David Barry

 

About me:
I first started playing Minesweeper seriously when I was in Muscat, Oman, where I graduated from high school in May 2002.  It was there that I achieved my best times of
2-15-50.

I now live in
Brisbane, Australia, where I study maths and physics at the University of Queensland.  I finished 4th in the theory section of the inaugural Australian National Physics Competition, open to 3rd year students in Australia and New Zealand, in December 2003, and 3rd in the experimental section (we worked in pairs in the lab, so it was more like tied-5th).  OK, no more bragging. :)  Because of all this maths and physics, and also because until recently, my mouse was very bad, I haven't been sweeping much since starting uni in July 2002.




A long, rambling and vaguely connected essay:

"Looking through my file today, I found an intermediate screenshot with a time of 46 (that was my best time at the time). Upon further scrutinization, I found that, with the difference of one mine, it was the exact same as Marko Vucolic's 19-second board."
-Matt McGinley,
4 September, 2000.

This is the earliest post I know of in the guestbook relating to repeat boards.  Lanyje also found some repeated boards at around the same time; Gernot Stania's 15 board became known as the dreamboard.

Further repeated boards were found on beginner.  Dan Cerveny posted this prescient message to the guestbook on
24 February, 2001.

"I'm a bit worried that people will look at all of the repeat boards used for many of the top intermediate records, then start playing in such a way that they will get a great score if they happen to get one of those friendly boards (i.e. figure out the two or three places to click that will open up most of the board). Some of those top boards obviously show up quite often, so playing specifically to get a 10 second game on one of them seems like it could work. This worries me."

Just over four months later (27 June), Matt McGinley used his knowledge of the dreamboard to get a 10, in what is perhaps the most famous Minesweeper video in existence.  Here is Matt's post to the guestbook, announcing the new record.

"Words cannot describe how I feel right now. My hands are shaking and my chest is full of butterflies. It has taken me almost a minute to type the above lines. "Why so nervous, Matt?" you ask? A new frontier has been opened upon the minesweeper board. I'm in my 30,000th click of the day and only a 21, 22x2 and a 23 to show for it. Well not as of five minutes ago. Do you want me to get to the point already? Too bad...We need more suspense... I was fairly sure that my 14 would never be broken. I was fairly sure the world record was safe (yes, the world record). I was fairly wrong. Do you want to know what my time was? Do ya really? Are you shaking in anticipation? I am...Well, here ya go. The new intermediate world record: 10 - 10 - 10 - 10 - 10 - 10 I have the video. Mail me or visit my site soon."

A day later Matt explained how he had done this.  "I flew through that board entirely from my subconscious..."  He had watched his 14 video on the same board many, many times, working out how to click more efficiently.

As an aside, it was the day after Matt's 10 that Dan used the term "dream board" for perhaps the first time; this appears to be the earliest use of the term in the guestbook.

This created a big controversy over who the best Intermediate player was.  Curiously, the record of 10 seconds was not in dispute in the days soon after its setting.  The debate was only over whether Matt's 10 and 14 made him a better sweeper than Damien's masses of sub-17's.

A few days later Owen Fox had made this insight.  "...[I]t seems that on my PC, I get 'loops' where a lot of boards that I played yesterday might come up today..."

Matt then noted that the game before the dreamboard was the same in both his 10 and 14 videos.  Thus the notion of board cycles was born.

In August, I did my first, fairly inaccurate study of the cycles.  I wrote a program that read the positions of the mines on each Intermediate board, and found that there were two main board cycles, of about 15000 boards in length.  This number is inaccurate, because the program wasn't always good at clicking the smiley face to start a new game.

Fast forwarding to Christmas 2002, and I went back to my old program after Tim Kostka had given exact numbers for Beginner board cycles.  I fixed my program, and found that there were two board cycles on Intermediate, one of 11643 boards and the other (containing the dreamboard) of 11676 boards.

Later, Georgi Kermekchiev analysed the cycles further and found sub-cycles, with board shifts.  I would give more details, but I can't find his analysis anymore.  From what I recall, each cycle had four sub-cycles, each of these sub-cycles being a given displacement from the others.  Board shifts were first raised by Matt McGinley soon after my initial study of the cycles.  Matt also found two board which differed by four mines, an unsolved mystery.

Now, these cycles are the ONLY reason there was a large effort to build a good Minesweeper clone.  The timer bugs are minor.  Steffen Stachna and Roland Seibt both broke into the single-digits on Intermediate by cycling through the boards and waiting for the dreamboard (roughly speaking).

If I may now digress for a paragraph, I wish to state that I believe Steffen and Roland's records should stand.  This was my original position; over time I came to agree with the majority that the method used was invalid.  Looking back over the arguments at the time, though, I feel that if Matt's 10 is counted, so should the 9's.

So the Minesweeper community wanted a better version of Minesweeper, one which didn't have board cycles on Intermediate.  And, preferably, one which didn't allow a foreign program to read where the mines are before you start playing.  (Such a program now exists for the standard Winmine.)

I have to admit that I haven't played either of the two main clones.  However, the use of (at least) one of them will become a standard requirement for top Intermediate players.

It may also come to pass that the clone will be needed for Expert times as well, because of the possibility of pre-knowledge of mines on Expert boards (even though there are no cycles in Expert).  I believe that Martin Toft Madsen has used this to play a very genuine-looking 37, though I haven't seen the video.

