Welcome to dxSudoku! Playing and practicing Sudoku strengthens our brains. When the problem solvers of the world have strong brains they make living here a better place for everyone.
This channel offers a large number of tutorial videos on specific puzzle-solving techniques and Sudoku concepts. The videos here are no nonsense with content solely focused on topic. Overlay graphics and animations are used to help you clearly see the patterns and concepts being demonstrated. dxSudoku videos are based on the following set of wiki pages: hodoku.sourceforge.net/en/tech_intro.php Or, google "source forge hodoku intro". I'm a big fan of Bernhard Hobiger, the creator of Hodoku. He is no longer with us unfortunately. When I started using Hodoku years ago, his way of thinking about Sudoku completely changed my way of thinking about it. Many of my videos take his ideas and add animated graphics to the explanations.
Problem Solvers Unite! Play Sudoku, save the world!
non capito --- how can a link be (bidirectional) strong and weak at the same time?? And what is the difference between strong link and either-or-link??
Any time a house has just two cells having the same possible candidate, in the house, one of the two must be the value of its cell. When this happens, if the first cell is "off", the other cell must be "on", hence a Strong Link. And if the first cell is "on", the the other cell must be "off", hence a Weak Link. And the links are bi-directional, both Strong and Weak links exist. Remember, there's nothing special about the meaning of the word "link". All it means is there's a relationship between two pencil marks or possible candidates. Concentrate on the relationship part and the not the visualization part or pattern. The link only exists because the relationship exists first.
I know this is old but I just found it. I wish you would have kept the row and column numbers marked so I could follow along quicker. When you say something like r2c5 us newbies have to count that out and in doing so can’t quite keep up with your explanations.
You make a good point. I should have thought of it. Now I'm thinking I will have to redo both beginner tutorials in the future. I'm currently working on my next book about solving ultra extreme puzzles. Once this get's done I will take a look at redoing the beginner videos. I'm really kicking myself with your comment because it's just such an obvious thing to do!!
I've been learning Sudoku for about a year now and always struggled when I came up to the extreme techniques. Always come back to your channel to review strategies and how to find chains/fish etc. Today was the day I jumped further into AICs and I've just gone through 3 Extreme puzzles on Hodoku using this technique in less time than it takes me to complete 1 (lower difficulty) puzzle using standard strategies. Amazing.
Wow, thanks for the great comment! I spent a long time developing the #107 AIC algorithm. When I finally got it working all of sudden I was solving Hodoku extremes I never thought I would solve!! I'm very pleased you succeeded on using this method. Sometimes it feels like there's 3 people in the world who actually get it. It's not the easiest thing to explain how to do. Thanks again for the great comment!!
Here's a nice Wiki page you might find useful: hodoku.sourceforge.net/en/tech_singles.php A "Full House/Last Digit" is pretty much the same thing as "Naked Single". I did a video where I explain how to solve a puzzle without using pencil marks. In this video, "Naked Single" is called "One Choice Remaining Type 1". I think this title is more descriptive. Type 2 in the video is a Hidden Single. Please try this video an let me know if it gave you a better explanation: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Cj4QcqLB5iI.html Also, here's another video I did titled "Beginner's Guide". I try to explain the Naked Single again this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-6WnFkrUt_10.html Sorry video #2 did not delivery for you. Hopefully these other 2 videos will fill in any gaps.
What's missing in this explanation at this point in the video is this: After coloring a candidate in dark red, if there's only one remaining candidate, then that remaining single candidate cannot be part of a triple. This is because a triple is composed of 2 or 3 cells. So when a cell as only one candidate, it can't be part of the triple. For row 5, here's the logic of my conclusion. 8 and 9 are disqualified because there are more than 3 cells having a possible 8 or 9 candidates. But now the 7 is disqualified from being in a triple because it's the only one remaining candidate in cell R5C1. The 2 is disqualified for the same reason in cell R5C7. And the 1 is disqualified for the same reason in R5C9. Since the 7 disqualified, that means there's only one candidate remaining in cell R5C8, which disqualifies the 5. With the 5 and 7 disqualified, there's only one remaining candidate in cell R5C6 which then disqualifies the 4. I've updated this video for finding Hidden Triple algorithm in the following video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-5pOJBMDxM04.html RU-vid does not let you edit a video so I had to create a new video. In this updated video, I talk about disqualifying candidates when a candidate becomes the sole remaining candidate in a cell. I hope this helps.
@@dxsudokuchannel Thank you for taking the time to explain why there is no triple there. I will watch your newer video. Honestly, I still do not trust my ability to find a hidden triple in advanced puzzles. I will do my best to reason a triple in order to eliminate a candidate and if I am not sure I am correct I will check the solution to see if I was right. This of course, indicates that I'm not sure and I am guessing. I do love Suduko puzzles and want to get better. Again, thanks for your response!
@@Bergarita Sometimes it's better to understand Hidden Triples in the context of subsets. Here's a video on using Naked Triples and Naked Quads as a better way of finding Hidden Triples using a concept I call "wrapping" which is commonly known as subsets: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-24JewFezAgM.html Keep working on it. Triples are tricky and just take some practice working on them.
