Blashyrkh here, fantastic solve Simon! The computer gibberish in the screenshot is actually _after_ giving it the center 5. Without that, the computer solver is totally lost. After all that talk about computers, I will have you know the puzzle is in actual fact (half) computer-generated! The break-in idea with the pointing high/low digits leaves some wiggle room for the precise layout of the thermos. So, I wrote a Python script that randomly places thermos (while keeping the break-in), and then has Rangsk fantastic solver check for uniqueness. This generated unique grids at a rate of about one per minute. After a while I got back to the computer to find a few dozen possible layouts. That left the task of selecting the one with the most appealing solve path. Most of the grids fell into two categories: a) Singles only after the break-in, which felt a little bland. b) Hardcore sudoku after the break-in, with multiple advanced techniques, which would have scared off any human solver. There were a select few however with solve paths like this one: Standard sudoku after the break-in, and then some sort of spicy stuff at the very end. This is exactly what I was looking for, and overall I liked this one best. By now I know three different ways to tackle the end: 1) The computer-y way: Finding a finned x-wing (whatever). 2) My way and that of many commenters: Colouring 12. 3) Simon's way: Finding multiple x/y-wings until the thing collapses :D
I'm loving the double meaning of center clash, with it applying to both the inital break-in to discover the 5, and to the impact 5 has as the central digit of the thermometers. Couldn't imagine a better name for this puzzle.
@@somebodystealsmyname no. It’s constraint based. The constraints remove choices. But towards the end, you could say that there is an element of suck and see. But from the output of the solver you can see if it is branching and bounding. It’s quite rare. The interesting part to me is that the solver is generic. You just have to add a new constraint type not change the solving bit. Fog of war will be hard to add but possible. The other interesting one is creation. Can you get it to work the other way
yea sometimes I am too busy to pay attention to the puzzle but put it on simply because Simon having so much fun doing them and just seemling like a nice blocke is just brightening the mood
Simon needed to read a lot of books, to get really good at speaking English like a poet, and then he needed to work at an investment bank, where he was dying of boredom, before he stumbled into live-solving variant SUDOKU every day during the so-called "pandemic" lock-downs. This channel is a confluence of several miracles, but now that it's here, life just wouldn't be as much fun without it...! Long live CTC...!
I'm very pleased with myself that I was able to solve this. A testament to CtC, which has taught me so much about how to approach puzzles and what questions to ask.
At least 75% of the time, when I get stuck and then I look at the video to see what I should have done, it was exactly what I already started to do but I didn't follow through with it. I feel like that's what this channel has taught me. That I pretty much know it. I just have to follow through with the thought. Think completely
Around the time I started watching this channel, there was a spree of puzzles that were similarly introduced as being resistant to computer solving. This intro made me feel nostalgic.
I was just thinking the other day that I'm not really interested in classic sudokus anymore unless it's watching Simon solving computationally difficult ones. Really nice seeing this puzzle!
lol, I was *just* talking to my wife about how insane Sudokus were getting back in the day by requiring jellyfish, swordfish, bent triples, etc., and how our broadened palette of variant rulesets was making it "easier" to find purchase...and then today! BENTIPPLES!
I was thinking similar to myself only yesterday! Mark mentioned a Jellyfish in his last video and it's been so long I can't remember what one is anymore. Back in the days that Mark would get stuck on a puzzle and just bifurcate and beat it into submission 😂
@@leftysheppey If I recall correctly, a jellyfish is two steps further down the rabbit hole than an x-wing, which is to say, four intersections of rows and columns that lock a digit into a subset of those intersections. That said, even if my explanation is correct, I would almost certainly never actually notice one on my own in a puzzle. lol
Hah. Conversations with my wife about sudoku are usually limited to "You spent two hours on a sudoku? Geez. Also, that's not how you pronounce Mephistopheles."
I used to avoid variant sudoku because I thought it made them too hard. When I started watching the channel four years ago I quickly discovered that complexity doesn't always equal difficulty, and that discovering beautiful interactions between the rules is way more satisfying than finding and applying complicated techniques that other people discovered for me. The reason computers get "confused" by puzzles like this is that they're really good at applying the techniques given to them, but really bad at creating their own techniques. The "trick" to this puzzle in inventing a new technique that is only applicable to this puzzle, which is why a human can solve it without complicated guesswork.
Once again, Simon proves his genius by spotting logic that I never could in a million years (nor can computers, apparently). At the same time, Simon aggravates me to no end by absolutely refusing to colour 1s and 2s, starting from 38:30 or so.
Likewise, I did it by coloring 1s and 2 without those final strategies. I prefer those strategies, but I couldn't because I missed placing a 9. (I still don't know whether the X-wing in 1s was superfluous.)
Problem with computers is, now that Simon spotted it, it can probably be added to a computer solving program, so that theoretically humans are “better than computers” just once for any kind of problem…
omg... this is the first time I finally understand what an X Wing is! That point about you have to look at the columns now was a huge lightbulb moment! I got so stuck at the end too lol
Said it before but it's worth repeating. I'd heard so many explanations on X-Wings that just failed to land with me. By contrast, the first time I heard the way Simon explained them not only could I spot them, I could use them. Same applies to his explanation of what he calls bent triples. He hasn't needed to break that out for a while but it's some thing that was an important addition to my toolkit.
I've had a go at puzzles where there was no obvious coloring solution and got stuck unable to spot the bent triple. Coloring worked today but I'm glad I watched the video. This time Simon's bent triple explanation clicked. I'm excited to try to use it on my own. Hopefully not too long before another bent triple puzzle is featured.
Normally on these "computer can't solve it" puzzles, I attempt to find something for a few minutes before I acknowledge that I'm not going to spot the hybrid "mutant--finned-jellyfish x skyscraper x 2-string kite" technique that Simon's going to pull out of nowhere and just enjoy the video. But on this puzzle I could feel there was some tension between the thermos, but couldn't quite get the foothold i needed. Started watching the video and as soon as Simon said "lots of low digits in the rows and high digits in the columns" the SET alarm bells went off in my head and my mind snapped into focus on the exact SET needed. Paused the video before Simon said "SET" and was off to the races. Amazing puzzle and a testament to what I've learned from countless hours following the channel.
