Minesweeper Super Challenge

Anything to do with minesweeper...
CineMassive150
Posts: 5
Joined: Wed Aug 12, 2015 2:28 pm

Re: Minesweeper Super Challenge

Post by CineMassive150 » Mon Aug 31, 2015 4:19 pm

Hi everyone,

We apologize for the slow response. As Curtis mentioned, we have been continuing to work out the details of the event, and we really appreciate all of your feedback and comments. We also had a few other points that we would like to get your feedback on.

First, in general, we wanted to collect your thoughts on whether the game will be simply too difficult to complete within 12 hours in its current form (262x718 board with 38,799 mines at expert level). If so, would it be better to increase the length of play-time or to reduce the difficulty of the game in some way? As a note, we have noted your concerns about the "3 attempts" rule, and will be discussing amending that rule based on your comments.

With that, as Curtis mentioned, we have also been evaluating whether the number of forced guesses in the current set-up is too high. If the number of guesses is as high as Curtis' estimate suggests (around 11), would it be preferable to use a version of the game that enables the guesses to be removed? If a no-guesses game is strongly preferred, we would certainly like to explore using freesweeper as a solution for this.

Another topic we'd like to discuss is creating a downloadable version of the game that has a scrolling feature and matches the specs of the game that will be played on the video wall. It sounds like freesweeper might be a good candidate for this if Cryslon could enable it to support a board of the dimensions we would need. We would like to get your suggestions on the best approach for this.

Thanks again to everyone for the input!

Zlix
Posts: 38
Joined: Fri Jun 13, 2014 12:29 pm

Re: Minesweeper Super Challenge

Post by Zlix » Mon Aug 31, 2015 5:17 pm

Hi.
First, I already said my thoughts about 12 hours limit, but I'll repeat. For me it would be too stretchy to have 12h total. On the other hand 12h playtime would be more than enough if you allow several breaks, especially if there are no forced guesses. I played minesweeper for many hours without a break, probably around 6-8h, but playing for best time does not require as much precision. You can make so many mistakes, all you need is one perfect board... Playing 12 hours straight with no mistakes is beyond my comprehension because I can't guess how will my wrist behave after few hours of "precision playing". My guess is that I would have play a bit slower and way more tense, multiply that by 12 hours... I would like to hear other sweepers opinions about this.

As of 3 attempts rule, it's not good as is. If you ask me I would allow unlimited number of attempts within first 2-3 hours. After that it's no longer solvable within current time frame even at fastest rate.

It would be nice if someone could calculate number of forced guesses needed to finish the board. I've been playing 58x118 board lately and I had from 2 to 10+ forced guesses by board. No-guesses rule would give us a fair chance to finish the board but It's not by the current sweeping rules so it leaves me with dilemma whether to play by the rules or whether to bend rules as much as necessary to get a fair chance.

Scrollable version would be great. That is how I imagined it in the first place.
I have one suggestion(again :)) It would be great if you could add a progression map. Hope it's proper term. By progression map I mean mini map with current progress(something like real time print-screen of entire board). If I haven't expressed myself properly, it's same as "progress" feature on Fill-a-Pix. Again, I'm not sure how much work it require but it could be useful if you skip a field or an area while playing.

EWQMinesweeper
Posts: 410
Joined: Sun Nov 30, 2008 11:50 pm

Re: Minesweeper Super Challenge

Post by EWQMinesweeper » Mon Aug 31, 2015 9:10 pm

CineMassive150 wrote:First, in general, we wanted to collect your thoughts on whether the game will be simply too difficult to complete within 12 hours in its current form (262x718 board with 38,799 mines at expert level). If so, would it be better to increase the length of play-time or to reduce the difficulty of the game in some way? As a note, we have noted your concerns about the "3 attempts" rule, and will be discussing amending that rule based on your comments.