I firmly believe that Stevan Gvozdenovic has tried to present a 41 video as a genuine time.  It is unfortunate because Stevan is one of the fastest sweepers around - his 44 is generally agreed to be genuine.  Previously, the community has been able to trust the top players; now we know that we can't always do this.

Of course, Stevan's 41 video was quickly shown to be faked (I will sheepishly admit to being fooled by it at first, though).  Stevan's technique was to create a solver that tries to mimic human mouse movement.  The problem he had was that the solver didn't do its mimicry very well - it was obvious that a computer was behind it.  I have ideas on how Stevan can improve his program.  Given that he still claims his 41 to be genuine, I believe that he will work out the same improvements he needs and will implement them, trying to claim more record-pace times.

I don't know if the solver will ever fool the whole community.  But it shows that the community will have to remain ever-vigilant, because the solver is just as capable of faking times on a clone as on the original Winmine.




The race to sub-50:

I preface this story by saying that the community needs to find a Japanese speaker.  There are at present no Japanese in Georgi's best ever list.  One would expect many more than this.

I joined Damien's expert records list as the second top player behind Lasse Nyholm, who had a best time of about 55 then.  In those early days of the Authoritative Minesweeper, my best time was 58 and Damien Moore's was 64.  Players not on the list included Chris Paradise (then world record holder, 51) and Vincent Yeh, who I think was at 52, and would join the community much later.

For many months, I closely followed Lasse.  He would break his best expert time, and within a few weeks I would break mine, normally ending up just one second behind him.  Lasse hit 52 on
18 November, 2000; I got down to 53 on 30 Nov.  At this time, Damien and Dan Cerveny were at 59 and Marc Schouten was at 65.  Then Lasse powered away, passing Chris Paradise's old record and getting a 50 on 9 January, 2001, followed five days later by this immortal guestbook posting:

"So, my dear minesweeper-brothers. I just took a 5 kilometre bikeride in minus 3 degrees celsius to find a computer with internet-access to tell you that the 50s-barrier has been defeated. It's been a strange minesweeper-day for me. I struggled with cold hands a could hardly get a sub-60, but after some hours and after getting a 51 I thought I would go to bed. Of course it was to tempting to keep playing and then a good board showed up. I started in the left side and went down to the corner, turned right and up, and when the lower part of the field was done I looked briefly at the clock that showed somewhere in the 20s. At this point my heart started dancing tapdance in my chest and I did my best not to get overexcited and miss a great score. So I took as few chances as possible and the white window showed with a new record. Here it is everybody - wait ....... Here it comes ..... fourty - freakin - seven 47 - 47 - 47 - 47 - 47 - 47 - 47 - 47 - 47"

Then followed the rapid rise of Sriram Sridharan, a good friend of mine in high school.  He and Damien were both at 56 in early February (Dan was at 57), but on 20 Feb, Sriram powered to a 51.  The next day, Damien got a 53, and the day after that, I got a 51 as well.  It was quite a few days of record-breaking for us Lanyje-chasers.

Marc had cracked the minute barrier in early January; he got down to 58 in early March.  On 13 March, Sriram became the second known sub-50'er with a 49.  About a week later, he made a 47.  This sparked a big debate on whether Sriram or Lasse was the best Minesweeper - Sriram had faster Beginner and Intermediate times, though Lasse had hundreds of sub-60'sSriram, somewhy, never submitted his time of 47 to Damien, though I strongly believe it to be genuine.  He just went into a lazy phase, I think.  Thus Sriram is left on the record lists at
2-11-49.

On 13 April, I was one click away from a 47.  Sadly for me, I have not come this close to breaking 50 since.  A small consolation for me was to own the record for the fastest expert video for a short while - my 50 video of 1 July was made before Lasse downloaded Camtasia.

By this time, Lasse had got his expert time down to his shoe size (or thereabouts), with a 44 on 20 June.  Three days later, Neil Collins scored a 52, and has since disappeared.  Dan and Damien both got to 52 by the end of July.

There followed a period of quiet near the top.  Marc got a 56 the day before the September 11 attacks, and Dan went from 52 to 51 in late October and then to 50 on 8 November.  Damien got a 51 ten days later.  Lukasz posted his times for the first time (I think) in the guestbook; he was at 55.  Vincent Yeh also made himself known to the community at around the same time, becoming the fastest known non-flagger with a 52.

In the leadup to Christmas, Lukasz gave himself a 53, and Dan gave himself a priceless 49 as presents.  Thus Dan Cerveny became the third known sub-50'er.  It looked like Dan would start to seriously challenge Lasse's position, for he scored a 46 on the eleventh day of Christmas.

This sojourn through history has now taken us into 2002, where I would like to stop, for the moment.  However, special mention must be made of Sorin Manea, who had a meteoric rise (or plummet?) to the 40's.  On
8 January, 2002, Sorin's best time was a 65.  His 49 came just six months later, on 12 July.  Damien had broken 50 in early March, and Lukasz did likewise not long after Sorin.  Poor Marc had to wait a while, but eventually both he and Reid Sinclair broke into the 40's.  Finally, I will note with patriotic pride that Dion Tiu became the first Australian known to sub-50.