Poe Price Check – Check Single Item and Stash Tab Items with Ease

@World_Destructor

I was trying to combine the following jewel mods according to:

https://pathofexile.gamepedia.com/Jewel

#% increased Attack Speed
#% increased Cast Speed
#% increased Spell Damage
+#% to Critical Strike Chance
+#% to Critical Strike Multiplier

However, it is messy if we want to aggregate them, e.g. for +#% to Critical Strike Chance, we will have:

'#% increased Critical Strike Chance with Cold Skills',
'#% increased Critical Strike Chance with One Handed Melee Weapons',
'#% increased Critical Strike Chance with Fire Skills',
'#% increased Critical Strike Chance with Two Handed Melee Weapons',
'#% increased Weapon Critical Strike Chance while Dual Wielding',
'#% increased Critical Strike Chance with Lightning Skills',
'#% increased Critical Strike Chance for Spells',
'#% increased Global Critical Strike Chance',
'#% increased Melee Critical Strike Chance',
'#% increased Critical Strike Chance with Elemental Skills'

For example, it is probably not wise to aggregate spells crit chance with melee crit chance. Does Fire skills include both attack and spell?...

Do you have any suggestion?

Fire skills apply to any skills tagged with Fire, so any skill in here:
https://pathofexile.gamepedia.com/Category:Fire_(gem_tag)
That includes Fire Strike, Burning Arrow, Flameblast and Fire Trap.

To solve that you need to catagorize the mods into:
[attack] critical strike chance
[spell] critical strike chance
And the skills or fire/cold/lightning damage should be handled as either or both.

For example:
'12% increased Weapon Critical Strike Chance while Dual Wielding',
'14% increased Critical Strike Chance with Lightning Skills',
'10% increased Critical Strike Chance for Spells',

[attack] critical strike chance 26%
[spell] critical strike chance 24%

Or:

15% increased lightning damage
16% increased Spell Damage while wielding a Staff
14% increased Melee Damage

[attack] lightning damage 31%
[spell] lightning damage 30%

I also want to suggest to increase the accuracy of pricechecking with calculating the probability for the item to roll equally as good or better.
For example:


With poeprices.info the recommended price is just 16c but it's a top 0,34% roll so I think it's probably worth triple of that, explained below.

maximum life rolls 40-70, so the chance for it is:
5 (equal or higher is 66-67-68-69-70) / 31 (70-40+1) x100 = 16%
Maximum % increased ES rolls 240-260% so 246% chance is:
15/21x100=71%
38% lightning resistance:
3/11x100=27%
chaos resistance:
1/13x100=7,7%

The calculation for it to roll equal or better would be to multiply it all without the x100:
5/31*15/21*3/11*1/13=0,0024 which's 0,24%.
I'd argue that the increased % ES is relatively insignificant compared to the other rolls though.
So I'd say it'll be:
5/31*3/11*1/13=0,0034 which's 0,34%
The chance for an equal or better Esh's Visage like this one is just 0,34%.
That'd mean that it'd take 294 divines or new Esh's Mirages (0,0034^-1) to get an equal or better one.

EDIT: On second thought, it probably isn't just triple the value, lol (priced it 2 ex for now).
I'm not entirely sure the math's correct on the chance for rolling equal or better though, but that shouldn't matter because it's still a good comparative way to compare rolls with.

@World_Destructor,

Your expertise is appreciated. For jewels, I have added the mods for crit chance attack and crit chance spell according to your suggestion. The same also goes to crit multiplier. As the first step, I will make the newly added mods not check by default. But it should pride ways to price based on aggregated mods. Later when the calculation works right, I will make the highest aggregated mod checked by default. The first step can be seen on the webpage as earlier as tomorrow, and the second step will follow.

The unique item in your example happens to be only two similar item been found. I have made a change to search for items which updated within 1 weeks. This will reflect more on the recent prices. I am watching whether this makes things better. Another way is to widen the search criterion if less than 5 similar items are found.

The probability idea for equal or better is brilliant! I don't know whether the rolling probability is the same across all the values from low to high. Since the mods' values are already included in the algorithm, the probability might be redundant. I am not sure yet, but will do some test.

I forget it every time to say but I'd really appreciate if you'd correct the quality by default to 20% just like poe.trade does.
Comparing rolls with variable quality isn't such a good idea as you might know from experience with using your own website for pricechecking.
An exception to this would be corrupted items though because they can't be modified on their quality.

I tried to widen the search but there were only 38 or so for sale at that time (implying it's not such a common item) and they weren't rolled even nearly as good (probably because they were sold) so that's why a % chance to roll equal or better would've been a nice tool to use to extrapolate it's possible value.
Maybe with that math you can even make graphs and such if you would want that, you probably need to apply a correction for the logarithmic scale though that you'd get from the calculation (a roll with 5% or equally good chance is 2x easier to get than a roll with 2,5% or equally good chance).

