Questions about Spider and its strategy

* A brief history

The solitaire game Spider has been around for nearly a century, though the earliest reference I can find is a November 1937 article in Ely Culbertson's Games Digest, (Vol. 1, Num. 3, pp. 42-43).  The author is not credited, but solitaire authorities Albert Morehead and Geoffrey Mott-Smith were on the Games Digest staff, so it may have been one or both of them.   The article describes it as a well-known game, and several sources reference an article in Redbook magazine which stated the game was a favorite of President Franklin D. Roosevelt (I cannot find a reference to the date of the Redbook issue).   It is also mentioned in passing in W. Somerset Maugham's 1930 travelogue The Gentleman in the Parlour, although there is another unrelated game confusingly called The Spider, which appears in some early books.   It seems highly likely that the game is an American invention. 

When Microsoft released Windows ME in 2000, it added Spider to its suite of free games.  This can be played in four-, two-, and one-suited versions, and dozens of computer implementations have followed suit.  This added to the popularity of what was already one of the most popular forms of solitaire (among two deck games, it vies with Forty Thieves and Miss Milligan among the most popular and best-regarded games), and at least in electronic form, it has jumped into the top echelon occupied by Klondike and FreeCell.   It has also spawned a number of variants, which we'll look at later in the article.

* What are the rules?

The standard form seen today is played with two full decks of 52 cards.  Shuffle and deal four rows of 10 cards face down, forming ten columns, then deal three more overlapping rows face down.  Deal four extra cards to the first four columns, then ten cards face up, one on each column, forming a tableau of 54 cards.  There are no foundations or waste, and the remaining stock of 50 cards is used only for a special kind of deal.   The columns can be packed downward in any suit (e.g. a seven at the bottom of one column can be placed on an eight at the bottom of another column.   Groups of cards in sequence can be moved as a unit if they are all in the same suit.  Partial sequences can also be moved; that is, a longer series of cards in suit can be split and part of it moved onto another card, even of a different suit.  This is frequently necessary when consolidating suits.  If a column is emptied, any uncovered card, or a full or partial sequence in suit, can be moved to the empty column.  When play is blocked, deal one card from the stock onto each column (you will do this five times).  A complete suite (13 cards in the same suit in sequence from king down to ace) can be removed from play.  The game is won if you can remove all eight suites (2 in each suit).  

Note: The earliest descriptions, in Games Digest and in Albert Ostrow's 1945 The Complete Card Player, specify a 50 card initial tableau, four face-down rows and one face-up row, with a sixth partial Spider deal of 4 cards at the end.   But by 1949, Morehead and Mott-Smith were describing the present-day tableau.

* How often can I win?

First of all, let's talk about the standard four-suit game, with no peeking (you can undo a move, but once you have turned up a face-down card there is no going back).    I realize that the game is widely played in a different way, allowing unlimited undos and letting the player make trial moves to see which one exposes the most useful cards, but we will concentrate on the standard version.  Having the opinion of several trusted experts, I am led to the conclusion that a skilled player can win perhaps one in three games.  This is in fact the figure given by Morehead and Mott-Smith in The Complete Book of Solitaire and Patience Games.  Some of their figures are well-known to be far off the mark, but Spider is a game the authors undoubtedly knew well and played often enough to have a solid estimate.  Unfortunately their tips are only mildly helpful (we'll mention one of those later).   I am sure it is not much less than that, although some sources give a dubious figure of 1 in 10.  I would be interested to find out that I am wrong, and that the actual win rate is higher than that (remember, no peeking).

If the win rate for the four-suit game is around 1 in 3, what is the win rate for two-suit Spider?   Surely it must be well over 50%, perhaps even 75-80%.   And what about the one-suit version?    I can win 95%; surely an expert can win 99%.   Is the win rate as high as FreeCell?   Are the strategies for two-suit and one-suit similar to standard Spider?

* What is the early strategy for play?