With the XY-Wing, the two pincer cells create the kill zone. Since R2C8 does not share a house with R4C2, the short answer is no. The kill-zone for the XY-Wing is any cell sharing a house with both pincer cells, R1C7 and R4C2 in this case. However, the better way to convince yourself of something like this is first pretend the pivot cell R1C2=5. Then pretend the pivot cell R1C2=6. In each case, R1C2=5, and R1C2=6, what happens to each pincer cell R1C7 and R4C2 ? This idea of pretending a cell is set to a value is key to doing logic with advanced chaining sequences. The XY-Wing is in a sense a type of chaining sequence with the way the logic causes a sequence of consequence.
Great question!! By definition, a triple is composed of 3 cells where each cell has 2 or 3 of the same set of 3 candidates. I know this is a bit of a mouthful. When you have a triple, you have 3 candidates that must be chosen over the 3 cells or the house will be missing a 1-9 number. Keeping this in mind, when you have a single digit, this means the other 2 cells must be Naked or Hidden Pair because with a single digit, if it were part of triple, you would have to choose the single number or the cell would be blank. Since the other 2 digits being a pair, this would mean your single digit must be a hidden single. If it is a hidden single, you choose it as the value of the cell. If it's not a hidden single, then we must conclude it's not part of a triple. The hidden triple search algorithm just helps you see what's going in the house. You still have to be mindful of what is possible and what is not possible. I hope this explains it. I'm surprised no one has ever asked this question before. I think it comes down to when you have single digit, you must choose it as the value of the cell so it can't be part of triple. A little round about logic here. I guess the only way to see what I'm saying is by practicing finding them but also being mindful of hidden singles. I hope this helps. If it's not clear, reply where it's confusing and I will try again. Sorry if it's not clear. Tough question to explain.
I love your explanations, but in the minute 12:14 is way more easy to see the x-wing, I'd like to see x-cycles examples with no other techniques at sight. 🙏
I agree. Finding the X-Wing in this case, at the 12:14 mark, is way easier. I found x-cycles and Nice Loops to be a bit tricky to find. The thing about the x-cycles and the Nice Loops is the way you do the chaining sequence. In computer programming parlance the common way is a Depth First Search (DFS) like algorithm. In my later videos, especially the ones on AICs (single pattern chaining sequences are a form of AICs), I started using a Breadth First Search (BFS). The result is the same candidates get removed. But the BFS chaining sequence is really fast and easy to do. Here's a video on it: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-yq59G7rDxVo.html And especially the following video is really fast for finding AICs: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-UDg_GjyOYaI.html This last video is the way I solve ultra extreme puzzles on my cell phone by poking my finger around the puzzle. You have to remember the starting candidate of the chaining sequence in your head which sometimes I have forget, but even still, it works really well. Thanks for the comment!
Hi, I'm assuming you are talking about the 2-4 in columns 1&2 of rows 2 and 7, and at the 5:13 mark in the video. If so, here is the problem with the 2-4, the 2-4 in row 2 are givens. All the uniqueness techniques are about having more than one solution. Or a situation where you have to make a guess to solve the puzzle. Since the 2-4 in row 2 are givens, there is only one choice with the 2-4 in row 7. Since there's no guessing required to solve the puzzle, this is not an Avoidable Rectangle. I hope I got the right question.
@@headslmmer33 I once had a huge argument with someone about the 2-2-2 rule. I was trying to convince them if it were not 2-2-2 it could not be ambiguous. I could see it intuitive but I don't think I was able to convince them with a proof or with language. At some point I may spend more time on this idea. At some point you need to take a look at AICs. AICs are somewhat like a universal puzzle-solving technique and they get past the UR type techniques. I still do UR Type 1s and 2s because they are so easy to find.
I wish you every success with your new book. Last week, I created my hardest Sudoku to date, with a HoDoKu rating of Extreme 22,068 or 21,968, depending on whether uniqueness techniques are allowed. Oddly, the higher grade occurs when uniqueness techniques are permitted. HoDoKu is not without its quirks. Anyway, I thought that because you are perhaps the only person in the world who could solve it, you might like to do so. I know it's not as hard as others you've done, but it's well into ultra-extreme territory. It's based on a Philip Newman puzzle called “Tatooine Oasis”, recently featured on Timberlake's channel. However, only the pattern of the given digits matches that of the original, whereas the actual digits are a lot more regularly ordered. In a sense, it's a sort of “Tatooine Evolution”: 000 000 600 002 300 007 010 040 050 080 050 040 007 600 005 000 001 300 500 008 000 006 700 002 070 060 090 000000600002300007010040050080050040007600005000001300500008000006700002070060090 This is actually one of 11 grids in the series that exceed 10,000 points in HoDoKu. The basic layout, with slight variations, has turned out to be a very rich source of ultra-extreme puzzles. Alas, the prettiest of the 237 grids are not among the hardest. By the way, I checked out your videos on Sudoku construction, and before watching them, I could not have imagined how much your methods might differ from mine. When I have a bit of time, I'll give you an outline of my approach.