Simon gave up on colouring too easily - much unlike his usual self - after going at eight-nines (@40:30). It really is like some of the old-school puzzles - colour the one-twos and note that r3c9 and r7c6 are different and looking at a 128 pencil mark in r7c9 which finishes the puzzle with, for the lack of a better desciption, alacrity. What a gorgeous puzzle!
That break-in was absolutely brilliant. After that I was able to solve it quite fast. I know I shouldn't brag, but I'm actually quite proud of that :).
I think I've become better at recognizing situations where the question of how many digits can certain cells share is very helpful. Finished in 21 minutes this time.
Same as my time, for same reason! I'm normally much slower and worse at all these, but managed that same insight on this one, that 125 must be the middle row, and that consequently the higher bulbs must be different, which quickly lets you place the 3/4 bulbs and fill everything in.
@@Cinakkuma its a little joke about strategies of x wings, y wings ang xy wings and a lot of other nonsense which are used to discuss how to solve really hard sudoku before Simon help us learn that some puzzles humans can solve that computers can't
19:30 Haven't you just proven that, as far as 5's go, there are four thermometers (all of them) with a 5 on it in a very restrictive place. Either the 5 is second to last on a thermo, meaning the bulb is a 1. Or the 5 is second from the start so the end of the thermo is a 9. Since there are 5's in restrictive places on all 4 thermos, the thermos are diagonally identical to each other except for the bulb and end digits. There will be two thermos going 1, 2, 3, 4, 5, X. And there will be two thermos going X, 5, 6, 7, 8, 9. And now it's just a matter of finding why one diagonal works and the other doesn't. Then you get a whole bunch of digits to play with.
I was so proud of myself for getting the break in, and following the logic around the paths.. then I got completely stuck. a bent triple.. how marvellous!
I finished in 54 minutes. That was such a cool break-in and overall a very enjoyable puzzle. I greatly enjoyed working out the limitations of the center square and was so happy to have gotten that 5 in the center. It may sound egotistical, but I am real proud that I was able to solve this puzzle, while a computer couldn't. It makes me feel like I am Kyle Reese fighting back against the machines. I very much enjoyed this one. Great Puzzle!
The criss-cross pattern with the high and low digits converging so that rows and columns needed to put 3 of their polarised-digits in the central box, such that there was an overlap, and thus '5' NEEDED to be in the middle of the grid, that was a fun conclusion to discover...!
No amount of Goodlliffing could help me break into this one. I stared at it for 15 minutes, and watched Simon make short order of it. He's a f'ing wizard.
Very nice puzzle. 24:33. Goodliffing; finding the low/high constraints on the central row and column; placing the 5; the rest of the high/low analysis in the cross boxes; reducing the thermometers; noticing the opposing 13s and checking a hypothesis. After that it's mostly cleanup.
27:30 finish. The way I figured out which of the thermos started with 1-2 was to see that if they were in the bottom left and top right, then the top left and bottom right thermos would both start with 3-5, which would place 1-2-3-5 in the center row of box 5. This means that it has to be the other way, and the bottom left and top right bulbs cannot be the same digit. So they are a 3-4 pair. An excellent puzzle, so much fun!
Same way for as I did it by the sounds of it. Took two minutes longer to complete than you though but then I messed about with breaking the 1,2,8,9's for a good 15 minutes...
1:30:20 - I can’t tell you how chuffed I am to have done that unaided. Massive thanks to Mark & Simon for giving me some of the skills needed to be able to solve these amazing puzzles.
Wicked cool puzzle, and awesome solve. I stared at it for quite a while, never came anywhere near the break-in, way above my pay-grade. With the break-in supplied, it mostly unfolds really nicely until the end-game bits. Where it gets sticky around 38:30, I never see things like the x wing 1's / 2's / the bent triple gambit. But I do love that Simon went there, because he's a f'ing wizard, and I love that he noted that there was probably an easier way to handle the end game. I colored 1's & 2's which placed 8 in c9r7, and it collapsed...
45:11 This was an insane ride. Started with a total Goodliffe of the thermos and gradually whittled them down with look-aheads at first before patterns started emerging. So clever how each part restricted and constructed the others.
It's not that the computer doesn't 'understand' set theory (unless it wasn't programmed to do so), it is that it looks for other ways first that are more common (as in a random puzzle is more likely to hold them) and uses those to solve the puzzle instead. It should be a very simple change in the program to make it show set theory as far as my understanding of how this solvers work. The solvers essentially go and mark each blank square with what it can be and eliminate possibilities, they do so by using a loop. The loop contains all known techniques and when it eliminates something it starts back from the beginning. So if the solver uses an X-wing to eliminate a possibility, it will go back to the start of the loop and look for naked singles again using less and less common logic. The computer commonly spots what I call an X in X, imagine 2 can only go in box 1 in the top row, same for 3 and 4, that is 3 numbers in 3 squares, so nothing else can go in there and everything that seems those 3 cells can't have any of those numbers. That is an XYZ wing, since we ran out of letters after Z, for 4 in 4 we call it WXYZ, VWXYZ wing for 5, etc. Importantly is the last technique the computer uses, backtracking, it essentially guesses a certain number in a certain box and sees if the puzzle solves or breaks, if it breaks we can eliminate that as a possibility. After guessing once we might need to guess again, this isn't really a problem, it is just if e.g. r1c1 is a 1 r2c2 can't be a 2, it can go fairly deep. The solver is using X in X first because generally X in X appears before any set theory, so whoever made it made it look for X in X before set theory, which leads to the solver finding solutions faster. Now it doesn't take long for the solver to find the solution to a puzzle, even just with pure guessing, but you wanna be efficient because you can make a program that uses this algorithm to look for puzzles with certain conditions that have a unique solution and there is gonna be quite a lot of them that fit them so you wanna look as fast as possible.
Very enjoyable solve. After looking at it for a second, I already knew what the "problem" is. Then it turned out not to be big enough problem. After thinking for a while, I realized I need to "brute force" it - just to see how exactly the limitations work. I filled the thermos with correct solution first try - well, something that looked possible, that is :) Only then I saw "the cell" - of course, every sudoku has 5 in the middle (not spoiler, that is common saying) With all thinking done ahead, just filling single cell quickly finished the thermos. Then it looked little annoying - the rest is regular sudoku - but actually still lot of very nice little logic discoveries were needed to finish the solve. Thank you for showing it, and Blashyrkh for making it!
