3BV difficulty measure accuracy

Want to calculate something or share your results?
Locked
Masterchifa
Posts: 4
Joined: Sun May 03, 2009 2:17 am

3BV difficulty measure accuracy

Post by Masterchifa »

I've been playing minesweeper for a few years now. About a month ago I came across this site and started using the clone to track my game progress. If I understand 3bv correctly as the # of left clicks required to clear the board. Then, as 3bv increases, time to clear the board should also increase in a somewhat linear fashion as long as the player continues to play at about the same speed. Or in other words, as difficulty increases, time increases.

However, I've played several hundred games so far and looking at the time v 3bv graph It's just a cloud of dots. There is no correlation between time and 3bv. I'd expect to see something like the RQP graph (except rqp is fairly linear against time because it's a little like plotting time v time). I play NF, usually win in about 140s, and my game speed has not changed that much over the past month. If there were some correlation between 3bv and time, it'd make sense and I wouldn't be here. But there is none. It looks like a cloud of random points. I can see how wasted clicks could affect the stats, but I don't think the effect would be that profound.

Can anyone shed any light as to why this is? Because if its true, then it calls into question the reliability of the 3bv and 3bv/s metric.

thanks,
Matt
User avatar
Harryck_Repse
Posts: 24
Joined: Tue Apr 14, 2009 5:49 am

Re: 3BV difficulty measure accuracy

Post by Harryck_Repse »

I think its probably because many low 3bv boards can have more dense, less spread out sections of mines, so it takes more pattern recognition or guessing to see where the clicks should be because there might be some pattern where you have to guess or lots of 4/5/6s near each other placed so you will take longer for solving those squares. In higher 3bv boards many of the mines are usually more spread out so there are lots of easy solving patterns like a section of 1s that are easy to see and click on because of obvious nearby mine.

Sorry if this didn't make any sense I have trouble explaining myself :P.
Masterchifa
Posts: 4
Joined: Sun May 03, 2009 2:17 am

Re: 3BV difficulty measure accuracy

Post by Masterchifa »

Good point, I hadn't thought about that. But I don't think that solves the issue, rather I think it's a good reason why it exists. If what your saying is right, then there is a range in difficulty between boards of the same 3bv value. If there is a range of difficulty under each 3bv value, then 3bv has the same problem that just measuring by time does: it still leaves uncertainty in the actual difficulty of the board.

My point is, that the proof for a difficulty metric lies in how well that metric correlates to time taken. I'm not saying it has to be linear, but it should trend in some fashion.

I think 3bv is a good step in the right direction, but I don't think that it is much better than time alone. Perhaps the frequency each #[1-8] appears should be added to the calculation.

Now, perhaps it is a good metric, but only for those who are really good and waste fewer clicks (and is not a good metric for "inefficient" players like me). If someone does have a 3bv vs time graph that is linear, I'd like to see it!

-Matt
EWQMinesweeper
Posts: 419
Joined: Sun Nov 30, 2008 11:50 pm

Re: 3BV difficulty measure accuracy

Post by EWQMinesweeper »

3BV is NOT an indicator for the difficulty of a board! the 'sad truth' is that we have not yet found a way to figure out the actual objective difficulty of a board. so far all we have are some single aspects that make boards easier or more difficult (such like high/low 3BV, # of openings etc)...but but how much do they affect difficulty and compared to what?

IMO difficulty of a board is still something very subjective that also depends on each sweepers individual preferrings in solving the patterns. eg. a board that might be unbelievable easy for one sweeper may have some patterns that another (faster) sweeper will maybe not even after 10 attempts in upk solve in a similar time.

when i try to rank some of my own games as difficult or easy i consider 3BV, 3BV/s, time, cl/s, # of mines i used, openings AND my subjective emotins towards whether the patterns are easy or not i had while solving that board
„Das perlt jetzt aber richtig über, ma sagn. Mach ma' noch'n Bier! Wie heißt das? Biddä! Bidddää! Biddddäää! Reiner Weltladen!“
KamilSaper
Posts: 149
Joined: Mon Dec 01, 2008 7:16 pm

Re: 3BV difficulty measure accuracy

Post by KamilSaper »

We will never know index of difficulty of a board, because as [name removed]said, one board can be easy for one player and difficult for other. Examples: I have 36s on exp , but I couldnt get sub40 on any of DIon's sub40 boards in upk, as well as I have 10x34 on int but I couldnt get 10s on any Manu's 10x6 boards.
Definitely better view on board's difficulty u will have when u look also on openings, every opening takes usually 2-3 clicks.
IMO better board difficulty formula would be 3BV + 2x number of openings.
0.49 - 7.03 - 31.13
NF: 0.49 - 7.03 - 31.51
User avatar
Tommy
Posts: 257
Joined: Mon Dec 01, 2008 9:22 pm
Location: Vienna

Re: 3BV difficulty measure accuracy

Post by Tommy »

I guess that one could see stuff like 3bv as a metric for the potential of the board, in other words, the difficulty of the board when playing with full information, playing NF.

Basically the fact that the distribution is so random here illustrates the fact that minesweeper is by no means a game with perfect information, and also that, well, solving patterns is a complicated process that takes time, especially in the beginning.