With that, as Curtis mentioned, we have also been evaluating whether the number of forced guesses in the current set-up is too high. If the number of guesses is as high as Curtis' estimate suggests (around 11), would it be preferable to use a version of the game that enables the guesses to be removed? If a no-guesses game is strongly preferred, we would certainly like to explore using freesweeper as a solution for this.
Is it possible to solve such a board with a total time below 12 hours? Absolutely. Is it likely that someone will complete such a board with possible forced guesses within a reasonable amount of time? Not very. Increasng the length of play time wouldn't increase the chance to win such a board by much. The only way to make completing such a board realistic would be to remove forced guesses by using freesweeper. It would be a different kind of challenge than the current largest board, which had forced guesses but if it's important to you that the board gets solved, then go with freesweeper. As long as it is made clear that there are no forced guesses I'm fine with that. Personally I'd prefer to have forced guesses and I would try to solve as much of the board as possible before taking any guesses.
Zlix wrote:Hi.
First, I already said my thoughts about 12 hours limit, but I'll repeat. For me it would be too stretchy to have 12h total. On the other hand 12h playtime would be more than enough if you allow several breaks, especially if there are no forced guesses. I played minesweeper for many hours without a break, probably around 6-8h, but playing for best time does not require as much precision. You can make so many mistakes, all you need is one perfect board... Playing 12 hours straight with no mistakes is beyond my comprehension because I can't guess how will my wrist behave after few hours of "precision playing". My guess is that I would have play a bit slower and way more tense, multiply that by 12 hours... I would like to hear other sweepers opinions about this.
It's all about the right preparation. Play a bit slower than usual, solves patterns the easy way, take short breaks whenever neccessary (1-2min is all you need). My 1000 Int marathon had barely any misclicks at all whereas nearly all my misclicks on 118x60 are the result of trying to solve the board in less than 1000 seconds.

I disagree that playing for best tme requires less precision than solving large boards. Playing for best time is all about precision. You solve patterns in a certain way that is fast but not guaranteed to win. You keep playing until the board fits the solve. Kamil and I could have easily won 10-15 sub50s per hour but instead chose to take more risks to win a sub40 every other day and some sub45s per day.
Zlix wrote:It would be nice if someone could calculate number of forced guesses needed to finish the board. I've been playing 58x118 board lately and I had from 2 to 10+ forced guesses by board. No-guesses rule would give us a fair chance to finish the board but It's not by the current sweeping rules so it leaves me with dilemma whether to play by the rules or whether to bend rules as much as necessary to get a fair chance.
There can be two categories: Standard and no guesses. The size of the board is enough that a successful solve would be one of the most amazing things in the history of Minesweeper.
Zlix wrote:I have one suggestion(again :)) It would be great if you could add a progression map. Hope it's proper term. By progression map I mean mini map with current progress(something like real time print-screen of entire board). If I haven't expressed myself properly, it's same as "progress" feature on Fill-a-Pix. Again, I'm not sure how much work it require but it could be useful if you skip a field or an area while playing.
Great idea.
„Das perlt jetzt aber richtig über, ma sagn. Mach ma' noch'n Bier! Wie heißt das? Biddä! Bidddää! Biddddäää! Reiner Weltladen!“

Zlix
Posts: 38
Joined: Fri Jun 13, 2014 12:29 pm

Re: Minesweeper Super Challenge

Post by Zlix » Tue Sep 01, 2015 6:53 am

EWQMinesweeper wrote: I disagree that playing for best tme requires less precision than solving large boards. Playing for best time is all about precision. You solve patterns in a certain way that is fast but not guaranteed to win. You keep playing until the board fits the solve. Kamil and I could have easily won 10-15 sub50s per hour but instead chose to take more risks to win a sub40 every other day and some sub45s per day.
Well, there is not really anything to agree or disagree here. There is a reason why two of you are top5 players and I'm only top100. You made a list of 10 players. Speed and time references should be made by slowest two instead of fastest two...

As of guessing vs. no guessing. It's really up to the organizers to decide what is the main goal of this challenge. If the main goal is to play by the current rules and see how close can we get, then go for forced guesses, if the main goal is to finish the board than no guessing is way to go.
EWQMinesweeper wrote: The size of the board is enough that a successful solve would be one of the most amazing things in the history of Minesweeper.
Totally agree. More I think about this closer I am to no guessing rule. Maybe if we finish this challenge successfully they can organize another challenge with guessing. It would be level up and probably the ultimate challenge and only those who successfully finish first challenge could participate in next.

aradesh
Posts: 90
Joined: Sat Aug 29, 2009 3:37 pm

Re: Minesweeper Super Challenge

Post by aradesh » Wed Sep 02, 2015 1:32 pm

Been having a think about this and it's just sinking in how enormous the board is. I think for anyone to have a chance of beating it in under 24 hours, it would have to have a density less than expert. Probably less than beginner density.

No way I'd be any good here, even with a lower density. I remember on mmo-minesweeper I had to restrict myself to opening no more than 1000 squares in one go, as I'd always lose focus and make a dumb error after any longer. Even then I still messed up regularly.

Maybe if the game would allow errors, and allow multiple players at once, and just see how few errors it can be solved in?

MichaelD
Posts: 6
Joined: Fri Apr 05, 2013 7:41 pm

Re: Minesweeper Super Challenge

Post by MichaelD » Thu Sep 03, 2015 4:34 pm

I would love to have a go at this but there are probably 10 already signed up.