Dude! It's a work of art!! That's amazing puzzle. Really nice constellation of givens. It's definitely on my queue to solve. In the 3 video of series, I talk about creating puzzles using masking. I'm in the process of developing some Python programs for creating puzzles. I would love to hear your approach so lets setup a zoom session. It can't be this month, I'm crazy busy. July is going to ease up for me. So I will contact you. Just so you know, you can send me messages on Reddit using the handle dxSudoku. Again, the puzzle you made is really great!! Keep going!!
This one confuses me a bit. Because if i got to that point in a puzzle, the first thing i would notice is the unique rectangle type 1 in column 3 and 5 and in row 4 and 7. With the unique rectangle logic of a puzzle having only one solution, the cell C5R4 should be a 9. Am i wrong?
At which mark in the video are you seeing the UR Type 1. There's no right or wrong way to solve a puzzle. If you see a UR Type 1, then use it to solve the puzzle and don't be confused. The UR Type 1 logic is easy to understand because it creates a set of 4 Naked Pairs if you do not choose wisely. The result of the Naked Pairs is there's no way to solve the puzzle without guessing, or in other words, the puzzle has more than one solution because you have two sets of choices on how to solve the puzzle. I did a video just on the logic of the UR Type 1 show the 4 sets of Naked Pairs if you are interested: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-RQmgJryCa-c.html With Bug+1, the logic is very similar except if you do not choose the one candidate in the cell having 3 pencil marks, what happens is you have N number of bivalue cells. With all the N cells having bivalue pencil marks, again, you have to guess how to solve the puzzle. And with N bivalue cells there are multiple guesses beyond just two like with UR Type1. So the logic behind all the UR techniques is we assume our puzzle has only one solution, and therefore, we can remove pencil marks based on this assumption. However, I have to honest with you. There are two opposing opinions on whether or not a puzzle having more than one solution is invalid. I am on the side that claims a puzzle having more than one solution is okay, valid, and is not a problem. To solve it, you just show a solution, it doesn't matter which one. As long as you have a solution grid that abides by the one of each number in each house rule your have solved the puzzle. I have solved many puzzles having multiple solutions. They can be quite challenging!
The beginning of the video 0:14 the first example. We have cells with 5 and 3 They're in 2 rows and 2 columns and one of them has a 9. With that logic the top right of the rectangle (row 4 column 5) should be a 9
Let's run the algorithm. Start with column 1, it adds rows 2 and 9. Next, add column 3. It adds rows 2, 5, and 8. So far we have 2 columns and 3 rows. Then add column 6 which adds rows 2 and 5. So we still have rows 2, 5, and 8. And now we have three columns 1, 3, and 6. We add the final column 7 which adds rows 5, 8, and 9. So we have four columns which is okay: 1, 3, 6, and 7. And we have four rows: 2, 5, 8, and 9 which is also okay. Nice catch! It looks like this would eliminate many of the same target candidates in the example.
Hi i'm not sure if i understand the pattern yet but is there a jellyfish pattern in Exercise 1 Column 5, 6, 8 and 9 Because their candidates are on 4 rows (Row 1, 2, 6 and 9)
I play Sudoku all of the time and today I heard of Jelly fishes for the first time. I've watched a few jellyfish sudoku explanations and now understand what a jellyfish is, sort of. But, I don't really understand why we need to find jelly or any other breed of fish? I simply look for which numbers belong in which small squares. I don't really give a flying fish what the pattern is called. So these videos are totally pointless. I just use simple logic and it works every time, without mistakes. So you can take your fishes to where the sun don't shine! ;-)
Here's a really good wiki page on the most common puzzle-solving techniques if you are interested: hodoku.sourceforge.net/en/tech_intro.php All the fish are X-Chain variants. Just look for X-Chains with Group Nodes and call it the day. Or even better yet, just use AICs and solve everything.
Merci pour le beau commentaire. Cette façon de trouver les triples cachés est en fait très utile. Bien que je recherche habituellement des triples nus, des quads et même des quintettes. Je les vois juste plus rapidement que de faire des éliminations de candidats.
The 69-R1C3-Pivot with 29-R6C3-Wing1 and 26-R1C8-Wing2 do form an XY-Wing. The kill zone would be the intersection point of the two wing cells which has the number 3 as a given. But the three cells do form an XY-Wing. When I created these practice exercises there could be more than one solution. If you are interested in creating your own practice exercises check out this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-jPssW_W3iH8.html
If the SE 11.9 puzzle there were lots of uniqueness techniques (type 1s and type 2s). However, theoretically, although I never needed to do it, you know how I have my cell-phone way of doing the AICs with extensions, you could use this style of doing of chaining do a next level up chaining sequence to prove the current Forcing Chain's verity or contradiction. But so far I've never seen anything so complicated that the first level Forcing Net required using AICs with extensions to prove out the Forcing Net. At most, when stuck, I've used X-Chains a lot. I think in this video there's a Skyscraper which is an X-Chain variant. The other question I've been pondering is the idea of branch points within the build-out of the chaining sequence, that is, the nature of ambiguity within the constellation of givens. And the last thing still hanging on thinking about, is doing Forcing Chains and Nets this way still a form of guessing. There are brief moments where I think it may be and then I argue myself back from the ledge with the realization the constellation of givens are dictating the sequence of the chaining sequence so it's logic. These three questions may be around for a while! Thanks for you comments!! I do think Forcing Chains and Nets are a sort of "end-of-the-road" when it comes to puzzle-solving techniques. I've been working on a new series of videos for doing Python programming with Sudoku. In this new series, I'm going to create a bunch of programs for creating new puzzles and providing more advanced analysis of Sudoku puzzles at a certain point in a solution. I'm also interested in an idea I've been kicking around with regards to classifying puzzles according to different sets of solution paths.
