnotes, maybe even as far back as beta that explains why we would see this behaviour. They may not have been talking about barter, maybe about another skill, but my memory is telling me they made a change to the calculations to make this behaviour instead of a raw curve. If anyone can remember what patch or post it was, linky linky please ^_^.
There's still a lot more that I need to look at on this to generate an accurate method to predict barter savings, but I figured I'd share the work so far.
Enjoy!