Difference between revisions of "Talk:Quality"

From RimWorld Wiki
Jump to navigation Jump to search
Line 96: Line 96:
 
|}
 
|}
 
:::I don't think updating the numbers on the page is a super high priority, but go ahead if you want.[[User:Cheldra|Cheldra]] ([[User talk:Cheldra|talk]]) 11:50, 4 October 2021 (UTC)
 
:::I don't think updating the numbers on the page is a super high priority, but go ahead if you want.[[User:Cheldra|Cheldra]] ([[User talk:Cheldra|talk]]) 11:50, 4 October 2021 (UTC)
 +
 +
::::Done. And although most of the changes were minor the work wasn't for nothing. When calculating the average results I noticed that they were significantly of in the inspired specialist table. In hindsight fairly obvious. They were above 6, while 6.00 would mean 100% legendary items. [[User:Ickputzdirwech|Ickputzdirwech]] ([[User talk:Ickputzdirwech|talk]]) 10:43, 5 October 2021 (UTC)

Revision as of 10:43, 5 October 2021

This table is outdated. Game code of 1.0 now clamps quality at 5 (masterwork) unless the pawn crafter has an "inspiration" which adds 2 grades to the crafted item. So crafting legendaries without inspiration isn't possible in 1.0, with inspiration pawn needs to craft excellent or masterwork tier for it to be internally promoted to legendary.—Preceding unsigned comment added by ‎Evzhy (talkcontribs) 21:45, 7 January 2019 (UTC)

I know the table is outdated, but unless someone good at datamining or experimenting manages to grab the values, we don't have the values for the new table so we can't make a new one.
Also, this talk page is broken, and you will need to manually sign by adding ~~~~ after your message; if you consent I will refresh this page by deleting it. PigeonGuru, Strategist ( talk | contribs | assist ) 14:06, 7 January 2019 (UTC)
No, we don't delete comments, learn to use Template:Unsigned. 21:56, 7 January 2019 (UTC)
I meant to say that unsigned comments aren't the only reason- the page is not updated and hence needs to be refreshed. But yeah, thanks for telling me that there is that template. PigeonGuru, Strategist ( talk | contribs | assist ) 22:41, 7 January 2019 (UTC)

Ok done [1] —Preceding unsigned comment added by ‎Evzhy (talkcontribs) 01:22, 9 January 2019 (UTC)

I'm new to the game, but how come noone here knew about a debug screen "Quality Generation Data" that shows all chances? —Preceding unsigned comment added by ‎Evzhy (talkcontribs) 13:18, 9 January 2019 (UTC)

So there is such a screen in dev mode. Maybe when I have the time I'll update the table. Alternatively, you can also help update it. PigeonGuru, Strategist ( talk | contribs | assist ) 14:02, 9 January 2019 (UTC)


2021, Version 1.3.3117

A long time passed since this discussion. Are the current values still correct? I tried to check the "Quality generation data" in 1.3.3117 but I couldn't access it. In 1.2.3062 I got the following:

Quality generation data 1.2.3062.png

It's not only rounded to two significant figures, it also differs in a lot of places by more than rounding errors. I suppose this in game table doesn't represent the actual chance but rather the percentage of items generated by the game. Can someone verify this? What is the actual formula? Ickputzdirwech (talk) 15:35, 3 October 2021 (UTC)

