- Pretty Good Solitaire

Klondike winnable percentage?

hi all,

I've recently got interested in playing klondike but I'm seeing a conflict between percentage winnable in play and what others are saying are winnable. I've learned that regular is more winnable than the "draw 3" mode. I just play regular and I make use of undo/redo. Anyways, I've seen winnable percentages quoted as low as 80% (in the PGS version) and as high as 91.5%, however after playing game numbers 0 to 482, my success rate has stayed between ranges of 97.52% and 98.37%. Interestingly, I've read that it is actually quite difficult to determine the true winnable percentage, although I thought that a brute force algorithm could surely achieve this? I'm not even confident that the games I've failed on are truly not winnable either, at least in some cases. Some moves that initially appear to be obviously "the way to go" actually prevent success, I found.


  • >I've read that it is actually quite difficult to determine the true winnable percentage, although I thought that a brute force algorithm could surely achieve this?

    This is actually a fascinating discussion, one that I have been having (mostly with myself πŸ˜‰) for nearly 40 years. (My first attempt to get an accurate approximation was on a Commodore VIC-20 back in 1982; I ultimately switched to a simpler game, Aces Up.)

    It is certainly possible to get a fairly accurate percentage, within a margin of error, with an algorithm. Obviously, the nature of random (technically, pseudorandom) deals means that an exact percentage is not mathematically possible for the general case. (It could be done precisely by adding constraints, say, winning percentage over the first 1000 deals with the Goodsol shuffle method.)

    A complicating factor in this case, as you noted, is that Pretty Good Solitaire (on Windows) has options which affect the winning percentage drastically. Did you know that Klondike was implemented in PGS more than 25 years ago? At the time, I doubt that accurate statistics were a primary consideration. Additionally, it is rather easy to cheat the statistics; I dare say every person in the main Klondike statistics (n.b., requires at least 50 games played) with a 100% winning percentage has either cheated or used an exploit.

    When we started implementation of Most Popular Solitaire (not even 20 years ago 😁), a higher emphasis was put on statistical accuracy, separating that single game into 4 different games based on options, Klondike (default: draw one, Kings Only) plus 3 bonus games: Klondike [easy], Klondike [deal 3], and Klondike [deal 3 easy]. We also added climb mode, which eliminated the "cherry picker" exploit, and we made it more difficult (though certainly not impossible) to cheat the statistics. (In researching Goodsol statistics for this reply, I see that the top climb mode score for one game is either an overflow bug, database corruption, or a cheat; I am not sure which. However, if those statistics were actually correct, that would indicate playing the game around the clock for more than 25000 years straight, which seems doubtful. 😜)

    Based on the climb mode scores, I see that just a few have reached 80% in the default Klondike game, the highest that I saw being 82.0% by a player who is active in this forum (RichardScotland) and an excellent player, so it is a trustworthy result. If you switch to Klondike [easy], you will see that results better than 90% are quite achievable. According to a research paper published a couple of years ago, the winnability of Klondike (in what they called "thoughtful" mode, the effective equivalent of using the undo feature) is 81.956% ± 0.096%.

    If you are really interested in pursuing this academically, you can find the source code for the algorithmic solver used to determine the above percentage here: (Note that this is not an endorsement; I have never looked at this code, and I had some quibbles with the paper itself, but it is certainly better than anything I have at the moment.)

    I hope this provides some illumination. πŸ™‚

    Gregg 🐡

  • If you'd care to post the game numbers of any that you have recorded as un-won, I and probably other members would be happy to give them a thorough going over. In other games I play I find that making moves that are counter-intuitive often proves effective. At least allowing others to play them will establish if any of them may actually be winnable. I am keen to try. πŸ˜‹

    A British prime minister, Benjamin Disraeli said something fairly accurate about statistics.


