Game that needs work: Werd
Nick Bentley
to give it's own entry to solicit some help. I've been on a two year
quest to design a perfect information (or as near as possible to
perfect information) word game that doesn't stink. This has turned
out to be a phenomenally difficult task, much harder for me that the
geometric/mathy abstracts that I'm used to designing. I've had plenty
of really nice concepts, but none of them survive formal playtesting.
So it is with this one. Nonetheless, I personally like this game
alot, and I'm still holding out hope that I can find some way to
correct the problem that's came up in playtesting.
The main problem is that it induces debilitating analysis paralysis in
my playtesters. Way too much. How can I reduce analysis paralysis,
without removing what I regard to be the defining element of the game,
which is that your score depends on how many letters your word shares
with your opponent's most recently played word (see rules below).
Anyway, I'm hoping that by raising the issue here, one of you guys
will have something in your bag of tricks that will simplify the
game. Here are the current rules:
1. It's played on two side-by-side grids, each 9 cells wide (although
it can be played on different grids, 9-cells wide seemed ideal.)
2. You own one grid, and your opponent owns the other.
3. Players take turns. On each turn you write down a word in the
upmost available horizontal row on your grid. Your score for turn is
the number of letters that your word shares with your opponent's most
recently played word, times the number of letters in the longest word
that you formed *vertically* when you added the new word
horizontally. For example, if you wrote "tan" on your first turn, and
then on your second turn, you wrote "cat" such that the "t" from cat
is directly below the "a" from "tan", then you would have formed the
word "at" vertically. Since "at" is 2 letters long, your score for
that turn is 2 times however many letters "cat" shares with your
opponent's most recently played word.
4. the first letter of your word may be in any cell, as long as the
word fits completely in your grid.
5. No word may be played twice, and you may not play a word that
contains in its entirety a word previously played by any player.
6. There's a bonus if your word is an anagram of your opponent's most
recently played word.
7. Game ends after say, 8 turns. Highest score wins.
8. If you think your opponent played a fake word, you can challenge
it, and if it is a fake word, you're opponent gets zero points for his
turn. If your challenge fails, you must skip your next turn. Fake
words *are not* removed from the board after they've been played.
9. Since the score is undefined for the each player's first turn, the
first turn of the game works differently:
-To begin, a word is picked randomly from the dictionary, and then
player 1 writes that word into his topmost gridline. He also adds any
number of points he wants to either player's score column. Then
player 2 decides whether to play as player 1 or player 2. This acts
as a pie rule, and the randomly chosen initial word ensures that each
game will proceed very differently.
-The score for player 2's first turn is the just the number of letters
that his first word shares with the randomly chosen word that player 1
wrote down on his first turn.
Now, somebody will suggest that I try simplifying the score, such that
your score is just the number of letters that your word shares with
your opponent's most recently played word. I tried it, and it doesn't
word. The game because much more stagnant. It needs some kind of
multiplier or modifier, but I think the particular multiplier featured
here is not right; it's a big source of analysis paralysis.
Nick Bentley
Torben Ægidius Mogensen
> How can I reduce analysis paralysis,
> without removing what I regard to be the defining element of the game,
> which is that your score depends on how many letters your word shares
> with your opponent's most recently played word (see rules below).
>
> Anyway, I'm hoping that by raising the issue here, one of you guys
> will have something in your bag of tricks that will simplify the
> game. Here are the current rules:
>
>
> 1. It's played on two side-by-side grids, each 9 cells wide
>
> 2. You own one grid, and your opponent owns the other.
>
> 3. Players take turns. On each turn you write down a word in the
> upmost available horizontal row on your grid. Your score for turn is
> the number of letters that your word shares with your opponent's most
> recently played word, times the number of letters in the longest word
> that you formed *vertically* when you added the new word
> horizontally.
What if you form two mor more vertical words? Does only the longest
count, or do you get extra for the others?
Let us say that the first word on your grid is "bat" and you play
"today" under it such that you form "at" and "to" vertically. Will this
score more than positioning the new word so you only form "by"?
> 4. the first letter of your word may be in any cell, as long as the
> word fits completely in your grid.
>
> 5. No word may be played twice, and you may not play a word that
> contains in its entirety a word previously played by any player.
It seems like you can obstruct by playing the words "A" or "I".
Granted, they won't score much, but some players may do this if they
can't play anything better. Maybe you should limit word lengths to
three or more letters (many word games do).
> 6. There's a bonus if your word is an anagram of your opponent's most
> recently played word.
>
> 7. Game ends after say, 8 turns. Highest score wins.
>
> 8. If you think your opponent played a fake word, you can challenge
> it, and if it is a fake word, you're opponent gets zero points for his
> turn. If your challenge fails, you must skip your next turn.
Why not let the falsely challenged word count double instead of letting
the challenger lose a turn? This seems simpler.
> Fake words *are not* removed from the board after they've been
> played.
