Abusing GemCraft (an Armor Games flash game)
Some of you have probably played the flash game
over at Armor Games.
a screenshot of the first board, to give you an example. It's an overhead view
of your wizard house and a path, which you must defend from monsters using
gems placed in towers. The gems shoot magical bolts at the monsters which
As would be expected, you start out with a very small number of waves and a
small number of weak monsters, and you get experience, levels, and skills over
time. Eventually you get up to harder boards that require a bit more thought,
with options for harder versions of each board as you go up in level.
It's an entertaining game and only takes a few hours to beat. If you haven't
tried it and have some spare time, you may want to check it out.
GemCraft Endgame - Our Story Begins
Up to endgame, everything looks normal; here's an
image of the final
board. Like all the others, the monsters walk the path and it's your job to kill
them before they get to you. But this board is special - very special - compared to
the rest of the game. For the past four days, I have been off and on playing through
this final board working out some interesting details and interactions.
Something I accidentally discovered the first time I played the endgame board is that
putting staggered slow and mana stealing gems in floor traps is a very effective
way to boost your mana. The first time I did it, I had half a dozen or so traps,
and was amazed at how much mana I was able to pull in.
Thinking about this later, I started to wonder just how far I could take this.
I started testing various combinations of gems in traps, and before long I'd
managed a configuration that had the entire floor space covered in traps filled
with high grade slow and mana stealing gems. "high grade" back then was grade
6 or 7, which cost 4000 and 8000 mana respectively to create. With the whole
floor space covered, I was pulling in several million mana over the course of the
board, wheras on normal boards a few hundred thousand is the best that could be
Time passed, and I discovered some new things: that prismatic gems, the ones
with all different colors, are actually pretty easy to build once you know the
right shortcuts. And while they don't really seem to be that powerful in
towers, they're really easy to upgrade. What took me longer to realize is that
when put in traps against high hitpoint monsters, they're stunningly, absurdly
In fact, they're powerful at a level I wouldn't have thought possible. The
area effect and chain specials hit multiple monsters, apparently hitting them
with all the other specials. Slow is of course useful, but strangely enough
shock is the real winner. Shocked monsters don't move -at all-. What I
started seeing is that a group of four or five monsters would all be shocked
at once, stationary over a single prismatic trap, until dead.
This was exactly what I needed for the final board! For this board, they keep
throwing wave after wave at you, each with an exponentially increasing number
hitpoints. Beyond a certain point, all you can do is slow them down on their
way you kill you, and get as many of them on the board as possible before
dropping the final gem bomb to win. (The more of them on the board when
you win, the higher your final score.)
With half a dozen monsters immobilized per trap, I should be able to pin at
least a few hundred monsters down before throwing the final bomb. Pinning
them also makes splash damage vastly more effective, which might even allow
me to kill some.
Meanwhile, the prismatic gems are also mana stealing full tilt from all the
monsters they've immobilized. On low level monsters, this mana stealing
isn't really worth much because the gems do too much damage and kill the
monster quickly. But on high level monsters that are immobilized, the gem
hits again, and again, and again, sometimes thousands of times, each one
stealing some mana. Very early on, I began gaining mana so quickly that
I could barely manage to spend it all before the bar was full again.
For the early boards, I turned the rendering quality up as high as I could
get it to slow things down. This dropped the frame rate to the point where
I could complete enough actions to actually spend all the mana that was
coming in. As more and more monsters and traps began to clutter the
screen, I was able to lower the quality settings to keep the game playable
but not running too fast.
Eventually though, I was running at lowest quality settings and the game
lag was simply too much. Fortunately, Linux saved me! The next day when
I got up, the game was fast - far too fast to play, and I had to raise the
quality settings again. It was then that I realized that the Firefox/Flash
audio driver had failed overnight, as it is prone to do on my system. With
the audio mixer broken, the game was much faster and able to handle many
more monsters on the board.
Eventually even with broken audio, the lag and delay were simply too much.
I eventually decided to stop upgrading and let the map run on its own. After
several hours, I decided to simply call as many remaining monster waves as
possible before dropping the final gem bomp to win the game.