My thoughts exactly!! I was literally stunned when I saw this result. I've been working on this puzzle forever! So the idea that this is even possible is very interesting to me. There's a certain about of complexity and ambiguity that can occur within the constellation of givens. So I've currently thinking on ways this could be leverage as a way of figuring the least amount of steps, or in this case, assumptions ever needed when solving a puzzle with a Forcing Chain and Nets. I also fascinated by the way this puzzle required 38 assumptions just to get 1 value set. So here is my challenge to anyone who tries to solve this puzzle. What is the best possible way with the least amount of work just to get ONE cell set to its proper value in the solution of this puzzle?
Yes, there may be more than 1. But in the exercise I only did one. If you are interested in generating your own exercises, I did a video on how to do it: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-jPssW_W3iH8.html You also might find this video interesting: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-di1aM5X0h6Y.html And this one will take you skills to another level: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-QvqGjXyPrIU.html Make sure when you search for X-Chains you also include Group Nodes in the chaining sequence. It opens up all kinds of new candidate eliminations.
Hidden Triple at 7:26-10:40 1. Examine this row. 2. These three squares are the only squares that can contain 2, 3, and 4 in this row. Therefore, the other notes in these squares can be erased.
You are correct. This is the definition of a Hidden Triple. The method in the video is just an aid in helping people see which cells contain a Hidden Triple. Here's another version of the same technique but for cell-phone apps: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-UDg_GjyOYaI.html You have good Sudoku eyes. If you haven't done already, you need to start looking at AICs. AICs are pretty much a universal puzzle-solving technique. So I end up doing AICs most of time rather than looking for stuff like XY-Chains, Uniqueness Techniques, Remote Pairs, and even X-Chains.
Yes, but the Quintet has 5 cells instead of four. But it's probably easier to look for a Hidden Triple than Quintet because a 2-2-2-2-2 Quintet would be very hard to see. Here's my new video on how to find Hidden Triples: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-33LnPOUH6Xo.html
Remote Pair before 7:50 1. A link is a line connecting two notes together. In a weak link the two notes cannot both be correct at the same time. In a strong link one of the two notes must be correct. 2. Connecting strong and weak links, in an alternating fashion, creates a "chain". Chains must always start and end with strong links. 3. Notice the chain of squares where each square only contains the notes 3 and 7. 4. Case 1: Imagine placing 7 on this end of the chain. Then, this note cannot be correct. 5. Case 2: Now imagine erasing 7 from that same end of the chain. Again, this note cannot be correct. 6. In either case this square cannot contain 3 or 7.
Which square cannot contain 3 or 7? At this mark in the video, I was showing either all the 37-greens are on, or all the 37-purples on on. It's one or the other. Also, take a look at this video. It's the same thing but searching for a good Remote Pair result like finding an X-Chain: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-d9IJdbFJ38w.html
Small error at the 1:19 mark of the video. There's actually four different types of contradictions. It is possible a contradiction occurs when there are no remaining open cells where a number can be placed. So the house will be missing one of the numbers 1 through 9. I will update the contradiction list in the next video.
9:27 "There is no kill zone at this point because the intersecting cell is between the two cells of the first strong link of the X-chain." I've no idea what you're talking about here. What do you mean by "the intersecting cell"? Why is R4C4 not in the kill zone (at this point) as it sees both the start and the end of the chain. This results in R6C6=9 and the outcome is the same your eventual conclusion.
Yes, your way of thinking about the logic works as well. In this section of the video I was trying to demonstrate what is known as a Nice Loop. But you are correct. You could have just stopped and kill the 9 in cell R4C4 which then opens up a Hidden Single with the 9 in cell R6C6. I'm surprised no one has ever made this comment before. This video is pretty old. Nice catch on your part!!
Ends up that brute forcing can be lots of fun in its own way! 👍 💡 Quick tip: To save time (and your mouse switch lifespan), for backtracking purposes you can make use of a nice little feature in HoDoKu which is located under the Puzzle menu and is called Create Savepoints 😉
Impressed that you find a way to solve these ridiculously hard puzzles. This is extreme example of Ariadne’s thread concept that Mepham from UK proposed when getting stuck in a puzzle. What are the standard notes you use when hitting a snag? Either backtrack or contradiction?