The 3-strikes rule is silly, there will be a lot more than 3 forced guesses on a board that size. Not to mention that we're all (I imagine) in the habit of making a guess when we don't strictly need to, when a square has a low chance of being a mine and a high chance of opening up space if not a mine.

I haven't followed the time calculations closely, but we can't expect the speed of solving a regular expert game to be mimicked with this. Even a poor game is still finished within a minute and a half, but that doesn't mean you can keep that up for 12 hours. A bad sprinter is still going faster than a marathon runner, is what I'm saying.

qqwref
Posts: 121
Joined: Thu Sep 23, 2010 4:17 pm

Re: Minesweeper Super Challenge

Post by qqwref » Sat Sep 12, 2015 12:34 am

CineMassive150 wrote:First, in general, we wanted to collect your thoughts on whether the game will be simply too difficult to complete within 12 hours in its current form (262x718 board with 38,799 mines at expert level). [...]

With that, as Curtis mentioned, we have also been evaluating whether the number of forced guesses in the current set-up is too high. If the number of guesses is as high as Curtis' estimate suggests (around 11), would it be preferable to use a version of the game that enables the guesses to be removed? If a no-guesses game is strongly preferred, we would certainly like to explore using freesweeper as a solution for this.
Curtis's estimate is actually very low. I have a program that counts the most common types of forced-guess patterns, and ran 1000 boards of the above dimensions. The average number of forced 50-50 guesses was 41.527.

With intermediate density (15.625% or 29394 mines), it is still about 9.658 guesses. That's a bit more plausible. With 25000 mines (13.3%) there are about 4.394 guesses, and with 20000 mines (10.6%) there are about 1.476. Of course, these lower densities may be too sparse to be fun - it's worth testing them out.
NF player. Best scores 1-10-39.

CBright
Posts: 39
Joined: Sun Nov 30, 2008 11:53 pm
Contact:

Re: Minesweeper Super Challenge

Post by CBright » Sat Sep 12, 2015 9:22 pm

qqwref wrote:Curtis's estimate is actually very low. I have a program that counts the most common types of forced-guess patterns, and ran 1000 boards of the above dimensions. The average number of forced 50-50 guesses was 41.527.
Awesome Michael, this is exactly the kind of information I was looking for. I should've mentioned that I expected my estimate to be low since it only used the board of the current WR, which I would expect to be 'nicer' than the typical board. Thanks a lot for running that simulation and posting your results! Did you do any analysis of where the guesses normally occur (i.e., how many occur along the sides)?

Having around forty independent 50-50 guesses makes the typical board of that size completely infeasible to solve, so I think this necessitates using a lucky mode (à la Freesweeper), unless the mine density is decreased as you pointed out.

qqwref
Posts: 121
Joined: Thu Sep 23, 2010 4:17 pm

Re: Minesweeper Super Challenge

Post by qqwref » Sun Sep 13, 2015 12:14 am

I didn't look at where the guesses are located, no. There are more guesses per square near walls and corners, because anything outside the board is essentially a mine, in the sense that you can't see the number. That only applies for 3 or so rows from the edge, though, so for a board like this I would assume that most guesses appear on the interior somewhere, just because there's so much interior.
NF player. Best scores 1-10-39.

CBright
Posts: 39
Joined: Sun Nov 30, 2008 11:53 pm
Contact:

Re: Minesweeper Super Challenge

Post by CBright » Sun Sep 13, 2015 9:34 am

qqwref wrote:That only applies for 3 or so rows from the edge, though, so for a board like this I would assume that most guesses appear on the interior somewhere, just because there's so much interior.
I'm not so sure about that - I don't have much intuition about how likely guesses are to occur in the interior, but it could conceivably be low enough such that there are rarely guesses in the interior. If that turns out to be the case, another way of making such large boards feasible to solve is to make the board wrap around on itself where the edges would be (like in the infamous "infinite solutions minesweper" mode) - topologically, playing Minesweeper on a torus. Any chance you would be willing to share your code somewhere?

qqwref
Posts: 121
Joined: Thu Sep 23, 2010 4:17 pm

Re: Minesweeper Super Challenge

Post by qqwref » Sun Sep 13, 2015 9:09 pm

Sure, here's the code I used. It's in JavaScript, which is not super fast, but it's good enough for a rough idea of how many guesses you can find.
http://pastebin.com/KCWXXLDC

If you want to change the size and number of boards, modify lines 33 and 34.
NF player. Best scores 1-10-39.

senfti
Posts: 19
Joined: Thu May 08, 2014 8:27 am

Re: Minesweeper Super Challenge

Post by senfti » Fri Sep 25, 2015 8:40 pm

I am thinking about implementing lucky/no-guess mode for MySweeper. But I'm not sure about how exactly this mode should work.