  • Haaaa! Self inflicted injuries it's called. I thought I'd search for an awkward looking Klondike deal. This is honestly the very first one I selected. I declare it as definitely unwinnable. #1709954943. Go ahead guys. Waste your time trying to prove me wrong. You won't. ☺️

  • edited September 2021

    Wow thanks for the great replies guys! Is starting at game #0 and increasing by one each time "climb mode"? If so I've currently played 0 to 551 with unsuccessful thus far on the following 15 layouts (for 97.29% success):

    2, 14, 79, 90, 246, 255, 283, 341, 363, 365, 376, 383, 504, 510, 539

    #1709954943 didn't seem difficult? I hope I have saved this correctly so you can see my moves using undo:

    Just to make it clear, I'm using regular (1 card deal) and NOT "king only" option.

  • Cheers. I was using the default Kings Only rule.

    Btw, game zero doesn't really exist. If you select it you'll get a different deal every time.

    I'll let you know pdq which games in your list I have managed to win. Thanks.

  • That would be the answer Richard, I'm not using Kings only option.

    376 in list above is possible I just discovered.

  • You are obviously a competent player as you didn't list any easy deals. I expected to harvest more winners than I did because you play the easier any card in a space option.

    Most of us who have been playing PGS since Noah was a boy deselect autoplay. Don't allow it to make decisions for you. Additionally, with that off, you can "worry back", taking cards from the foundations back into the tableau. That can be useful.

    I have saved the games I have won. You appear to already know how to use them. That is a great start. You will see from my screenshot which deals you should replay. If you would like any solutions just shout.

  • I'm quite impressed that you got any at all! I look forward to retrying the ones you got. I have autoplay switched off, but trigger it manually quite often. Sometimes I use undo if I think it did something I didn't want or I simply want to check what it did move.

  • edited September 2021

    Thanks. Undo and redo get used a lot by me. It saves a lot of time if you are not constantly replaying games to get back to the same necks of the woods.

    Glen McD ! My partner Liz was born a McD. Now you're curious which one. 😊

  • Thank you so much for getting those out Richard. I ended up getting all 5 out, but having to learn two new tricks to get there. I wasn't watching for when to keep a certain suit exposed, so that they could be transferred to the foundations. The other trick is when you get stuck, to look for cards of same number and color, where only one came from tableau and the other didn't need to come from draw pile. Solution is to undo and leave on draw pile in this case.

  • Absolutely correct. It also helps when you know a game is winnable. I always use every tool in the box when trying to win some difficult deals in a whole bunch of different games.

    Welcome to our Forum btw. Collectively we will tackle any problem games that anyone cares to post. We have a high strike rate.

  • >Is starting at game #0 and increasing by one each time "climb mode"?

    Basically, although as Richard pointed out, deal #1 is the first deal; there is no deal #0.

    The games are played in order (so everybody plays the same deals). The score is the total of all deals, so there are two approaches: either play for the highest score on each deal (my preference), or play more deals more quickly. Skipping a deal earns a score of 0 (to prevent cherry picking). Opening saved games is prohibited (to prevent saving an almost finished game and loading it in climb mode for a fast time). Each score has a verification code to resist tampering, and when tampering is detected, including missing scores, the statistics are deleted entirely (preventing a common form of log cheating: erasing losing scores). It is enough to keep honest players honest πŸ˜‰, and on a fair footing.

    >Most of us who have been playing PGS since Noah was a boy deselect autoplay. Don't allow it to make decisions for you. Additionally, with that off, you can "worry back", taking cards from the foundations back into the tableau. That can be useful.

    A little known feature of all games except the flagship Pretty Good Solitaire for Windows (i.e., all games with Climb Mode) is the temporary autoplay disable key. Holding down [Ctrl] on Windows, or [Command] on Mac, disables autoplay for any moves made while the key is held down. I keep autoplay enabled, but when I want to worry back or autoplay a different card, I simply hold the control key while adjusting the cards as I want them, then release and continue. Note that cards do not automatically play when the key is released, but they do after the next move.

    Just a little bonus information for players who read the forum. πŸ˜‰

    >Welcome to our Forum btw.

    Ditto. πŸ˜€

  • As Gregg has highlighted, Climb Mode is by far the best and fairest way to play solitaire. As everyone plays exactly the same deals, cheating is totally eliminated. Climb Mode stats are 100% accurate.