Thank you for comment. I will have to research Ariadne's thread concept. I will get back to you on that one. I feel like I've been trying to solve this puzzle for 2 years. It's like an old friend. It came from this thread: forum.enjoysudoku.com/the-hardest-sudokus-new-thread-t6539.html There's some really great solvers and creators posting in the enjoysudoku.com threads. I'm not the strongest mathematician but I've spent a lot of time doing computer programming. As you know, my #109 video I gave a general puzzle solving algorithm for what to do when you get stuck. For puzzles where the AIC logic based techniques don't work, I would use IBF only as a last resort. As I think you know I'm a big fan of Bernhard Hobiger (the guy who created Hodoku). His tutorial wiki pages are my favorite: hodoku.sourceforge.net/en/tech_last.php In this page, Hobiger says, "Brute Force is not really a technique: Place a digit in a cell and look, whether you get a solution or not." The thing about the puzzle in this video is I'm not sure it can be solved using logic alone. For a month I tried 3+ assumption AICs and went no where. But as I've stated in the video, for human being to solve this puzzle at all even if it is by Brute Force is something in my opinion. For this SE 11.9 level puzzle, I would just like to see one piece of logic capable of removing one pencil mark from the puzzle. At least then a person would have a fighting chance with lots of work. Let me know if you find something. Maybe I'll post a thread on reddit or enjoysudoku on this very question.
I just realized I did not answer your second question. First, IBF never get's snagged. You just keep backtracking every time you get a contradiction until you solve it. I did notice some cells where better choices than others. I think there might be some better rules around choosing the next cell in the sequence to reduce the overall amount of work in finding a solution. I tried to make my algorithm simple by just choosing the nearest bivalue or bilocation cell. But I think a better rule might be choosing the cell that causes the most number of pencil marks to be removed and/or the most number of values to be set.
You are not going to believe this but I've actually found a way to knock off a few candidates and open up a Hidden Single with this Ultra Extreme Level 3 puzzle. It did require me changing the way I do Forcing Nets. So I may do a new video on my improved algorithm for finding Forcing Nets.
Yeah this is an interesting approach. Bilocation is like the 1 in the bottom row? Ive only ever used this method to disprove candidates so I would start on one candidate earlier like one of the 2/4/8 that sees the 2/6 and set it to 2 in order to try to prove it wrong using the naked 6 but its not much easier and the result is just one less candidate where as your approach eventually does just solve the whole thing. In this example I would have instead set the 2/6 to 2 like you did because the 4/8 pairs looked so good, but there is no way I would have gotten anywhere as far as you did if I hadn't watched the video! Your method taught me to be more aggressive as in the example I would have opted to use the 7/6 on the first stall opting to keep the 4/8 pair in tact/ambiguous for as long as possible. Also with the way ive always done it before I would look for a Turbot, Kite, or W-chain type of diagonal relationship as a starting point because of how many houses it influences. But i guess that really doesnt apply here either so thank you very much!
Thank you for your thank you! I like how you have created your own insights into the technique. Keep building the strength in your insights. I've had some philosophical discussions with MVE. After my discussions, I've come to agree that MVE is a specialized form of the Brute Force technique. But with MVE it only takes 30-50 or so path changes to find the solution to the hardest puzzle. Where as Brute Force, may try have thousands of paths before finding the solution path. MVE is the only way I've been able to solve the hardest puzzles like this one: 12.3.....4...5......6..17....1..68..3...4..7....2...5..1....9....9....68.....9..7 Some people are critical of MVE because it is not solving the puzzle by logic but just pushing through the constellation of givens until the solution grid is realized. So in response to these criticizes I've developed a new technique outlined in this video similar to MVE but candidates are removed from the puzzle solely based on logic: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-yq59G7rDxVo.html I've done AIC videos in the past, but this #106 is different. Then, in order to solve even hard puzzles with my new method of doing AICs, I extended my AIC technique in this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-kjvSRMdEAXs.html "Multiple Assumption AICs with Extensions" is very similar to MVE but candidates are removed from the puzzle solely based on logic. Although I have yet to figure out how to solve the hardest puzzles using Multiple Assumption AICs with Extensions I am pretty confident it will solve it. So far I've only been able to solve puzzles having 2 AIC assumptions. I am looking for one requiring 3. The puzzle I referenced above requires at least 4. Having 4 assumptions is a crazy amount of work which is why I haven't solved it yet. Try the technique shown in video #106. I use it all the time when I get stuck solving normal puzzles. Normal meaning Andoku 3 Hard and Very Hard levels. I solve Sudoku puzzles on my cell phone while I'm watching TV at night. Sometimes it's hard to keep the AIC assumption in my head clearly but I'm getting better at doing it while watching TV.
@@dxsudokuchannel Nah you lost me on the second video :D . AIC is exactly what I meant I was doing, just didnt know it was called that. But I have to say having done a ton of them you can develop the ability to see regular AICs. I will still do them now but I would way rather do this MVE method than MAIC! The logic is less messy but basically the same, just that you set the value to ON and test it for a solution (think whole board nice loop) instead of testing to see if the OFF gives you a few kills (BVE might just give you a kill as well so not sure what put ppl off). I like to think of it as kinda like solving a derivative of the current puzzle, you still should be using advanced technique while doing it and that really separates it from pure brute force. On the other hand I actually always called my AICs brute forcing. In my head I would say: alright there is nothing more I can do to this with my current level of knowledge so time to do AIC.
@@dacrazyrazor Either way is good. AIC or MVE. I knew this guy who was really into uniqueness techniques. So he would solve puzzles with unique rectangles before anything else. Which "second video" are you referring to? Video #106? If it's confusing I may redo it.