Expertise? xD
Well, I'm modest but sure, glad to help.
I help you to help me to pricecheck accurately anyway. :)

EDIT: To clarify if you don't get what I mean, imagine there are only 3 items for sale for a unique item and the lightning resistance can roll from 30% to 45%.
The 3 that are for sale have:
33% for 30c
36% for 32c
39% for 37c

My item rolled 43% so I can't pricecheck unless I can extrapolate:
33% = (16-3)/16 = 0,8125
36% = (16-6)/16 = 0,625 (roughly 25% less chance relative to 32%)
39% = (16-9)/16 = 0,4375 (roughly 35% less chance relative to 35%)
43% = (16-13)/16 = 0,1875 (over 50% less chance relative to 38%)

It's possible to roughly estimate the value of my item now:
33%=30c, relative difference between 33% and 39% is roughly 50% less (0,8125 and 0,4375), difference between 33% and 39% in price is 7c and since 43% is relatively 50% less chance than 39% (0,4375 and 0,1875) similar to 33% and 39% the estimated value would be 37c+7c which's 42c.
I deliberately kept the math simplified (since it will actually require a logarithmic correction) so it's not exactly correct.
You can also offer an alternative to include the price of divine orbs in the calculation but that generally doesn't correlate well and can be confusing, only if the item is exceptionally rare it'll somewhat correlate.

I would really like to be able to see data of the most valuable corrupted implicits versus how many of them are listed on poe.trade, very similar to what you have on the website from perandus league about the enchants on helmets. I would appreciate some input on how I could maybe put together my own program so that I can input what unique item I have, and then get results on all of the values of each implicit.

@World_Destructor,

To calculate the attributes with 20% quality, I think I need to first get the base of the item and then 1% quality increases 1% base. After this, add the flat increased values back and multiply it with the percentage multipliers. I am slowly implementing this. One thing not clear is that where ggg does the round to make the value an integer (most of the time).

For the percentage, really cool idea. If the value reaches max-1 or max, the value may increase exponentially for some mod on an item. The extrapolation would not always work. Similarly, at the lower end of the range, the effect can be flat instead of even linear. This may involve too many rules to set up. We are working on models for rare now. Maybe after this we should work on the unique items. Let's me think about this.

"

Ravenstaver wrote:

I would really like to be able to see data of the most valuable corrupted implicits versus how many of them are listed on poe.trade, very similar to what you have on the website from perandus league about the enchants on helmets. I would appreciate some input on how I could maybe put together my own program so that I can input what unique item I have, and then get results on all of the values of each implicit.

Back in Perandus, I think I did the enchanted mods with the frequency count (popularity) and cost (selling price).



For corrupted implicits, I think it is better to count the most valuable mods by Subtype (e.g. Ring, Helmet, Boot, etc.) The easiest way is to sort them by the median prices within one Subtype. It will not be very accurate, while it can show some interesting findings. You can even do this for each unique item. Maybe can also plot the number of items verse median value on a 2D plot to check whether the rarer an implicit mod the more expensive it gets.

"

SlugPranker wrote:

@World_Destructor,

For the percentage, really cool idea. If the value reaches max-1 or max, the value may increase exponentially for some mod on an item. The extrapolation would not always work. Similarly, at the lower end of the range, the effect can be flat instead of even linear. This may involve too many rules to set up. We are working on models for rare now. Maybe after this we should work on the unique items. Let's me think about this.

The funny thing is that you exactly describe what would turn out to be the model you get from using the calculations I suggest.
For example, a unique can roll from 1 to 10% (10 possible numbers).
1% is priced 11c and 6% is priced 16c.
Then applying my extrapolation calculation for a 10% roll would give a price of:
1%=10/10=1 (priced 11c)
6%=5/10=0,5 (2x less likely to roll equal or higher than 1%, priced 16c)
10%=1/10=0,1 (5x less likely to roll equal or higher than 6%, extrapolation price: 5/2*5c=12,5c more than 1% so 12,5c+11c=24c rounded up)

If it was linearly then if 1% is 11c and 6% is 16c then 10%=20c but it's calculated to be 24c.
It's a bit more advanced statistics.
To really use it correctly you need to use logarithms though, this isn't 100% correct but for clarity sake.

@World_Destructor

Yes, that right. There are two things I am working on. (1) the mod ranges kinda changes among leagues, and also over time. Need a reliable source to get the correct numbers; (I have a fraction, but not very reliable I guess) (2) For a certain mod value, there are different prices from different sellers most of the time, hence some averaging, median, etc. need to be used. An alternative I can think of is very close to our effort to build models on rare items. So in the same time we are exploring that possibility. I will keep you updated while we are making progress. Thanks again for the very good suggestion.

Btw, please read more about our modeling effort here if interested.

[Removed by Support]

Site was experiencing some performance issues which caused return of result very slow. Now the speed has been improved.