My thoughts about lucky/no-guess mode :
The easiest way would be to take only simple guess patterns into account, but this way you still might be forced to guess, especially on the second click.
Another approach would be to "help guess" everytime when guessing is needed, but you need no risk management this way and I'm not sure if an implementation working in real-time on huge boards is possible.
The probably best and also most difficult approach would be to move a mine only if a wrong guess was made on the square with the highest likelyhood of beeing no mine.

I tried 100x100 with expert density many times and I was never able to solve it. My best attempt ended with a wrong guess after 50 minutes on the second last 50-50 guess (3BV was 3512).

qqwref
Posts: 121
Joined: Thu Sep 23, 2010 4:17 pm

Re: Minesweeper Super Challenge

Post by qqwref » Fri Oct 02, 2015 2:51 pm

No guess patterns is pretty easy to do, and does help a lot. I implemented that in my javascript program (http://mzrg.com/js/mine/multi.html, choose the "no guess patterns" game), so it's not too hard and doesn't require that much computational power. There would still be several guesses, or areas where you would end up stuck, but I don't know how many - it does depend on how good your library of guess patterns is. In that program I only have the most common 2 patterns, and in the pastebin code above there are 3 patterns.

"move a mine only if a wrong guess was made on the square with the highest likelyhood of beeing no mine" might sound OK, but I think it would make for a rather unpleasant game. Not only would you have to solve every logically solvable square before starting to guess - it's all too easy to miss one, even on 16x30 - but if you were wrong about the probabilities you will be back where you started in terms of risk. In any kind of complicated situation, unless you want to whip out the spreadsheet program and spend hours staring at the board, you'd be better off just guessing.

One feasable option for a no-guess game is, if someone ever clicks on a mine, and there is some way to move the mines around so that all the visible clues still apply without killing the player, do that. So you could go through the board in any order, treating guesses like they aren't there, and as long as you didn't make a logical error (click a square that must be a mine) you're good. I'm not sure exactly how this is coded, but I think I've seen it done before.

On the other hand, you could just generate a board that requires no guesses at all, a la Tatham. It's pretty complicated code, since it essentially simulates playing the entire thing with humanlike logical deductions, but the program is open source. I just generated a 300x160 board with 10000 mines (in Google Chrome, on the javascript version) in a few seconds - not bad!
NF player. Best scores 1-10-39.

senfti
Posts: 19
Joined: Thu May 08, 2014 8:27 am

Re: Minesweeper Super Challenge

Post by senfti » Sun Oct 04, 2015 2:47 pm

MySweeper 1.2.0 supports lucky/no-guess mode now (http://www.minesweeper.info/forum/viewt ... f=26&t=273).

I didn't implement just a pattern approach because my solver approach also solves all standard guess patterns easily. This solver is able to find the probability of being a mine for squares in a "guess-area". This "guess-area" contains all unopened squares which are not separated from the guessed square by a border of at least 2 opened/correctly flagged squares (this way the guess-area can be treated independently from the rest of the board). For the guess-area all possible permutations of the mines are checked and with this information the likelihood of being a mine can be calculated for all squares. If the guessed square has the lowest likelihood of being a mine in the guess-area, then the mines in the guess-area are reordered to another randomly chosen permutation.

Problems of this algorithm:
- it has exponential runtime (at the moment exponential with the number of processed unopened squares, this can be reduced to the number of number of unopened squares at the border of the guess-area): to avoid lags there is a maximum of 50000 permutations (for example this is a guess-area of 18 squares with 9 mines within)
- to give the player no additional information it has to be assumed that the number of mines in the guess-area is know by the player
- to find the border of the guess area it's important that all known mines are flagged (else they are treated as unknown squares); so always flag all know mines before guessing, else even simple guess patterns may no be found; wrong flagged squares are a problem, so you won't have luck if a wrong flagged square is found
- the possibility of successive guesses is not taken into account when calculating the probabilities

If you want to read the source code of my lucky mode implementation, download the source code of MySweeper. The lucky mode parts are in MinePermutator.h/MinePermutator.cpp and in Board.cpp line 593-794.
Attachments
guess-area.png
marked squares are part of the guess-area
the black line is the border
guess-area.png (4.31 KiB) Viewed 6031 times

Post Reply