@@dxsudokuchannel also unique rectangles feel like a gimmick a little because a puzzle should be constructed well enough that they dont come up. Of course it is a valid technique that should be learned so I guess it would also be bad if it never came up. :D
If you want additional practice, you can use Hodoku's Learning mode to create more exercises. Here's a video on the details of using Hodoku's Learning mode: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-jPssW_W3iH8.html The important thing to understand about Sashimi X-Wing is it's an X-Chain where one of the nodes is a "Group Node". When I am solving puzzles, I no longer look for Sashimi X-Wings. What I do is I look for X-Chains where any of the nodes might be a Group Node. Here's a video I did on Sudoku Links which has a section where I talk about Group Nodes: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-cVwV3MXXx5c.html And here's video showing my latest X-Chain search algorithm ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-di1aM5X0h6Y.html When I do X-Chain searches and include looking for Group Nodes sometimes there are unnamed patterns having more than just one Group Node. Once you get really good with X-Chains including ones with Group Nodes, you should study AICs: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-yq59G7rDxVo.html X-Chains use a single digit. AICs are X-Chains using many different numbers in the chaining sequence. Also, the way I show doing X-Chains in videos uses a Depth First Search (DFS) search algorithm. In my AIC videos, I use a Breadth First Search (BFS) search algorithm. The BFS algorithm is very powerful and you catch more candidates that can be removed from the puzzle.
Dear David, I've been working on a few variations of a 2024 SGP puzzle (Round 2, Puzzle 2), and I think I've come up with something a little special. I think you may be the only person in the world who could solve it. 020 400 700 003 050 009 100 006 080 009 000 008 060 090 020 300 000 900 070 300 005 800 020 400 006 001 030 020400700003050009100006080009000008060090020300000900070300005800020400006001030 HoDoKu rating: Extreme 14126. Andrew Stuart rating: No Solution Found.
It's a hard but I think I can do it. Hodoku is showing 14 Forcing Chains and 1 Forcing Net. My first pass was multiple AIC with r1c3 <> 5 followed by the bivalue cell 37-r3c5. I was able to knock the puzzle down to 7196 on Hodoku. I will keep working on it and post you the results here.
Hi Julian, I solved the puzzle you posted: drive.google.com/file/d/12JoNInN9P4RcWZJ3e-pzCQJpaRIuw8G9/view?usp=sharing Let me know if you can access the PDF okay. I did it in 6 steps but I think the 2nd step was a mistake. I think if I work at it I could probably do it in 4 or 5 steps. It took three Multiple AICs followed by three Single AICs according to my techniques. Still, at 14,126, it was a pretty tough puzzle. Normally, when I do multiple assumption AICs, there are only 2 assumptions needed. The next level of difficulty to find would be a puzzle that requires 3 assumptions in order to solve. I'm thinking a Hodoku puzzle around 20,000 with maybe 1 Brute Force technique.
@@dxsudokuchannel Hi, David, I opened the PDF with no trouble. I'm so impressed, and I'm most grateful that you took the time not only to do the puzzle, but also to explain how you did it. The Skyscraper, after all that fine crafting, was a little cherry on the cake. Finding a 20,000-point grid would be something of a miracle for me, as I usually develop on an aesthetic basis rather than a logical one. I did do one SET-based series of grids from scratch, but in general I have a very _ad hoc_ approach, and interesting logic comes sporadically as a bonus. Even finding a 14,000-point puzzle came as quite a surprise. I had already found an 11,000-pointer in the same series, but somehow failed to save it. Of course, if I happen across another ultra-extreme grid, I'll pass it on to you, but it could be quite a while in coming. Thank you for the PDF, which I will examine carefully in the coming days.
@@AnonimityAssured Hi, I did a whole series on creating puzzles from scratch you might find interesting: ru-vid.com/group/PLDT9hh28q4mxUzH5XcOfpbWRYOlT0fa6G If you follow my way of creating puzzles, what you would do to get puzzles having higher difficulty scores but still maintain the symmetrical beauty, is you would remove a set of 2 givens at a time and test if the puzzle is solvable with one solution. This video shows my method of making asymmetrical puzzles: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-lqhGB3iVKOU.html In the 3rd video of the series I create a puzzle having an Hodoku score of 12999. I'll play with the last one you gave me an see if I can create a 20K one from it.
for me it's still unclear whether I have to check for box criterias when I found a hidden pair in the same box and what I may remove then. For example: I find a hidden pair for the last column in the right corner that is fully in the box. The hidden Pair is (2,9) there. Is it a problem that other (2s) or (9s) are in the same box not interfering with the column? If not, may I remove the digits from those 2 cells in the column but also from the other cells those (2s) and (9s) in the same box being around? I certainly know that when I find a naked pair in a box that I can remove these numbers without any doubt from all other cells in that box, but this is still uncovered here and would be so useful to know.... That's why I find this video as many others way too vague. You need to explain stuff with a focus on special features, limitations and stuff, if you really want to make them more successful than the others. Too many feel like university lectures where people either want you to make the brain processes yourself to earn the degree or people lacking common sense on having empathy what people don't understand.