I'm glad to see a video where Simon does teach us X-wing logic! I got stupidly stuck on how to logically solve the end of this puzzle, and resorted to "Well there's a lot of pressure on R8C8 (the 1-9 in box 9)" and realized 1 didn't work and 9 solved the puzzle. It worked, but still. Out of all the set theory tricks and mathematic combinations, we haven't seen much of what can solve a non-variance sudoku problem like this one. Edit: Also, to add what Simon says at the end of the video about "I hope I didn't overcomplicate it," I actually did throw that ending into a sudoku solver, and indeed, what Simon did is the correct method to find the solution. It forces you to find the two X-wings, then solve the XY-wing (or "bent wing" as he called it) to eliminate 8 as a candidate from R7C7. It wasn't more simple than what he did.
I see him start unwinding things at 30-odd minutes and then notice that there's still about 15 minutes more to the video... I was excited to figure out what happened. 😀
A phenomenal 23:37 solve on this. I just grok'd the crucial concept early by luck and kept just finding the right areas to focus on. Pretty chuffed about this one.
Simon: making super complicated logic Also Simon: saying that basic solving techniques (e.g. x-wing) is more complicated than the rest of the solution 😂
Computer are the most idiotic form of existance ever. If they seem smart, it's just because some nerd wasted too much time studying math, to apply it in the craziest way possible to make something work with only 1 and 0
Finished in 20:17 and I'm really proud of that. Wish I knew the double-click trick on the thermometers though, I could have saved a couple more seconds! I generally use tactics more like Simon's rather than Mark's but I almost always pencil mark thermos. So much logic seems to jump out from that
I used set theory to break in - made a set of rows 4&7 and a set from columns 4&7, and you can see that there must be a 5 in two of the thermo high ends, and a 5 in two of the thermo low ends, which got me to where Simon was ~20 minutes in. I found that easier than what Simon ended up doing This was a really great sudoku - not quite my favourite among recent puzzles, because once the thermos were done it became more about spotting things and less about deduction, but it was a great example of a puzzle where I really had to use types of reasoning I learned on this channel.
I think you must mean rows 4&6, cols. 4&6. Just sketched it out quickly and there are only 4 degrees of freedom left over when you take out the thermo cells....that is, the six leftover cells in the rows must exceed the six leftover cells in the columns by a minimum of 32. Since the maximum difference is 2x(24-6), that doesn't leave much room....
31:18 for me. I could just see that 5 in the middle too and then the thermos seemed obvious because the other way around would prevent 3 in the middle box
I love how your last few puzzles have been pure suduko puzzles with a simple rule set, yet challenging. Much better than your previous videos where there is are about 15 lines of text for the rules
27:10 How to find out that the particular choice of 2 vs. 5 doesn't work in a different way? First I clipped down a few more candidates: the 79 pair clipping block 5 column 4, followed by the 7s in block 5 column 6 clipping the rest of column 6. Then I *tried* to make R6C1's 134 resemble more the 14 and 13s in the other bulbs. The puzzle responded one better: 1 in the 134 cell broke the puzzle. 32:30 After filling the thermos, I centermarked the grid, finding some pairs and triples and placing digits in the process. I found a (probably superfluous) X-wing in 1s. I colored 12 pairs, changing to 1-not1 pairs, and that collapsed the rest of the puzzle. 38:40 Much more complicated? Come on, now. There's an X-wing in 1s. 43:20 Earlier on, I missed placing 9 in R3C7. Consequently, I never got the X-wing in 2s, and never got the Y-wing (bent triple) which would have collapsed the puzzle. Fortunately, I tried an alternate path.
27:22 I saw this due to high/low colouring. Where red means that a cell is the higher options notated in the cell and blue means that it's the lower options. Curiously, looking back, I can't recall you ever using this technique.
28:02 for me. Nifty puzzle! The X-wings and bent triple at the end weren't necessary. Just colour the 1s and 2s, and you'll find that r7c9 sees both, so can't be either, and the puzzle falls down.
I had an easier time with the break-in than Simon because I was looking at rows 4 and 6 and reasoning that the two subsets could only overlap in 3 digits, which readily brought ne to the same result. I didn't hold a candle to how smoothly Simon finished the puzzle after the thermos were done.
does anyone else think it’d be hilarious if Simon and Mark streamed the game Thank Goodness You’re Here? I’d love to see the opinions of my two favorite Brits on it 😂
I've just had a look at the preview to this and I have to say I would LOVE to play this on a stream. Can anybody who has played it advise us on whether it would be suitable for our audience?
@@CrackingTheCryptic there’s certainly a lot of innuendos and jokes that would go over younger viewers’ heads, but nothing explicit and no curse words.
Fun puzzle. I found it pretty approachable and finished in 42 min broken into a lunch session and finished after work. Update: I used set to achieve the same break in as Simon by comparing rows 4 and 6 to columns 4 and 6. This forced the 5's onto the wings of the thermometers, the initial box 5 constraints and the extreme thermo runs based on the placement of the 5s.
27:37 for me - and I didn't get stuck when I had that mess of 1289 cells at the end. :) To get the 5 in the middle, I wasn't looking at the numbers on the thermos, I was looking at the numbers that go in the empty cells. For example, in boxes 2 and 8, four of those empty cells were from 1234 and one cell from 56789, and in column 5, there are four cells from 1234; when taken together, each of box 2 and 8 give two cells from 1234 and the two cells from 56789, which leaves the three cells in box 5 also coming from 56789. Then when you do boxes 4-6 and row 5, you get the cells R5C4-5-6 being from 12345, which forces R5C5 to be 5.
At about 42:00 I realised that, if R1C3 is an 8, then R4C1 is a 9, and vice versa - meaning you can put a 1 in R1C1 - and the rest falls in place from there.
nice one, was quick to spot the break in, even the logic where the 1s have to go on the thermos is something i quickly got. Didn't see the bent triple in the end and had to watch Simon show me. So don't feel bad Simon, it seems that level of complication in the end was necessary ;) Thanks for the puzzle and the solve!
I ended up staring at that final state of the puzzle for a good 15 minutes having no idea what to do, then decided to color the squares, upon which I immediately noticed that r2c2 in box 7 could only go in r9c1, therefore in box 4 it could only go in r6c3. Since r2c2 could only be a 1 or 8, and r6c3 could only be an 8 or 9, it must be an 8, and everything unwound from there.
