It’s official. I believe many Spider Solitaire programs are biased – both for mobile phone and PC. Biased means if a player wins too many games, then the software will stack the cards against him in subsequent deals. Apologies if this comes across as a broken record, but “spider-software-bashing” is what inspired me to create this blog in the first place.

Let me state at the outset this is only my belief, and it comes with a number of caveats. I won’t expect regular readers to agree with my POV, but at least they can hopefully see where I am coming from. Following the rot13(erirefr fuvg fnaqjvpu) principle, I will start with a large number of caveats, then bring in the good stuff, and finally summarise these caveats. I assume we are always playing at the four-suit level (although it is possible to perform similar experiments with less than four suits).

The Caveats

• Firstly, I don’t expect regular players to be able to reproduce my results. You will need a very good aptitude at the game. In particular you must be able to win practically every game at the four-suit level with enough perseverance and skill. Fortunately undo/restart is permissable (but not reject). Once you reach that ability you also need a “burn-in period” i.e., play enough hands to let the software deduce you have a good very aptitude for the game.
• Secondly, I don’t have very many data points. The point of this article is to merely explain why I believe many programs are biased, and I am not claiming conclusive proof. Ideally, if time permits, I would prefer to play 100 games instead of 10. Oh, did I mention I am returning to work on Monday? In any case, if you are intelligent enough to play decently at the four-suit level then you are intelligent enough to draw your own conclusions.
• I have no convincing explanation of what a software developer would gain by rigging the RNG.
• A conclusive proof will entail some detailed knowledge of statistics (e.g. what is the meaning of statistically-significant, Komolgorov-Smirnov test, etc). I am nowhere near being in a position to submit a paper to Journal of the American Statistical Association.
• I tested this on an earlier version of Ninja Monkey’s code, not the one I’m currently using now for the current game.
• If you are stuck at 10-20% win rate then the bias is either non-existent, or doesn’t concern you. If you are winning above 20% then this is probably relevant. If you are serious about improving your game then all bets are off: your win rate is irrelevant and this is most definitely worth paying attention to.
• I write this with a heavy heart – one of my readers plays a certain online server, which I now suspect to be biased.

Now Onto My Actual Claims

The certain online server in question is this one. Initially I thought this was kosher after some “experimentation”. However, one of my recent losses cut a little too close to the bone. I then made a serious effort to win 10 consecutive games for the first time – using RESTART if necessary. Note that the software forbids undoing 2 or more moves in a row (and single undo is illegal after dealing a row or removing a suit). I immediately sensed something was off.

I did manage to win all 10 games, but most of them were practically unwinnable without undo/restart. And not for lack of trying. Needless to say, if a Spider Solitaire program penalises a player who wins a few too many with undo/restart, then you can guess how it will treat a player who wins a few too many without these privileges.

Bad Beat

As you might have guessed, I recorded the identity of every card in each of the 10 hands and ran Ninja Monkey’s algorithm. NM played each hand 50 times, for a total of 500 hands. Only for 2 hands did NM manage to win at least once. Whereas for random hands, NM should practically always win at least one in 50. I had similar results for a number of different programs.

Admittedly I may have tilted after a few games and failed to play to the best of my ability. But if anyone accuses NM of doing the same to corroborate my claims I don’t wanna hear about it.

I should point out one or two hands were grannies. Not exactly walkovers, but I would expect even my grandma will have a legit chance of winning sans undo. The term GRANNIE was apparently coined by a Scrabble expert – if you get all the power tiles then even your grandma can’t rot13(shpx) it up. In any case, one of my victories can safely be described as an EARNING.

My gut feeling says the software company wants to “equalise the players” so that (ideally) every single hand is either won by every player or lost by every player. Hopefully I don’t need to explain why this is a very bad idea.

Since it is beyond the scope to discuss the detailed mathematic specificity (with apologies to Kevin Rudd), I will simply present the initial game state for each of the 10 hands. I recommend the reader play the hands out for himself – first attempt to win without undo/restart, and if that fails then see how easy/difficult it is to win with undo/restart. The reader can also perform a control experiment by dealing 10 random hands. For extra credit: try to guess which two hands gave Ninja Monkey at least one victory in 50 tries.

List Can Be Any Colour As Long As It’s Black

As a result of this experiment, I have personally black-listed a number of Spider Solitaire programs, both for PC and mobile. From now on I only accept recognised programs like Microsoft Windows, or programs with source code (e.g. a GitHub project).

Games that “look and feel slick” are a red flag. Most of the bad programs I have encountered go beyond the basic “six-week-internship-prove-you-know-Javascript” thingy. While it is true that correlation is not causation, I would rather put my money on something that looks tacky instead of slick. Yes, the tableau cards do not flip automatically which is admittedly stupid, but at least you know what you’re getting. And the developer seems like a cool guy – anyone who’s willing to publish their name like that is almost certainly kosher. I found two look-and-feel-slick mobile phone apps, both dodgy and both had a similar “auto-complete” function when every column was either empty or reduced to a single straight-flush of any length. They’re probably from the same company.

Conversely if something looks like a basic implementation of Spider from someone who is learning the ropes of version-control in GitHub, then it is reasonable to assume there is no effort to punish a player who wins too often.

Yes, I hear you say, what possible motive do software developers have for punishing players who perform too well? True, not all Spider apps offer players a chance to enter competitions for “free” provided they sign up for PayPal, give away credit card details, their home and work addresses and the number of times they use the toilet every day. I was asking the exact same question and don’t have an answer. What I do know is that my numbers don’t lie – if Ninja Monkey says something is off, then Ninja Monkey says something is off.

IMHO, the onus should therefore be on software developers to prove they are kosher. Guilty until proven innocent if you pardon the inversion of a terrible cliché. If you can write a decent half-page essay on poker player Veronica Brill without help from Google, then you will certainly know where I am coming from.

Conclusion

I have decided many Spider Solitaire programs are not worth playing for both mobile phone and PC and I have blacklisted a number of those. But I wish to reiterate the fact this is my opinion only and my claims may well be difficult to substantiate for average players. Do not expect to see my name in lights if you subscribe to Journal of the American Statistical Association. But at least the reader has something to go on if he wishes to investigate for himself.

Finally, I wish to apologise if this comes across as a long-winded rant. But I believe this topic was worthy of more than the usual number of paragraphs.

And if you believe that some program is biased then you can also sit back and listen to the beat.