I think the table that the game generated for you, and also the table on the Quality page right now, were found through simulating the creation of many items and counting the results. Since there's inevitably some variation to doing that, the results will be slightly off unless. I poked around in the source code to find the actual algorithm for determining quality, and tried simulating it myself: https://github.com/Cheldra/rimworld_stuff/blob/master/quality.py. I think it's either really hard, or else mathematically impossible, to find the probabilities analytically. The results with 10,000 simulations per level were about the same:
0 [64.83, 30.07, 4.91, 0.2, 0.0, 0.0]
1 [43.31, 43.67, 12.12, 0.89, 0.02, 0.0]
2 [20.25, 53.23, 23.47, 2.96, 0.09, 0.0]
3 [9.1, 50.71, 33.41, 6.5, 0.27, 0.0]
4 [4.78, 45.17, 39.52, 9.94, 0.58, 0.0]
5 [2.24, 34.93, 47.05, 14.58, 1.18, 0.01]
6 [0.93, 24.25, 52.16, 20.41, 2.23, 0.02]
7 [0.42, 15.45, 53.47, 26.68, 3.9, 0.07]
8 [0.13, 9.11, 50.67, 33.62, 6.32, 0.15]
9 [0.06, 5.64, 46.85, 38.18, 9.02, 0.26]
10 [0.02, 3.27, 40.52, 43.66, 12.05, 0.47]
11 [0.01, 1.8, 32.13, 49.23, 16.08, 0.75]
12 [0.0, 1.0, 24.74, 52.52, 20.54, 1.21]
13 [0.0, 0.62, 19.91, 53.72, 24.18, 1.57]
14 [0.0, 0.4, 15.68, 54.41, 27.31, 2.2]
15 [0.0, 0.21, 12.17, 53.62, 31.24, 2.77]
16 [0.0, 0.12, 9.24, 52.46, 34.43, 3.75]
17 [0.0, 0.08, 6.89, 50.43, 37.93, 4.66]
18 [0.0, 0.03, 5.04, 47.43, 41.3, 6.19]
19 [0.0, 0.02, 3.57, 42.57, 46.24, 7.6]
20 [0.0, 0.01, 2.44, 37.56, 50.4, 9.58]
Cheldra (talk) 23:40, 3 October 2021 (UTC)
Ok just as I suspected. Thanks for the confirmation and the calculations! I checked the above linked reddit post again and the author states in this comment, that the current values were also generated by a c++ program. There is no link to it or any mention of how many simulations they did, however. Your 10,000 simulations per level sound pretty solid. I'm tending to update the table with your values. Sure some variations are unavoidable but discrepancies of up to 0.3 (level 5, poor) seam quite high for that many simulations. Ickputzdirwech (talk) 10:30, 4 October 2021 (UTC)
In case you want it, here's also the means and std deviations observed after repeating those 10,000 simulations per skill level 10 times:
level 0 1 2 3 4 5
0 64.65+-0.1 30.16+-0.1 4.98+-0.04 0.21+-0.02 0.0+-0.0 0.0+-0.0
1 43.44+-0.14 43.56+-0.13 12.11+-0.1 0.87+-0.02 0.02+-0.0 0.0+-0.0
2 20.28+-0.14 53.18+-0.13 23.52+-0.08 2.93+-0.05 0.09+-0.01 0.0+-0.0
3 9.14+-0.11 50.76+-0.17 33.39+-0.17 6.41+-0.07 0.29+-0.01 0.0+-0.0
4 4.81+-0.06 45.18+-0.16 39.4+-0.17 10.0+-0.09 0.61+-0.02 0.01+-0.0
5 2.26+-0.04 34.73+-0.08 47.14+-0.11 14.66+-0.11 1.2+-0.04 0.01+-0.0
6 0.98+-0.03 24.24+-0.12 52.15+-0.13 20.39+-0.09 2.21+-0.04 0.03+-0.0
7 0.38+-0.02 15.51+-0.12 53.32+-0.11 26.84+-0.12 3.87+-0.05 0.07+-0.01
8 0.13+-0.01 9.01+-0.1 50.76+-0.1 33.52+-0.08 6.43+-0.07 0.15+-0.01
9 0.06+-0.01 5.59+-0.08 46.97+-0.11 38.17+-0.08 8.94+-0.08 0.26+-0.01
10 0.02+-0.0 3.33+-0.04 40.15+-0.1 43.84+-0.12 12.21+-0.04 0.45+-0.02
11 0.01+-0.0 1.86+-0.05 32.17+-0.17 49.06+-0.11 16.14+-0.12 0.76+-0.02
12 0.0+-0.0 1.0+-0.03 24.58+-0.1 52.66+-0.13 20.57+-0.11 1.19+-0.05
13 0.0+-0.0 0.63+-0.03 19.99+-0.11 53.88+-0.2 23.88+-0.13 1.62+-0.03
14 0.0+-0.0 0.4+-0.01 15.82+-0.09 54.27+-0.15 27.35+-0.14 2.15+-0.04
15 0.0+-0.0 0.24+-0.01 12.29+-0.14 53.84+-0.21 30.78+-0.15 2.85+-0.07
16 0.0+-0.0 0.14+-0.01 9.28+-0.09 52.38+-0.16 34.53+-0.17 3.67+-0.08
17 0.0+-0.0 0.08+-0.01 6.91+-0.06 50.3+-0.12 37.93+-0.14 4.77+-0.1
18 0.0+-0.0 0.04+-0.01 5.0+-0.04 47.67+-0.15 41.23+-0.17 6.06+-0.09
19 0.0+-0.0 0.03+-0.0 3.54+-0.06 42.73+-0.16 46.04+-0.15 7.67+-0.08
20 0.0+-0.0 0.01+-0.0 2.44+-0.06 37.4+-0.14 50.56+-0.1 9.58+-0.05
I don't think updating the numbers on the page is a super high priority, but go ahead if you want.Cheldra (talk) 11:50, 4 October 2021 (UTC)
Done. And although most of the changes were minor the work wasn't for nothing. When calculating the average results I noticed that they were significantly of in the inspired specialist table. In hindsight fairly obvious. They were above 6, while 6.00 would mean 100% legendary items. Ickputzdirwech (talk) 10:43, 5 October 2021 (UTC)