After solving it just for fun I ran the solver from sudokumaker on it. It took 0.2 seconds. Brute force, okay. But I've had that brute forcer timing out on some of my puzzles.
Haven't watched Simon's solve yet, but it looks like I may not be a computer. I got the "difficult" part that the computer apparently must have been hopeless at (eight from five won't go, so...) reasonably quickly, and then found myself with a classic sudoku with twenty-four given digits that took me ages. Eventually, I sort of bifurcated on the two possible values I then had for r2c2, and found that one of them caused a clash of 2s in c5 _just_ within the limits of my working memory, so it wasn't _real_ bifurcation. (At least, that's what I'm going to claim, very sheepishly.) Now, let's see how the maestro did it... Edit: 38:19 is precisely where I got stuck, so now I'm all eyes... Right, I didn't know it was useful to be bent when doing sudoku. Very interesting.
the good old RSTUVWXYZ wing on the digits 1-9. To see this we have to ask a facetious question how many 1's would we expect in these 9 columns in a complete puzzle and you can see that forces the 1's into these 9 rows......
I went the long way round at the beginning - if 5 wasn't on the thermos (therefore in box 5 in line with the thermos), you end up with either too many high digits in columns and/or too many low digits in rows.
I tried once an 8 digit anti knight sudoku. Then I looked at how a computer approached it. It spent about 100 lines explaining to me why a specific digit could not occupy a specific square. And yes, being a computer is more about brute logical force than elegance and aesthetics.
Delightful puzzle. Difficult but not devastating. I think, more than any other "miracle" type sudoku I've seen on this channel, this seems the most implausible, that JUST those thermometers create a unique solution without so much as a knight's move constraint, or anything else, in the rules or puzzle. It seems absurd that this could solve at all.
Round about the 40:40 mark, when Simon decides colouring the 89 pairs is not going to work, he could have looked at colouring the 12 pairs instead. It all drops out pretty quickly from there as it resolves the 1s and 2s pretty cleanly. Not quite trivial, but you don't need to hunt down that bent triple.
at 20:00 why do the 5's in box 2 & 8 have to be on the thermos? if you make r4c3 a thermo from 2 to 7, r6c7 a thermo with 123567, and the other two thermos going from 4 to 9, then there's enough room to place the 5's elsewhere, or am I tripping?
Took me about an hour and a half, but I did it! I had to backtrack to filling in the thermos TWICE -- the first time, I made a mistake on the deduction about the 3's on the bulbs, and the second time I made a typo that led me down a wild goose chase for 15 minutes until I hit a contradiction. But I got there!
i downloaded your SUDOKU puzzle app, but how can i create my own varient puzzles to send others to solve or potentially show Simon (would be a dream), at the moment i just use a whiteboard to write my puzzles on!
Blashyrkh here, fantastic solve Simon!
The computer gibberish in the screenshot is actually _after_ giving it the center 5. Without that, the computer solver is totally lost.
After all that talk about computers, I will have you know the puzzle is in actual fact (half) computer-generated!
The break-in idea with the pointing high/low digits leaves some wiggle room for the precise layout of the thermos. So, I wrote a Python script that randomly places thermos (while keeping the break-in), and then has Rangsk fantastic solver check for uniqueness. This generated unique grids at a rate of about one per minute. After a while I got back to the computer to find a few dozen possible layouts. That left the task of selecting the one with the most appealing solve path. Most of the grids fell into two categories:
a) Singles only after the break-in, which felt a little bland.
b) Hardcore sudoku after the break-in, with multiple advanced techniques, which would have scared off any human solver.
There were a select few however with solve paths like this one: Standard sudoku after the break-in, and then some sort of spicy stuff at the very end. This is exactly what I was looking for, and overall I liked this one best. By now I know three different ways to tackle the end:
1) The computer-y way: Finding a finned x-wing (whatever).
2) My way and that of many commenters: Colouring 12.
3) Simon's way: Finding multiple x/y-wings until the thing collapses :D
Thank you for your comment!
That particular solver is lost. I've a solver that uses a different approach, using a MILP algorithm and it doesn't have a problem
I'm loving the double meaning of center clash, with it applying to both the inital break-in to discover the 5, and to the impact 5 has as the central digit of the thermometers. Couldn't imagine a better name for this puzzle.
@@adenwellsmith6908 Isn't that solver than just guessing? Any solver can solve this sudoku if it guesses.
@@somebodystealsmyname no. It’s constraint based. The constraints remove choices. But towards the end, you could say that there is an element of suck and see.
But from the output of the solver you can see if it is branching and bounding. It’s quite rare.
The interesting part to me is that the solver is generic. You just have to add a new constraint type not change the solving bit.
Fog of war will be hard to add but possible.
The other interesting one is creation. Can you get it to work the other way
"I dont think this is the most intelligent way to go about this, to just stare at it"
Well thats disheartening, that was my strategy
It’s crazy how a british man solving a sudoku can make you so happy. This channel is golden 💘
Well said! This is a wickedly cool puzzle. Simon's joy as the logic unfolds is so charming...
yea sometimes I am too busy to pay attention to the puzzle but put it on simply because Simon having so much fun doing them and just seemling like a nice blocke is just brightening the mood
Simon needed to read a lot of books, to get really good at speaking English like a poet, and then he needed to work at an investment bank, where he was dying of boredom, before he stumbled into live-solving variant SUDOKU every day during the so-called "pandemic" lock-downs. This channel is a confluence of several miracles, but now that it's here, life just wouldn't be as much fun without it...! Long live CTC...!
I just wanted to let you know that I find your username unusually fun to say. That is all
@@Kelarys you made my day 🤣
"I am not going to write an 8 in this square"
*puzzle finished with 8 in that square*
NostraSimonus?
His final digit was also an 8, and it would have been even more hilarious if he'd referred to that final cell.
"I know exactly what Mr. Goodliffe would do! I'm not going to do that!" (does it immediately, and uses it to great effect during the puzzle solve)
Lol yeah was going to comment that he probably would've been effed if he didn't go full goodlifffe
Computer must have chugged a case of Red Bull, because it was using all the wings.
