How does the latest future of the fortress reply affect the science done in this thread?
The reply:
Let's see, fort scholars... they do activity cycles, the length of which is 1-2 days (whether they are pondering or discussing etc.) Once they get through 50 cycles, it rolls 0-50 vs. the number of completed cycles minus 50 to see if they get "breakthrough credit." So at 51, they have a 2% chance, and at 100, they have a 100% chance. Then, it resets the cycle number to zero and gives them breakthrough credit, based on a skill roll plus 100 (for discuss, the other researchers contribute half of their summed skill rolls.) Based on the difficulty (1-4) of the topic, total lifetime breakthrough credit is then assigned a number of 50-sided dice. An easy topic is dice=credit/2500, then /5000, then /10000, then /20000 for level 4 topics. The number of dice cannot exceed 10. Then roll these dice -- if you get a 50 on any of them, discovery! This is a bit archaic, and I'm not suggesting it works particularly well. But that is how it works. Also: if they fail to get the breakthrough after the 50-sided rolls, they have a 2% chance of switching topics, or if their credit exceeds 100000, they always switch topics (though they keep the credit, so returning to the topic later gives them a decent chance at breakthrough.)
I went ahead and added this quote to the wiki page on
Topics. I wasn't sure whether to add it to topics or knowledge, but I figured better to get it on the wiki sooner rather than later and let someone else who knows better organize it. I think the next logical thing to do is to work out the implications of this reply in regards to the science in this thread.
Important question: does it look like the credits Toady mentioned correspond to the research points therahedwig was tracking?
The way knowledge works is that you have that knowledge tech-grassland, with it's topics. Each histfig keeps track of which topics they do and don't know, as well as which books they've read. It also keeps track of a knowledge goal for the histfig and how many times they've performed a research action(in the DFHack api reffered to as times pondered, but it also goes up when they participate in a discussion regarding their research goal). After a certain amount of research actions (DFhack comment suggests between 40-60, I myself suspect this is tied to academic skill) the scholar gains research points in said topic(the amount of research points is definitely tied to skill level, with dabbling giving in the thousands, and competent/skilled three times as much, but there's no set numbers). According to the DF hack comments, 100K research points is necessary for a topic to be researched, and the histfig will then write a book about it.
There's still a lot I don't understand, for example, there is a certain amount of variance in how much pondering is necessary to get research points, and the research points themselves also vary for a single scholar. One time a skilled (lvl 4) geographer got 4700 research points, and the subsequent time 3700 research points.
Also does anyone have past research or FotF statements on how Toady typically handles adding skills to rolls? I think this has a big impact on optimal strategy for developing scholars and researching topics. In particular, based on therahedwig's research on xp gain, it heavily influences the tradeoff on dwarf's discussing more vs. dwarf's pondering more.
therahedwig has noted that research takes a long time. But with the latest information from Toady, we can use a bit of statistics to get an exact probability distribution of how long for level 1/2/3/4 topics...
- Research takes a long, long, long time. I haven't ever seen the research progress on a legendary something-or-the-other, so no idea how much effect this has. However, as long as the scholar doesn't die, they keep their research progress, suggesting that you could expel a scholar if you think a siege will end your fort, and then start a new one elsewhere and hope the scholar comes to your new fort.
Some initial math, with unskilled rolls, as a proof of concept:
So for totally unskilled rolls each 1-2 cycle averages (0*1/51+.02*1/51+.04*1/51+.06*1/51... +1.00*1/51)*100=50 research credits per cycle.
This means an additional dice every 50 cycles for level 1 topics.
So after the first 50 cycles, (cycles 51-100) the odds are now 1/50 chance of breakthrough for each cycle.
However there is also the 2% (1/50 chance) of changing research topics.
Cycles 101-151 have a 1-(49/50)^2 odds for each cycle getting a breakthrough. And so on...
I can a graph showing the exact odds on each cycle and then make a cumulative probability distribution from that...
It does indeed look like it takes a long time for even level 1 research topics to get a breakthrough... and early on, there are even odds between making a breakthrough and changing to another research topic entirely.
Important Edit: I think I made a major error in my understanding of what Toady said... I misread the description of the roll each cycle that it was vs. the cycle counter and I also missed that it takes 50 cycles before the first roll. This increasing the research time required even more.
Edit 2: So with a corrected understanding of what Toady was saying...
With zero contribution from skill rolls it takes an average of 50+8.54 cycles to get a breakthrough credit. This means an additional breakthrough dice every 1464 cycles for level 1 topics, every 2528 cycles for level 2 topics, every 5056 cycles for level 3 topics, and every 10112 cycles for level 4 topics. From that point, it is an average of 2528 cycles to get a breakthrough, or to switch intentionally topics.
Author
Topic: Libraries and Books (Read 17822 times)