New board difficulty

Want to calculate something or share your results?
Locked
KamilSaper
Posts: 149
Joined: Mon Dec 01, 2008 7:16 pm

New board difficulty

Post by KamilSaper »

Ive noticed interesting fact (picture) connected to distance, taking it into account + 3BV would be better than normal 3BV, but I dont know yet how to connect it. Write your ideas. :D
Both boards on the picture have 3bv=5, but its obvious one of the boards is easier than the 2nd one.
Attachments
qwerty.JPG
qwerty.JPG (58.88 KiB) Viewed 5011 times
0.49 - 7.03 - 31.13
NF: 0.49 - 7.03 - 31.51
EWQMinesweeper
Posts: 419
Joined: Sun Nov 30, 2008 11:50 pm

Re: New board difficulty

Post by EWQMinesweeper »

as path/distance depends a lot on where you start i think it is too random to be really important for board difficulty
„Das perlt jetzt aber richtig über, ma sagn. Mach ma' noch'n Bier! Wie heißt das? Biddä! Bidddää! Biddddäää! Reiner Weltladen!“
User avatar
RonnyDeWinter
Posts: 76
Joined: Mon Dec 01, 2008 9:11 am
Location: Netherlands
Contact:

Re: New board difficulty

Post by RonnyDeWinter »

You'll still have easy boards with big distances and difficult slow boards with small distances. Simpy too many variables into this equation to come up with something really accurate and still being understandable.

Off course you could throw the history files of many players on a big pile and see what the avarage effect of 3BV, nr of openings, number of numbers, size of the openings and nr of islands (etc.etc.) is on your time and 3BVs, but I don't think anything as complex as that will not interest the majority of players.
NF 1 (0.96) + NF 15 (14.20) + NF 61 (60.18)

All my minesweeper records
KamilSaper
Posts: 149
Joined: Mon Dec 01, 2008 7:16 pm

Re: New board difficulty

Post by KamilSaper »

Anyway one of the boards I showed on a picture is possible to win in 1s, and the 2nd not really in normal mode :D Both boards have 1 opening and 5 3bv.
@[name removed]: What I dont like is that u r always the 1st person to say that my idea is bad :mrgreen:
0.49 - 7.03 - 31.13
NF: 0.49 - 7.03 - 31.51
EWQMinesweeper
Posts: 419
Joined: Sun Nov 30, 2008 11:50 pm

Re: New board difficulty

Post by EWQMinesweeper »

@kamil: so what? the 3BV of a board won't change when you solve it this or that way, but the path to solve it depends too much on in which order you solve the patterns.

for NF beg boards i might be easy to determine the most efficient path - try to find the same thing for an FL solve on an int board and you will see that it is too comlex
„Das perlt jetzt aber richtig über, ma sagn. Mach ma' noch'n Bier! Wie heißt das? Biddä! Bidddää! Biddddäää! Reiner Weltladen!“
User avatar
RonnyDeWinter
Posts: 76
Joined: Mon Dec 01, 2008 9:11 am
Location: Netherlands
Contact:

Re: New board difficulty

Post by RonnyDeWinter »

A long time ago I wrote a piece of software in Java to find the shortest path between between an X number of locations. It was using a kohonen network of 'neurons' to solve this 'travelling salesman problem.

A bit like this program:
http://www.vias.org/simulations/simusoft_travsalm.html

This way you could roughly calculate the shortest path around the board (without looking at the solving sequence). Problem is that it would take a LOT of calculation to do the same on a minesweeper board with 100-250 3BV. Still it would, in combination with the 3BV, probably be a better estimation for the boards difficulty. I'm not planning to program it though, cause it was almost 8 years ago since I wrote that and I can't program windows based application either. :mrgreen:

If some genius around here can combine a kohonen network with 'board solving neurons' to solve a board with the shortest path possible he would definitely earn my respect. :ugeek:
Last edited by RonnyDeWinter on Fri Jun 05, 2009 8:50 pm, edited 1 time in total.
NF 1 (0.96) + NF 15 (14.20) + NF 61 (60.18)

All my minesweeper records
User avatar
Tommy
Posts: 257
Joined: Mon Dec 01, 2008 9:22 pm
Location: Vienna

Re: New board difficulty

Post by Tommy »

A possible optimization would be to group 3bvs into clusters, so, for example, you could specify that adjacent non-opening 3bvs will be counted as one cluster. Not sure how exactly to connect them though, but I'll think about it.
Remember that the algorithm doesn't have to be ultrafast (it won't, and we have 3bv for sth that can be calculated at runtime). Maybe it's better to postprocess only boards that have some significance (highscore/similar).

I'll post nmore on that topic later on.
Don't anthropomorphize computers - they don't like it.
Locked