37:25 if you colour in the 12s, you can quickly see that r7c9 sees both colours, which unlocks the rest of the puzzle.
That's actually what I did and came here to write 👏
I'm very pleased with myself that I was able to solve this. A testament to CtC, which has taught me so much about how to approach puzzles and what questions to ask.
I feel the same and I couldn't have said it better myself.
Was going to write the same. My solution wasn’t as elegant, but I got there and three years ago I would never have seen the first digit.
At least 75% of the time, when I get stuck and then I look at the video to see what I should have done, it was exactly what I already started to do but I didn't follow through with it. I feel like that's what this channel has taught me. That I pretty much know it. I just have to follow through with the thought. Think completely
Around the time I started watching this channel, there was a spree of puzzles that were similarly introduced as being resistant to computer solving. This intro made me feel nostalgic.
My first was him demonstrating how a computer completely failed to use the corner/ring set trick. I was immediately hooked.
I was just thinking the other day that I'm not really interested in classic sudokus anymore unless it's watching Simon solving computationally difficult ones. Really nice seeing this puzzle!
lol, I was *just* talking to my wife about how insane Sudokus were getting back in the day by requiring jellyfish, swordfish, bent triples, etc., and how our broadened palette of variant rulesets was making it "easier" to find purchase...and then today! BENTIPPLES!
I was thinking similar to myself only yesterday! Mark mentioned a Jellyfish in his last video and it's been so long I can't remember what one is anymore.
Back in the days that Mark would get stuck on a puzzle and just bifurcate and beat it into submission 😂
@@leftysheppey If I recall correctly, a jellyfish is two steps further down the rabbit hole than an x-wing, which is to say, four intersections of rows and columns that lock a digit into a subset of those intersections. That said, even if my explanation is correct, I would almost certainly never actually notice one on my own in a puzzle. lol
one day, Simon will be forced to resort to a 9th-dimensional wonkyfish
Hah. Conversations with my wife about sudoku are usually limited to "You spent two hours on a sudoku? Geez. Also, that's not how you pronounce Mephistopheles."
I used to avoid variant sudoku because I thought it made them too hard. When I started watching the channel four years ago I quickly discovered that complexity doesn't always equal difficulty, and that discovering beautiful interactions between the rules is way more satisfying than finding and applying complicated techniques that other people discovered for me.
The reason computers get "confused" by puzzles like this is that they're really good at applying the techniques given to them, but really bad at creating their own techniques. The "trick" to this puzzle in inventing a new technique that is only applicable to this puzzle, which is why a human can solve it without complicated guesswork.
I loved that the second half of the puzzle was plain sudoku. I couldn’t believe how much time was left after the thermos were done.
Once again, Simon proves his genius by spotting logic that I never could in a million years (nor can computers, apparently).
At the same time, Simon aggravates me to no end by absolutely refusing to colour 1s and 2s, starting from 38:30 or so.
He ponders colouring 8s & 9s, and I'm shouting try it with 1s and 2s instead. No bent triple necessary (although that was neat).
Likewise, I did it by coloring 1s and 2 without those final strategies. I prefer those strategies, but I couldn't because I missed placing a 9. (I still don't know whether the X-wing in 1s was superfluous.)
Problem with computers is, now that Simon spotted it, it can probably be added to a computer solving program, so that theoretically humans are “better than computers” just once for any kind of problem…
I love you guys and thanks for widening my horizon in the world of sudoku
omg... this is the first time I finally understand what an X Wing is! That point about you have to look at the columns now was a huge lightbulb moment! I got so stuck at the end too lol
A beautiful start and a beautiful finish for that puzzle! Even if there was an easier way, I liked yours.
Said it before but it's worth repeating. I'd heard so many explanations on X-Wings that just failed to land with me. By contrast, the first time I heard the way Simon explained them not only could I spot them, I could use them.
Same applies to his explanation of what he calls bent triples. He hasn't needed to break that out for a while but it's some thing that was an important addition to my toolkit.
Is not coloring an easier way to find “bentipples”?
@@tessabrisac7423 Not for me, but I can see how it would be that way for some people.
I've had a go at puzzles where there was no obvious coloring solution and got stuck unable to spot the bent triple. Coloring worked today but I'm glad I watched the video. This time Simon's bent triple explanation clicked. I'm excited to try to use it on my own. Hopefully not too long before another bent triple puzzle is featured.
Normally on these "computer can't solve it" puzzles, I attempt to find something for a few minutes before I acknowledge that I'm not going to spot the hybrid "mutant--finned-jellyfish x skyscraper x 2-string kite" technique that Simon's going to pull out of nowhere and just enjoy the video. But on this puzzle I could feel there was some tension between the thermos, but couldn't quite get the foothold i needed. Started watching the video and as soon as Simon said "lots of low digits in the rows and high digits in the columns" the SET alarm bells went off in my head and my mind snapped into focus on the exact SET needed. Paused the video before Simon said "SET" and was off to the races. Amazing puzzle and a testament to what I've learned from countless hours following the channel.
Simon gave up on colouring too easily - much unlike his usual self - after going at eight-nines (@40:30). It really is like some of the old-school puzzles - colour the one-twos and note that r3c9 and r7c6 are different and looking at a 128 pencil mark in r7c9 which finishes the puzzle with, for the lack of a better desciption, alacrity. What a gorgeous puzzle!
That’s how I finished it too.
That break-in was absolutely brilliant. After that I was able to solve it quite fast. I know I shouldn't brag, but I'm actually quite proud of that :).
I think I've become better at recognizing situations where the question of how many digits can certain cells share is very helpful. Finished in 21 minutes this time.
Same as my time, for same reason! I'm normally much slower and worse at all these, but managed that same insight on this one, that 125 must be the middle row, and that consequently the higher bulbs must be different, which quickly lets you place the 3/4 bulbs and fill everything in.
At minute 39 of the video, I contemplated whether it was finally time to use the fabled RSTUVWXYZ-wing.
🤣🤣🤣🤣🤣🤣
FUNNY!
I instead used the rarely discussed RSTUVWXYZ bent nonuple!
What does those letters mean
@@Cinakkuma its a little joke about strategies of x wings, y wings ang xy wings and a lot of other nonsense which are used to discuss how to solve really hard sudoku before Simon help us learn that some puzzles humans can solve that computers can't
At 44:53 Simon says he is sorry he went to that level of complication. The computer confirms he did need to do so. :)
More elegant way was to color r9c5 and look for it around the grid
19:30 Haven't you just proven that, as far as 5's go, there are four thermometers (all of them) with a 5 on it in a very restrictive place.