Sorry my explanation is not clearer. For many of the techniques, the only way I was able to master them was by using Hodoku's Learning Mode. With Hodoku's Learning Mode, it solves the puzzle to the point where the very next technique to do is the one you are trying to learn. The beauty and benefit of this is when you do Learning Mode 10 times in a row not only to master the technique but you have insights because you've seen so many flavors of the same technique. See video #24 on this topic. Hidden pairs are tricky to find. But it is important you understand the concept of what is a house in Sudoku. For a Hidden Pair to occur, there must only be only two pencil marks for each of the two numbers in the house. If there are are one it's a Hidden Single. And if there are more than two pencil marks, then the number can't be part of a Hidden Pair. Take a look at my #40 video titled Beginner's Guide. I explain more concepts which may help you. Also, there's another attempt at explaining what is a Hidden Pair and how to find them. Again, when I get feedback like yours, I try to improve my explanations. Also, my book has a pretty good explanation of Hidden Pairs because the person editing my book made me rewrite it until it made sense to her and she doesn't play Sudoku.
I did another video on Hidden Triples that you might find helpful: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-33LnPOUH6Xo.html This video shows me finding Hidden Triples with real-time video capture. I talk through the process so you can see my algorithm in better detail.
I went fast on this video because I wanted to show solving a puzzle having an SE rating of 9.1 under 8 minutes. For a more detail understanding of what I am doing, here is a list of links. The first is a wiki page on chaining techniques. This is a really good introduction into chaining sequences: hodoku.sourceforge.net/en/tech_chains.php This wiki page will show you Strong and Weak links. Here's my video on the different types of links used in chaining sequences: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-cVwV3MXXx5c.html Before can master AICs, you need to master X-Chains: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-GRAhOv4L1E8.html And here's an improve search algorithm for X-Chains: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-di1aM5X0h6Y.html In this video I use a technique called Multiple Assumption AICs because this puzzle has an SE rating of 9.1 and its the only way to solve it from my perspective. Computer programs have no trouble. But for humans, my method of doing AICs takes some practice but at least it can be done by a human being. Before you learn Multiple Assumption AICs, there's two videos to watch for single assumption AICs: First video #106: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-yq59G7rDxVo.html Second video #107: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-sC7WjuqGx-w.html #106 will teach the basics of how I do AICs. My method of doing AICs is very effective because I use a Breath First Search (BFS) algorithm. With BFS, I expand the chaining sequence out in every direction. It's easy to do and you find more candidates to remove with your effort. #107 is an improved version of #106. AICs are Sudoku's closest thing to a universal puzzle-solving technique as far as I can tell. Here's an interesting wiki page I like on this idea: www.paulspages.co.uk/sudokuxp/howtosolve/niceloops.htm Nice Loops are a special case of AICs. With my method, you find all the eliminations without having to identify any loops. There are many ultra advanced puzzle solving-techniques. The first wiki page I referenced has many of them. However, with my way of doing AICs, you will be able to solve a very large class of ultra extreme puzzles. In the following video, I give a detailed tutorial on Multiple Assumption AICs and I talk about puzzle difficulty levels: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-kjvSRMdEAXs.html If you get stuck on anything, create an "imgur" image and post the question here with a link to the image. I will generally respond pretty quickly unless I am on a fishing trip.
The reason the 7 can't be used in a triple is because of the 6. The 6 has 5 possible candidates in the house making up block 6. So the 6 is disqualified as being part of a triple. When the 6 in cell R4C7 is removed, it leaves a single 7 candidate. Anytime a single candidate show up in a cell, it means it can't be part of a triple because triples are composed of 3 cells having 2 or 3 candidates not 1. I emphasize this more in the next version of this video here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-5pOJBMDxM04.html And this idea of only 1 left means a candidate can't be part of a triple is shown really well in this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-33LnPOUH6Xo.html I hope this helps. Just post again if anything is not clear.
Thank you for putting up such a great video in finding Hidden Triples. I have tried the algorithm using Hodoku in PC and Andoku3 in cell phone. They worked perfectly. Here is my experience in using your method. In Hodoku, under Options, I have to uncheck “Show Deviations”. Otherwise, Hodoku may recognize a deleted candidate is an error but still shows up the deleted candidate in red color. As demonstrated, you opened the Notepad to record the first deleted candidate and its cell location for subsequent restorations. Paper and pen are not handy when playing sudoku in cell phone. In playing Andoku 3, I simply remember the first deleted candidate. I wonder if you know of such recording function available in Andoku 3. I am using Version 1.36.0. I wonder if you could kindly make a video to demonstrate the algorithms in finding a Hidden Quadruple in PC and cell phone. Thanks again.
Thanks for the great comment. You are spot on with the Hodoku settings. I change the deviation color to white. I should have mentioned it. You mentioned finding Quads and today was just trying to come up with a way. Off the bat, more than 4 cells are disqualified. But there needs to be some other rules for it to be effective. Mostly just houses with 8 cells because Naked Pairs and Naked Triples are easier. I'll keep you posted.
So I've been thinking about looking for Hidden Quads and here is my conclusion. You would never look for Hidden Quads. With 7 open cells you would look for Hidden Triples using the algorithm in this video. Or you would see a Naked Triple wrapping a Hidden Quad. For 8 open cells you would always look for a Naked Quad first before looking for a Hidden Quad.
btw, I'm still working on the video for the puzzle you gave me. Video #109 was huge amount of work. This video was relatively easy. I have solved your 2nd puzzle you gave me several times now so I am confident I will be able to create the video with no issues. I'll start on this new video this weekend. It should go smoothly.