  • Thanks guys! πŸ˜€

  • I was expecting there to be a "start climb mode" somewhere in menus but find nothing like this? What am I missing please?

  • >I was expecting there to be a "start climb mode" somewhere in menus but find nothing like this? What am I missing please?

    I assume that you (like the majority) are using the Windows version of Pretty Good Solitaire. If so, it is not there. πŸ™ I apologize if I was not clear enough. In the main product, you can play the same deals by selecting deal #1 and always using 'Next' (rather than 'New'). Deal numbers are consistent among the different products and platforms (save the very rare bug 😌).

    For other desktop Solitaire games, including Pretty Good Solitaire Mac Edition (Mac), Goodsol Solitaire 101 (Windows/Mac), Most Popular Solitaire (Windows/Mac), and FreeCell Plus (Windows/Mac), you should find a 'Climb' icon on the main toolbar that can be toggled on and off, or a 'Climb Mode' option on the 'Options' menu.

    Our mobile Solitaire games, Pretty Good Solitaire Touch Edition (iPad), Pretty Good Solitaire Mini (iPhone), and the iPad versions of the smaller games, use a hybrid system that defaults to something like climb mode, but with the flexibility to choose arbitrary deal numbers (not counted in the climb statistics) and to go back to replay a deal number (and replace its score). (I think this may be the future.)

    I simply invoked Climb Mode here to suggest that its statistics are more likely to be accurate and make a clear distinction between the different Klondike options. Sorry for any confusion.

  • For anyone like you (Glen?) who prefers to compare their results honestly with other players, I'd suggest you buy the Goodsol Solitaire 101 collection. It costs next to nothing anyway. I have that collection as well as Pretty Good Solitaire. To be honest I tend to play the latter more. The 101 collection saves games in a different format so it's handy for me to be able to play someone's problem game in whatever format they are using. Incidentally, you can still play random games within 101. You have both options. Greg will keep you right. It's his baby. 😊

  • Thanks guys. Yes, I do use the Windows version of PGS and have essentially used the manual climb mode recently. I don't feel compelled to use something other than PGS to compare my speed with others but it is quite helpful for me to learn if any of the games I haven't succeeded at are possible, as Richard and I achieved yesterday.

  • Here's a bunch more I couldn't get out for anyone that wishes to try. Playing regular klondike (draw 1), not king-only and I don't use a solver but do make use of undo/redo:

    575, 613, 696, 845, 879, 901, 968.

    Maximum success rate was 98.46% after games 1-844.

  • edited October 2021

    Your apprenticeship is still not complete, oh international man of mystery. 😊 The two that you missed are both difficult. I'll have to replay the others because I labelled one of them as a .......

    and I can't remember which. (Sad old man).

    I posted the Jeff Beck version on this track as I have always particularly liked him.

    Btw,I DO use a solver. I keep it between my temples. 😁

  • Btw,I DO use a solver. I keep it between my temples. LOL!

    Richard, it reminded me of the time when I was in Jamaica, I could convert the exchange rate in my head faster than someone could do it on their calculator. Somebody in my party asked, "What's the trick?"

    I said, "It's called a brain!"πŸ˜›

  • edited October 2021

    Pity you left it there eh Ken? 😁 Seriously though, the modern reliance on calculators and spell checkers must surely put youngsters at a disadvantage when they don't have them to hand. I've always said that playing darts is almost as much about mental arithmetic as it is skill. I used to be pretty decent at darts. A friend from the pub lamented to me one night that he had never scored a 180 in his entire life until that particular evening when he'd managed it twice ........ "and you won both those games". πŸ˜ƒ

    What time is it in Florida?

  • Wow they were both hard but particularly 696. Much appreciated Richard.

  • edited October 2021

    I'm guessing that #575 was your "definintely maybe" game Richard. I tried again, it starts off like a bull at a gate but there seems no way through still.

  • Yep. That was the one. Two black queens did the damage. Had there been only one I could have got the KD onto the foundations and then, who knows? Games where you cannot empty a column are almost always going to be trouble.

Sign In or Register to comment.