Either the 5 is second to last on a thermo, meaning the bulb is a 1. Or the 5 is second from the start so the end of the thermo is a 9.
Since there are 5's in restrictive places on all 4 thermos, the thermos are diagonally identical to each other except for the bulb and end digits.
There will be two thermos going 1, 2, 3, 4, 5, X. And there will be two thermos going X, 5, 6, 7, 8, 9.
And now it's just a matter of finding why one diagonal works and the other doesn't. Then you get a whole bunch of digits to play with.
The comparison to natural moves in chest is apt. And finding beautiful solutions is an admiral pursuit for our community indeed
Simon has mentioned that really great puzzles are a discovery, unearthed more than constructed. This was one of those
simon: lets look at every digit except 6!
me: no!!!!! you want to look at 6!!! it will solve the puzzle. lol
or not. lol those last 4 digits are stubborn
Great break in!
At the end I used colouring of 12 which helped to eliminate some options in box 9.
I was so proud of myself for getting the break in, and following the logic around the paths.. then I got completely stuck. a bent triple.. how marvellous!
I finished in 54 minutes. That was such a cool break-in and overall a very enjoyable puzzle. I greatly enjoyed working out the limitations of the center square and was so happy to have gotten that 5 in the center. It may sound egotistical, but I am real proud that I was able to solve this puzzle, while a computer couldn't. It makes me feel like I am Kyle Reese fighting back against the machines. I very much enjoyed this one. Great Puzzle!
great vid, love how you got caught up there at the end. reminded me of the videos from when i first started watching
The criss-cross pattern with the high and low digits converging so that rows and columns needed to put 3 of their polarised-digits in the central box, such that there was an overlap, and thus '5' NEEDED to be in the middle of the grid, that was a fun conclusion to discover...!
23:08 ... an incredible break-in, and a challenging (but totally fair) ending
Nice puzzle!
Proper Goodliffing at 8:07
No amount of Goodlliffing could help me break into this one. I stared at it for 15 minutes, and watched Simon make short order of it. He's a f'ing wizard.
Great puzzle! Fantastic setting by Blashyrkh!
That bent triple at the end is also called a Y-wing.
Nice, solved in 30:16! I am quite pleased with myself for spotting the logic around the 5's so quickly in this one.
Great puzzle and solve. Great puzzle title as well, gave me just enough information to look for the break in.
Very nice puzzle. 24:33. Goodliffing; finding the low/high constraints on the central row and column; placing the 5; the rest of the high/low analysis in the cross boxes; reducing the thermometers; noticing the opposing 13s and checking a hypothesis. After that it's mostly cleanup.
So enjoyed watching Simon solve this. He makes it look so easy.
27:30 finish. The way I figured out which of the thermos started with 1-2 was to see that if they were in the bottom left and top right, then the top left and bottom right thermos would both start with 3-5, which would place 1-2-3-5 in the center row of box 5. This means that it has to be the other way, and the bottom left and top right bulbs cannot be the same digit. So they are a 3-4 pair. An excellent puzzle, so much fun!
Same way for as I did it by the sounds of it. Took two minutes longer to complete than you though but then I messed about with breaking the 1,2,8,9's for a good 15 minutes...
Very fun puzzle from you Blashyrkh!! Clever use of length and placement of Thermos!! Thank you Simon for showing us the computer gibberish!! 😁
38:13 today. I've lost to Simon hundreds of times over the years, but this is the first time I lost by two seconds.
1:30:20 - I can’t tell you how chuffed I am to have done that unaided. Massive thanks to Mark & Simon for giving me some of the skills needed to be able to solve these amazing puzzles.
Wicked cool puzzle, and awesome solve. I stared at it for quite a while, never came anywhere near the break-in, way above my pay-grade. With the break-in supplied, it mostly unfolds really nicely until the end-game bits. Where it gets sticky around 38:30, I never see things like the x wing 1's / 2's / the bent triple gambit. But I do love that Simon went there, because he's a f'ing wizard, and I love that he noted that there was probably an easier way to handle the end game. I colored 1's & 2's which placed 8 in c9r7, and it collapsed...
you guys should do another Sandra and Nala puzzle. I miss those.
I love their puzzles!
@@katiekawaii I lost my oldest dog back in June. I was kind of hoping there would be one since then. I would also love a S&N DLC for the CTC app.
45:11
This was an insane ride. Started with a total Goodliffe of the thermos and gradually whittled them down with look-aheads at first before patterns started emerging. So clever how each part restricted and constructed the others.
It's not that the computer doesn't 'understand' set theory (unless it wasn't programmed to do so), it is that it looks for other ways first that are more common (as in a random puzzle is more likely to hold them) and uses those to solve the puzzle instead. It should be a very simple change in the program to make it show set theory as far as my understanding of how this solvers work.
The solvers essentially go and mark each blank square with what it can be and eliminate possibilities, they do so by using a loop.
The loop contains all known techniques and when it eliminates something it starts back from the beginning.
So if the solver uses an X-wing to eliminate a possibility, it will go back to the start of the loop and look for naked singles again using less and less common logic.
The computer commonly spots what I call an X in X, imagine 2 can only go in box 1 in the top row, same for 3 and 4, that is 3 numbers in 3 squares, so nothing else can go in there and everything that seems those 3 cells can't have any of those numbers. That is an XYZ wing, since we ran out of letters after Z, for 4 in 4 we call it WXYZ, VWXYZ wing for 5, etc.
Importantly is the last technique the computer uses, backtracking, it essentially guesses a certain number in a certain box and sees if the puzzle solves or breaks, if it breaks we can eliminate that as a possibility. After guessing once we might need to guess again, this isn't really a problem, it is just if e.g. r1c1 is a 1 r2c2 can't be a 2, it can go fairly deep.
The solver is using X in X first because generally X in X appears before any set theory, so whoever made it made it look for X in X before set theory, which leads to the solver finding solutions faster. Now it doesn't take long for the solver to find the solution to a puzzle, even just with pure guessing, but you wanna be efficient because you can make a program that uses this algorithm to look for puzzles with certain conditions that have a unique solution and there is gonna be quite a lot of them that fit them so you wanna look as fast as possible.