Like many players, I am lousy at Spider, particularly the standard four-suit game.  Strangely, there is a serious lack of strategy guides for Spider (there is a recent book by Steve N. Brown, but it slow going without electronic deals to follow), though I have managed to find a few articles on the Internet, though at least one of these concentrates on endgame play. I can follow most of this, but I rarely get far enough in a deal to put their strategies into practice.   Most of my deals are hopelessly blocked by the third or fourth deal from the stock.  None of the standard works which contain fully played-out deals include Spider (Patience Games by Cavendish, The Complete Patience Book by Basil Dalton, or  How To Win at Solitaire by Walter B. Gibson).

So I am going to ask questions about strategy and see if anyone can help work out a basic strategy.    Let's start first with some basic questions and my guess at the correct answer.

Here is a deal from Pretty Good Solitaire, number 440954453.  There is only one move I am 100% sure is correct, so I have already packed the queen of clubs onto its king and exposed the seven of hearts.    

spider01.gif (72437 bytes)

What do I do now?   I understand that I am trying to uncover as many cards as possible, pack in suit wherever possible to gradually build up suit sequences, and to create empty columns in the tableau as quickly as possible.   So should I then start to build a mixed-suit sequence on the ten of hearts?   This will be hard to move until later, when I hope to have empty columns, but it at least will expose three more cards in turn.   As luck would have it, there is a nine of hearts under the nine of diamonds; if I had another ten I could put the nine of diamonds there and put the nine of hearts on its ten.  One tip given by Morehead and Mott-Smith is to make plays that can be unmade (putting a seven on one of two available eights) above all other plays, even plays in suit.  

So now I logically want to move the eight of spades.  Is it better to put it on the nine of diamonds -- an already mixed sequence -- or should I prefer multiple shorter mixed sequences to a single longer mixed sequence?  I put it on the nine of diamonds, exposing the jack of clubs, which I immediately pack on its queen (exposing a king of diamonds).   The mixed sequence 8-9-10 cannot be moved onto the jack, so I add the seven of hearts to it (exposing an ace of diamonds).

spider02.gif (100568 bytes)

Now I can only make plays with the ace and two.   Since the three of diamonds is slightly closer to an empty column (only four cards underneath), I assume it is better to pack the two of clubs onto one of the threes of spades (arbitrarily the leftmost one), exposing a jack of hearts, and packing the ace of diamonds onto the two, exposing a six of clubs which I add to the 7-8-9-10 sequence, exposing an ace of spades.     Have I done anything terribly wrong?  I've only exposed eight cards, and I already have to deal from the stock, giving this:

spider03.gif (125310 bytes)

I pack ten of clubs, ace of clubs, two of hearts, six of clubs, and finally ace of spades onto the two of hearts, emptying the fifth column.   I then use the empty column to move 7H-8S onto the 9D in column eight, and 9D from column seven onto TC in column one, and the TH onto its jack in column three.    I don't see how to use the empty column to move any further mixed sequences, so I am going to fill it.    At least one source advises against filling empty columns with kings, but how on earth am I ever going to clear column two otherwise?   I would like to put the king of hearts there and pack the nine underneath onto the 10-J sequence.  But I will trust that advice is correct, and instead I move the 6C-7C there, pack the 3D (exposing a ten of spades) and the AC-2C on the 4C.  Now I pack 9D on TS, and add the TC to the J-Q-K in column nine, and then use that ten to shift the 9D-TS back onto the JC in column 1.   I add the exposed 8D onto its nine, exposing a jack of diamonds.    I reempty column five by packing 6C-7C onto 8D.  

spider04.gif (186294 bytes)

Again I cannot see how to do any useful consolidation with the empty column, so I am going to fill it, moving a card -- JD-- from the column with the fewest face down cards.   [Is this correct?  It is the opposite of what I believe is correct strategy in Klondike].    This exposes a 3C, and I pack the A-2 on top of it.    I will not give the details of the rest of the deal as I played it, but merely show the gruesome ending:

spider05.gif (298743 bytes)

I did not manage to complete even a single suite of 13 cards (this happens to me much more often than not), and I still had 27 cards covered at the finish.    In more than a few deals, I do not even manage to get an empty column at any point.

* Why must I fill empty columns before dealing from the stock?