@@dxsudokuchannel That's welcome news. I'm in awe of the power of your methods. My own Sudoku speciality is notation-free speed solving of easy and low-end medium puzzles. I've become quite adept at it, having competed (very successfully, according to the app I use) in 52 anonymous online tournaments over the last 15 months or so. Alas, one of the downsides of such an activity is that it kills all patience. When I watch videos in which solvers say things like “This 3 blocks this cell, and this 3 blocks these two cells, …”, I find myself thinking “Just put in the 3!”
I have been trying to slove Hodoku (unfair 1302). Correct me if I am wrong. It seems finding the hidden quadruple is the only path to solve this puzzle.
@@igbertau-yeung4874 Sorry, I'm confused. Where did you find the puzzle? Did it appear on this channel? Did I post it? Can you copy the grid into your reply? Here's an example, adapted from the 2024 Sudoku Grand Prix, Round 1, Puzzle 3. It's considerably than the original: 060 200 007 400 003 060 005 010 900 800 000 204 000 050 000 907 000 003 002 040 600 070 300 008 600 005 070 060200007400003060005010900800000204000050000907000003002040600070300008600005070 HoDoKu rating: Unfair 1374. Andrew Stuart rating: Very Hard Grade 203.
Here's what I believe is happening. After the 8 and 9 are disqualified because there are too many in the house to be part of the triple, it leaves a single 1 candidate in cell R4C4. When there is only one candidate in a single cell, this disqualifies the candidate from being part of a triple. This is because a triple is composed of 3 cells each having 2 or 3 candidates per cell. So having 1 candidate means the 1 is not part of a triple. After marking the 1 in red, the rest of the remaining candidates get disqualified because they become part of cells having only 1 candidate. I just created a new video showing this same algorithm with a real-time demonstration for working with cell-phone apps. Take a look at this new video and I think it will clear up any confusion you may have with this example: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-33LnPOUH6Xo.html Just post here or on the other video if you still have any questions. I hope this helps.
It's very close. I was just using the new algorithm on my cell-phone last night. I didn't expect it but I found a triple on the puzzle on I was working on with it. The video should be out this week.
It would be helpful for all players if you could make another video to clarify the steps. At mark 16:35, you mentioned players are able to use the same steps in cell phone app. For me, I am unable to find the website to download HoDoKu in cell phone. The only one I can find is Andoku which does not provide any available features to execute the steps you mentioned. If possible, please provide the android download link for HoDoKu and also make a video to demonstrate how the steps work in the cell phone app. In the future, I wonder if you could make videos to demonstrate how to play HoDoKu cell phone versions as most people have cell phone available all the time. My other suggestion is to provide playlist links for same category for all your videos so that users can learn all the similar features without searching. For example, under this video to provide links for dxSudoku #7, #8, #11, #25, #50, #51 etc. as they are all related to Triples.
Hi, thanks for all the great feedback. Hodoku only works on PCs running Java. For cell phones, I would recommend Sudoku Joy for iPhones and Andoku 3 for Android devices are the ones I like. You are correct in saying Andoku does not support candidate level coloring features. I've done a number of chaining videos working with cell-phone apps. So I have a good way in mind on how to do the algorithm for finding Hidden Triples using Andoku or any cell-phone app. So I will do a new video just after my current video is done. You can't play Hodoku on a cell phone as far as I know. You may want to post this question on Reddit (www.reddit.com/r/sudoku/) someone may have done it. I do have a playlist for all my videos on triples: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-BJncFpOsoDw.html&pp=gAQBiAQB Since I got involved with doing Sudoku, I'm always amazed by with just a 9 x 9 grid and a few simple rules there's just so much to talk about!!
I agree, it's a bit confusing. What I am doing is clicking on the 3 x 3 cells in the "Toggle Candidates" area. What is not shown is I kept a copy of what was in each cell so I knew what to click on in the "Toggle Candidates" area. I think I need to redo this part of the video. A better way is once the 4-5-7 hidden triple in R1C123 is identified, it should be written down. And then use Hodoku's undo button repeatedly until there's nothing to undo. Then after knowing the where 4-5-7 hidden triple is located, the candidates that need to be removed from R1C123 could be removed. Again, I apologize for this not being clearer. Conceptually, or logically, the way to think about it is if anything other than 4-5-7 occurs in the cells R1C1, R1C2, and R1C3, the house making up row 1 will be missing a 4, 5, or 7 from the set of values chosen for the 9 cells making up the house (or another number will appear twice, or one cell will be blank). What makes a hidden triple a triple is the 3 identified cells MUST be used for 4, 5, or 7 or bad things happen. I hope this helps. If not, please let me know. I'm currently working on a new video with AIC chaining sequences. But I would definitely do another video just on this way of finding hidden triples if you are interested.
I am unable to figure out how you can quickly add back candidates or pencil marks for cells 1,5 through 1,8. If possible, please show the steps in detail. Many thanks.