Very enjoyable solve.
After looking at it for a second, I already knew what the "problem" is. Then it turned out not to be big enough problem. After thinking for a while, I realized I need to "brute force" it - just to see how exactly the limitations work. I filled the thermos with correct solution first try - well, something that looked possible, that is :) Only then I saw "the cell" - of course, every sudoku has 5 in the middle (not spoiler, that is common saying) With all thinking done ahead, just filling single cell quickly finished the thermos. Then it looked little annoying - the rest is regular sudoku - but actually still lot of very nice little logic discoveries were needed to finish the solve.
Thank you for showing it, and Blashyrkh for making it!
So.......Goodliffing gets us started. Submit, Simon! Succumb to the glory that is hyper notating!
I'm glad to see a video where Simon does teach us X-wing logic! I got stupidly stuck on how to logically solve the end of this puzzle, and resorted to "Well there's a lot of pressure on R8C8 (the 1-9 in box 9)" and realized 1 didn't work and 9 solved the puzzle. It worked, but still. Out of all the set theory tricks and mathematic combinations, we haven't seen much of what can solve a non-variance sudoku problem like this one.
Edit: Also, to add what Simon says at the end of the video about "I hope I didn't overcomplicate it," I actually did throw that ending into a sudoku solver, and indeed, what Simon did is the correct method to find the solution. It forces you to find the two X-wings, then solve the XY-wing (or "bent wing" as he called it) to eliminate 8 as a candidate from R7C7. It wasn't more simple than what he did.
I see him start unwinding things at 30-odd minutes and then notice that there's still about 15 minutes more to the video... I was excited to figure out what happened. 😀
The X-wings and Y-wings (aka bent triples) are back. Yeaaayyyyy.
00:35:32 for me. Loved the break in and even then it was a great puzzle afterwards! Kind comment.
I managed to solve it in 19:01. I didn't fully verify it was the only solution, but my intuition at the start turned out to be correct.
A phenomenal 23:37 solve on this. I just grok'd the crucial concept early by luck and kept just finding the right areas to focus on. Pretty chuffed about this one.
This one is pretty approachable, with very nice logic. Managed to be faster than Simon for once.
Simon: making super complicated logic
Also Simon: saying that basic solving techniques (e.g. x-wing) is more complicated than the rest of the solution 😂
Or as he would normally say "shame on them, making me do Sudoku in a Sudoku puzzle"
Simon smarter than a computer, confirmed.
Computers aren’t very smart actually
@@ahouyearno☝️🤓
@@ifroad33 imagine calling out someone for being nerdy in this particular comment section.
No. He is smarter than the people who program the computers.
Computer are the most idiotic form of existance ever.
If they seem smart, it's just because some nerd wasted too much time studying math, to apply it in the craziest way possible to make something work with only 1 and 0
Solved in 75:18! I love these puzzles with simple rules :)
Finished in 20:17 and I'm really proud of that. Wish I knew the double-click trick on the thermometers though, I could have saved a couple more seconds! I generally use tactics more like Simon's rather than Mark's but I almost always pencil mark thermos. So much logic seems to jump out from that
I used set theory to break in - made a set of rows 4&7 and a set from columns 4&7, and you can see that there must be a 5 in two of the thermo high ends, and a 5 in two of the thermo low ends, which got me to where Simon was ~20 minutes in. I found that easier than what Simon ended up doing
This was a really great sudoku - not quite my favourite among recent puzzles, because once the thermos were done it became more about spotting things and less about deduction, but it was a great example of a puzzle where I really had to use types of reasoning I learned on this channel.
I think you must mean rows 4&6, cols. 4&6. Just sketched it out quickly and there are only 4 degrees of freedom left over when you take out the thermo cells....that is, the six leftover cells in the rows must exceed the six leftover cells in the columns by a minimum of 32. Since the maximum difference is 2x(24-6), that doesn't leave much room....
31:18 for me. I could just see that 5 in the middle too and then the thermos seemed obvious because the other way around would prevent 3 in the middle box
I love how your last few puzzles have been pure suduko puzzles with a simple rule set, yet challenging. Much better than your previous videos where there is are about 15 lines of text for the rules
I can accept that Simon is human. Mark, I'm not so sure about.
27:10 How to find out that the particular choice of 2 vs. 5 doesn't work in a different way? First I clipped down a few more candidates: the 79 pair clipping block 5 column 4, followed by the 7s in block 5 column 6 clipping the rest of column 6. Then I *tried* to make R6C1's 134 resemble more the 14 and 13s in the other bulbs. The puzzle responded one better: 1 in the 134 cell broke the puzzle.
32:30 After filling the thermos, I centermarked the grid, finding some pairs and triples and placing digits in the process. I found a (probably superfluous) X-wing in 1s. I colored 12 pairs, changing to 1-not1 pairs, and that collapsed the rest of the puzzle.
38:40 Much more complicated? Come on, now. There's an X-wing in 1s.
43:20 Earlier on, I missed placing 9 in R3C7. Consequently, I never got the X-wing in 2s, and never got the Y-wing (bent triple) which would have collapsed the puzzle. Fortunately, I tried an alternate path.
I’m here for “Simon wings “ everyday! 😂❤
27:22 I saw this due to high/low colouring. Where red means that a cell is the higher options notated in the cell and blue means that it's the lower options. Curiously, looking back, I can't recall you ever using this technique.
That was a very cool puzzle. I got more stuck in the middle than I should have but after bit of a break I saw what I'd missed.
28:02 for me. Nifty puzzle!
The X-wings and bent triple at the end weren't necessary. Just colour the 1s and 2s, and you'll find that r7c9 sees both, so can't be either, and the puzzle falls down.
The humour of this is Simon's solve is the same as the computer's ...
I had an easier time with the break-in than Simon because I was looking at rows 4 and 6 and reasoning that the two subsets could only overlap in 3 digits, which readily brought ne to the same result. I didn't hold a candle to how smoothly Simon finished the puzzle after the thermos were done.
I would like to humbly nominate Ben Tipples as an honorary mascot of the channel ❤
does anyone else think it’d be hilarious if Simon and Mark streamed the game Thank Goodness You’re Here? I’d love to see the opinions of my two favorite Brits on it 😂
I've just had a look at the preview to this and I have to say I would LOVE to play this on a stream. Can anybody who has played it advise us on whether it would be suitable for our audience?