>
> 9. Since the score is undefined for the each player's first turn, the
> first turn of the game works differently:
>
> -To begin, a word is picked randomly from the dictionary, and then
> player 1 writes that word into his topmost gridline. He also adds any
> number of points he wants to either player's score column. Then
> player 2 decides whether to play as player 1 or player 2. This acts
> as a pie rule, and the randomly chosen initial word ensures that each
> game will proceed very differently.
> -The score for player 2's first turn is the just the number of letters
> that his first word shares with the randomly chosen word that player 1
> wrote down on his first turn.
I think some of the problem with analysis paralysis is if you have too
many choices that are not obviously bad.
One change that might reduce the number of sensible choices is to make a
word score the number of letters it shares with the opponent's word
_minus_ the number of letters it has that the opponent's word does not.
For example, if the opponent wrote "bats" and you write "basin" you get
3-2 = 1 point, so it would be better to write "bans" instead, which
scores 3-1 = 2 points. Obviously, this doesn't account for vertical
words nor that "stab" or "tabs" seem like even better choices.
The idea is that you would get no benefit for looking for longer words,
so you stop searching for better solutions earlier. Also, you can
remove the extra bonus for anagrams.
You could also make play simultaneous: After each round, you swap grids
and secretly add a new word underneath the bottom word. You reveal
simultaneously and score. You could add that a player has to finish in
no more than 30 seconds (or 15?) once the other declares that he is
done.
Simultaneous play also allow more than two to play, since there is less
down time. You could also do away with the random seeding and let each
player write the first word before passing his grid on to the next. You
can require these initial words to be exactly (say) four letters long
and give penalty to players that use the same initial word.
Torben
Nick Bentley
> count, or do you get extra for the others?
Only the longest counts. I did try a version where you got extra
points for more vertical words, but this made the analysis paralysis
*even worse*. As it is, the main source of analysis seems to be this
vertical word scoring mechanism, and allowing multiple vertical word
scores just made it worse.
> It seems like you can obstruct by playing the words "A" or "I".
> Granted, they won't score much, but some players may do this if they
> can't play anything better. Maybe you should limit word lengths to
> three or more letters (many word games do).
There is a strong incentive not to obstruct in this way. If you play
only tiny words, and your opponent plays only long words, your
opponent will have a great deal of freedom in how he constructs his
words, which will lead to long vertical words that you will have an
extremely hard time matching. Consistent short words seem to lead to
bad defeats.
> Why not let the falsely challenged word count double instead of letting
> the challenger lose a turn? This seems simpler.
I chose the challenge rule to match scrabble, since this is what
people are familiar with. But agree that yours is probably a bit
better, not in the least because it means the stakes are higher for
longer words.
> One change that might reduce the number of sensible choices is to make a
> word score the number of letters it shares with the opponent's word
> _minus_ the number of letters it has that the opponent's word does not.
I tried this once before, and I had mixed feelings about it. One
issue I had is that it made the game feel too static to me. What I
mean is, in a game with scoring, it helps if there's some potential
for big, game-changing scores. Like when you bonus and triple word
score at the same time in scrabble, or huge captures in Go. This
version of the game didn't have it; it was more a game of inches,
which felt a bit tedious. I should mention however, that I didn't
play this version with the anagram bonus included, and that might
change things. So I'll go back and try it.
> The idea is that you would get no benefit for looking for longer words,
> so you stop searching for better solutions earlier. Also, you can
> remove the extra bonus for anagrams.
Even with the original rules I posted above, there's little incentive
to maximize word length. The reason is that, if you make a word
that's longer than your opponent's word, you don't get any extra
points for the extra letters, and in addition, you are setting your
opponent up for a big score on his next turn. On the other hand
though, it is easier to generate long vertical words with long
horizontal words, so there is *some* incentive to go long. Not sure
how much yet though.
Because the vertical scoring seems to be the source of the problem, I
think your instincts are right on. I just wish that your subtraction
idea could be modified somehow so that big scores become possible.
Another good thing about the vertical scoring, which I wish there was
some way to preserve, is it introduces an obvious an understandable
long-term strategy component. Since it takes many turns to construct
a long vertical word, you have to plan ahead to try to make it
happen. That is one aspect I like.
> You could also make play simultaneous: After each round, you swap grids
> and secretly add a new word underneath the bottom word. You reveal
> simultaneously and score.
I like this idea alot. I'll definitely try it out. thanks.
> You could add that a player has to finish in
> no more than 30 seconds (or 15?) once the other declares that he is
> done.
I'm a little hesitant to impose time limits. Feels a little
artificial to me. Good games seem to impel you to making moves
without being forced to.
> Simultaneous play also allow more than two to play, since there is less
> down time. You could also do away with the random seeding and let each
> player write the first word before passing his grid on to the next. You
> can require these initial words to be exactly (say) four letters long
> and give penalty to players that use the same initial word.
I like this idea as well, and will try it.
Thanks very much.