Morehead and Mott-Smith specify that all empty columns must be filled before dealing ten new cards from the stock.   This rule is followed by most, though not all, computer implementations of Spider.  What is the purpose of this rule?   Is it just to make the game harder?   Would it be better to deal into an empty column if you were allowed to?   The rule can potentially block you from winning late in a deal if you discard so many completed suites that you have less than ten cards left while there are still cards left in the stock.

* What are some variants of Spider?

The most popular variant by far, found in Microsoft's version as well as many other packages, seems to be the two-suit version, identical to the standard game, but with a modified double deck of 104 cards, consisting four complete sets of two suits (typically hearts and spades).   This is essentially equivalent to a game in which the object is to form sequences in one color, and in which partial sequences in one color can be moved  as a group.   Pretty Good Solitaire includes a game called Tarantula, where partial sequences in color can be moved, but complete suites must be in a single suit.    Two-Suit Spider is considerably easier than standard Spider; even an inexpert player like me can win one out of three.

Even easier is the one-suit version (usually dealt from a deck of 104 spades).    I imagine an expert can win virtually every deal, but I lose occasionally, and I have encountered a deal, number 74,219,361,482,650 in Hardwood Solitaire III, that I cannot find a way to solve. 

Although the rules of Spider are pretty standardized, one occasionally sees odd variants.  In some computer versions, such as Troy Miller's 1992 version and Vince Cecala's 1993 version, both for Windows 3.1, the columns with an extra card are one, four, seven, and ten, rather than the first four columns (this makes no difference from a mathematical or strategic standpoint).   George F. Hervey's books on Solitaire (Teach Yourself Card Games For One -- 1965, and The Illustrated Book of Card Games for One -- 1977) have an easier layout, starting with four rows of ten cards (instead of five rows of ten and a partial sixth row of four).   This gives the player an extra full deal of ten cards and a partial deal of four cards from the stock.  This is the same initial layout as Forty Thieves; perhaps Hervey got the two confused?  

I have never seen it implemented in a computer version, and the layout of 54 initial cards must be considered the standard.

Hank Mishkoff suggests a way to make Spider more challenging (he played it with the two-suit variation, but even one-suit is challenging).    Deal out the entire deck before you move any cards.   I suspect the strategy of this variant differs greatly from the standard game.

* What other games are related to Spider?

The two most common one-deck adaptations follow the rules of Spider exactly, but use different layouts.  Spiderette (an extremely difficult game I have never won), starts with the normal 45-card tableau of Klondike.  Will O' The Wisp, a much more reasonable game, invented by Albert H. Morehead and Geoffrey Mott-Smith, starts with seven columns of three cards.

The best one-deck relative, in my opinion, is Toby Ord's Autumn Leaves, described in detail in another article on this site.   This uses a unique rule where packing is allowed only downward in suit, but not necessarily consecutively (e.g. an eight of spades can be packed on any higher spade).

A very interesting two-deck adaptation is Mrs. Mop, invented by Charles Jewell and first described in George F. Hervey's 1965 Teach Yourself Card Games For One.  (The game is named after a character in a British radio show of the 1940's, It's That Man Again.)  It is a completely open game, one of relatively few open two-deck games.  The entire double deck is dealt into eight rows of 13 cards; play then follows the rules of Spider.  The early strategy resembles Beleaguered Castle: it is necessary to clear an entire column almost immediately and another column or more fairly soon afterwards.   Once there is space to maneuver, the strategy can turn to consolidating suites of 13 (which resembles Simple Simon, but the suites in Mrs. Mop are removed when complete).  Jewell's rules require sequences in suit to be moved as a whole (that is, a sequence of cards in suit cannot be split, with only part of the sequence moved).   At first I thought it would make the game impossibly hard: partial sequence moves seemed necessary to consolidate longer sequences in suit, both for making extra space and for eventually completing suites of 13 cards.   However, I have now found that the game is playable with the stricter rule, though it takes some more care in manipulating sequences (e.g., T9876 and 7654 in the same suit must eventually become part of two separate suites).

This article is copyright 2021 by Michael Keller.  All rights reserved.  This file was revised on April 25, 2021.