@@CrackingTheCryptic there’s certainly a lot of innuendos and jokes that would go over younger viewers’ heads, but nothing explicit and no curse words.
Fun puzzle. I found it pretty approachable and finished in 42 min broken into a lunch session and finished after work.
Update: I used set to achieve the same break in as Simon by comparing rows 4 and 6 to columns 4 and 6. This forced the 5's onto the wings of the thermometers, the initial box 5 constraints and the extreme thermo runs based on the placement of the 5s.
28:40 C1R6 pencil mark (compare to the other bulbs) was the HUGE cluebat I needed to break the symmetry
27:37 for me - and I didn't get stuck when I had that mess of 1289 cells at the end. :)
To get the 5 in the middle, I wasn't looking at the numbers on the thermos, I was looking at the numbers that go in the empty cells. For example, in boxes 2 and 8, four of those empty cells were from 1234 and one cell from 56789, and in column 5, there are four cells from 1234; when taken together, each of box 2 and 8 give two cells from 1234 and the two cells from 56789, which leaves the three cells in box 5 also coming from 56789. Then when you do boxes 4-6 and row 5, you get the cells R5C4-5-6 being from 12345, which forces R5C5 to be 5.
I think I was a bit lucky with seeing the break-in immediately, got 37:53. I'm pretty happy with that, it's a very nice puzzle.
At about 42:00 I realised that, if R1C3 is an 8, then R4C1 is a 9, and vice versa - meaning you can put a 1 in R1C1 - and the rest falls in place from there.
nice one, was quick to spot the break in, even the logic where the 1s have to go on the thermos is something i quickly got. Didn't see the bent triple in the end and had to watch Simon show me. So don't feel bad Simon, it seems that level of complication in the end was necessary ;) Thanks for the puzzle and the solve!
I ended up staring at that final state of the puzzle for a good 15 minutes having no idea what to do, then decided to color the squares, upon which I immediately noticed that r2c2 in box 7 could only go in r9c1, therefore in box 4 it could only go in r6c3. Since r2c2 could only be a 1 or 8, and r6c3 could only be an 8 or 9, it must be an 8, and everything unwound from there.
Beautiful puzzle.
After solving it just for fun I ran the solver from sudokumaker on it. It took 0.2 seconds. Brute force, okay. But I've had that brute forcer timing out on some of my puzzles.
I unwound the end by colouring the 1s and 2s, and there was an x-wing of a colour in boxes 3 and 9, making the 2 in box 6 that colour.
What a fantastic puzzle.
Once I hit 38:00, I looked at the remaining pair in row 5 and colored the hypothetical cascade. One of the two led to a break.
Solved in 31:30. I had to test some cells to see the break in. And the ending was a bit of a snag, too
Haven't watched Simon's solve yet, but it looks like I may not be a computer. I got the "difficult" part that the computer apparently must have been hopeless at (eight from five won't go, so...) reasonably quickly, and then found myself with a classic sudoku with twenty-four given digits that took me ages. Eventually, I sort of bifurcated on the two possible values I then had for r2c2, and found that one of them caused a clash of 2s in c5 _just_ within the limits of my working memory, so it wasn't _real_ bifurcation. (At least, that's what I'm going to claim, very sheepishly.) Now, let's see how the maestro did it...
Edit: 38:19 is precisely where I got stuck, so now I'm all eyes...
Right, I didn't know it was useful to be bent when doing sudoku. Very interesting.
38:30, getting the 5 in the middle was surprisingly quick, then like 25 minutes doing the rest of the puzzle.
41:37 for me. This was a really clever idea.
the good old RSTUVWXYZ wing on the digits 1-9. To see this we have to ask a facetious question how many 1's would we expect in these 9 columns in a complete puzzle and you can see that forces the 1's into these 9 rows......
Maybe a computer can't solve the puzzle elegantly, but it certainly can pencilmark the thermoes elegantly.
I went the long way round at the beginning - if 5 wasn't on the thermos (therefore in box 5 in line with the thermos), you end up with either too many high digits in columns and/or too many low digits in rows.
thank you for posting this - I could not for the life of me figure that out and he didn't say .. I was getting a little annoyed ;-)
I tried once an 8 digit anti knight sudoku. Then I looked at how a computer approached it. It spent about 100 lines explaining to me why a specific digit could not occupy a specific square. And yes, being a computer is more about brute logical force than elegance and aesthetics.
Delightful puzzle. Difficult but not devastating. I think, more than any other "miracle" type sudoku I've seen on this channel, this seems the most implausible, that JUST those thermometers create a unique solution without so much as a knight's move constraint, or anything else, in the rules or puzzle. It seems absurd that this could solve at all.
Round about the 40:40 mark, when Simon decides colouring the 89 pairs is not going to work, he could have looked at colouring the 12 pairs instead. It all drops out pretty quickly from there as it resolves the 1s and 2s pretty cleanly. Not quite trivial, but you don't need to hunt down that bent triple.
at 20:00 why do the 5's in box 2 & 8 have to be on the thermos?
if you make r4c3 a thermo from 2 to 7, r6c7 a thermo with 123567, and the other two thermos going from 4 to 9, then there's enough room to place the 5's elsewhere, or am I tripping?
Finished in 31:33 with help from the video.
of course I fully marked the thermos!! Done in 22:13.
Took me about an hour and a half, but I did it! I had to backtrack to filling in the thermos TWICE -- the first time, I made a mistake on the deduction about the 3's on the bulbs, and the second time I made a typo that led me down a wild goose chase for 15 minutes until I hit a contradiction. But I got there!
I tried coloring 1s and 2s and it wasn't very fruitful. Let's see if Simon went with a similar way to how I solved it.
Looks like I followed Simon pretty well. Took a few steps in a different order but basically equivalent. I'm just not nearly as good at scanning.
21:10 here, really groked the break in quite quickly, then lost a bit of time looking for a bent triple at the end.
i downloaded your SUDOKU puzzle app, but how can i create my own varient puzzles to send others to solve or potentially show Simon (would be a dream), at the moment i just use a whiteboard to write my puzzles on!
11:22 for me. Very fun puzzle!!