I'd be interested in results from better players in general, because I'd guess that the better a player is, the less random the distribution gets (no offence!).
Also, I guess that if one compared flaggers and NFers, one could get an idea of how distorted 3bv (as a stat) gets by flags.

And I'd make a graph of my games now if I could easily and fast, but thats not the case (I use vsweep and am on linux at the moment, so all I could use is the csv file, and I should actually study :P)
Don't anthropomorphize computers - they don't like it.
Masterchifa
Posts: 4
Joined: Sun May 03, 2009 2:17 am

Re: 3BV difficulty measure accuracy

Post by Masterchifa »

I see, thanks for the input. I suppose I assumed that 3bv was the metric for difficulty because it seems to be the preferred measurement instead of time. It seemed that the 3bv/s value was more important than lowest time because it accounts for board difficulty (at least for NF players). I was just pointing out that it isn't as reliable compared to time.

thanks,
Matt

btw, I'd still be interested in seeing 3bv vs time plots if anyone has one that isn't a cloud.
EWQMinesweeper
Posts: 419
Joined: Sun Nov 30, 2008 11:50 pm

Re: 3BV difficulty measure accuracy

Post by EWQMinesweeper »

i play by far more arbiter than clone (~17000 games vs probably less than 1500) but this pic leaves room for interpretation :lol:
Attachments
3bvvstime.png
3bvvstime.png (42.12 KiB) Viewed 12295 times
„Das perlt jetzt aber richtig über, ma sagn. Mach ma' noch'n Bier! Wie heißt das? Biddä! Bidddää! Biddddäää! Reiner Weltladen!“
Masterchifa
Posts: 4
Joined: Sun May 03, 2009 2:17 am

Re: 3BV difficulty measure accuracy

Post by Masterchifa »

Interesting. Do you play NF? It looks like one can almost plot the difficulty range encapsulated by each 3bv value.

Also, Does anyone know if there is a way to import the history from clone 2007 into excel? (I checked the file, but it's binary and not ascii) I'd like to experiment with different equations to see if I can find a better difficulty metric.

-Matt
Attachments
minesweeper.JPG
minesweeper.JPG (27.02 KiB) Viewed 12299 times
KamilSaper
Posts: 149
Joined: Mon Dec 01, 2008 7:16 pm

Re: 3BV difficulty measure accuracy

Post by KamilSaper »

And my cloud ;)
Attachments
time x 3bv.JPG
time x 3bv.JPG (97.88 KiB) Viewed 12292 times
0.49 - 7.03 - 31.13
NF: 0.49 - 7.03 - 31.51
User avatar
Tommy
Posts: 257
Joined: Mon Dec 01, 2008 9:22 pm
Location: Vienna

Re: 3BV difficulty measure accuracy

Post by Tommy »

I'm not sure about the rest, but I know for certain that vsweep makes csv files.
Don't anthropomorphize computers - they don't like it.
EWQMinesweeper
Posts: 419
Joined: Sun Nov 30, 2008 11:50 pm

Re: 3BV difficulty measure accuracy

Post by EWQMinesweeper »

@matt: i'm a convinced flagger :lol:

@tommy: arbiter->game->save stats CSV

i have 17026 games in my arbiter stats...but i have no clue how to turn the data from the stats into a diagramm :oops:
„Das perlt jetzt aber richtig über, ma sagn. Mach ma' noch'n Bier! Wie heißt das? Biddä! Bidddää! Biddddäää! Reiner Weltladen!“
LTsully
Posts: 22
Joined: Mon Dec 01, 2008 2:56 am

Re: 3BV difficulty measure accuracy

Post by LTsully »

Openings also make a huge difernce in the dificulty of a board.
User avatar
RonnyDeWinter
Posts: 76
Joined: Mon Dec 01, 2008 9:11 am
Location: Netherlands
Contact:

Re: 3BV difficulty measure accuracy

Post by RonnyDeWinter »

If you look at your cloud of dots, make sure you only compare the dots of about the same color. If you do that you will most certainly see that ON AVERAGE lower 3BV boards result in faster times, just like less openings result on average in faster speeds/times.
NF 1 (0.96) + NF 15 (14.20) + NF 61 (60.18)

All my minesweeper records
EWQMinesweeper
Posts: 419
Joined: Sun Nov 30, 2008 11:50 pm

Re: 3BV difficulty measure accuracy

Post by EWQMinesweeper »

RonnyDeWinter wrote:If you look at your cloud of dots, make sure you only compare the dots of about the same color. If you do that you will most certainly see that ON AVERAGE lower 3BV boards result in faster times, just like less openings result on average in faster speeds/times.
the brighter the dots the more recently the games were played?????

btw, kamil's and my pics show that the relation between 3BV and time does not depend on the different levels
„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: 3BV difficulty measure accuracy

Post by RonnyDeWinter »

That's correct. That's why you'll always see more bright dots on the left (if not you can better quit playing, cause then you aint improving. ;) )
NF 1 (0.96) + NF 15 (14.20) + NF 61 (60.18)

All my minesweeper records
Locked