Screenshots of Endgame
Here are some screenshots with comments to give you an idea of what exactly
Image 01 -
this is the board near wave 148. I believe the monsters here are
on the order of a couple million hitpoints each. Every gem on the board,
except for the ones in my inventory, are grade 11 prismatic or higher. About
half the gems in towers have been upgraded to grade 12. I hit the mana pool
spell so many times that it costs 50k to do it again. I should probably
have hit it more before stopping.
Image 02 -
around wave 150 you get a handful of swarms that really inflate
the monster count on the board. I was thinking that these would be too much
and that they'd push probably halfway through the board.
Image 03 -
around wave 153, showing the swarm just before it starts to fall
Image 04 -
Wave 155 - the swarm waves actually never got past a third of
the path. After they cleared, the monster count started to drop dramatically.
Amazingly, the gems were keeping up with the incoming monsters.
Image 05 -
On wave 163, I gave up in disgust. At this point, the frame rate
was approximately one frame per second, even with everything set to the minimum
and with the audio in its broken condition. It was just too much effort to
build and upgrade gems. About a quarter of the board you see here is grade
13. Every screen update adds about 3000 mana to my total.
The board is 15x17, with 13 squares unusable for towers or traps. The total
gem count is 242 gems of grade 12 or higher. A grade 11 costs about 130k
mana, a grade 12 about 260 k mana, and a 13 about 520k mana. My estimate of
the total mana cost of the gems in this image is approximately 80 million
I figured it would take several hours for the monsters to walk the entire
path and get to the end. The plan is that when they get near, I'll drop
the last gem bomb and finish, and hopefully end up with a score somewhere
Image 06 -
I decided to go on a quest to see
what higher grade gems look like. This is a shot of a grade 16, which cost
an estimated 4.2 million mana to create. Note that I collected this 4 million
mana in the span of two waves, and that I neglected to spend at least another
four million that was lost because I was writing this up and not paying
Image 08 -
Wave 175. They're halfway across the board now. Each frame update
adds over 5000 mana to my total. Note that the monsters in the displayed
wave (wave 179) have 33 million hitpoints.
Image 09 -
Wave 181 or so. It occurred to me that I could wait a week for
the game to complete naturally, or I could do the unthinkable: call as
many waves as possible, as quickly as possible, then drop the final gem
bomb just before the clear and obvious end. This shot is the beginning of
that process. Unfortunately, it's taking on the order of 30 seconds per
wave calling them as fast as I can, and it's slowing down substantially
as the monster count increases.
Image 10 -
I've now managed to call up to wave 202. It's a testament to
the developer of this game that it's managed to hold up under all this
abuse. It's taking well over two seconds per frame to do display updates.
Image 11 -
What's this? Is that the end of the waves I see there? Is that
Image 12 -
We're definitely out of waves. Take a look at the hitpoints on
those runner monsters - over a billion each.
Image 15 -
The view just before the end. Take a look at how many monster health bars
are visible through the gem firepower display.
Image 16 -
this is just after I drop the final gem bomb to kill everything
on the board. in the center you can see the kill notify just starting to
appear: "1032 kills in a row". There were 1032 monsters on the board
when I dropped the bomb.
Image 17 -
The final score/victory page. Gems created: 22153. Longest kill
chain: 1032 monsters. The gem creation count all by itself represents
over 80 million mana spent on gems, and matches up nicely with my estimate
from image 05 above.
My conclusion is that it might actually be possible to beat this map
legitimately, by killing all the monsters in every wave, if you
have the patience to never stop spending mana on gem upgrades. I doubt
anyone will do it any time soon.
The major concern is that as the monsters increase in hitpoints, your
mana gain per hitpoint decreases due to upgraded gem damage; further, gem
upgrades only give about 50% increase in damage for every factor of two
spent. Since mana is dependent on the number of hitpoints the monsters
have for mana stealing, and the amount you can upgrade is dependent on
your mana, these two factors are multiplied. Double the monster hitpoints,
and you end up getting 66% as much mana for upgrades that yield about 66%
the damage/cost ratio as your previous.
In short, for every factor of two increase in monster hitpoints, your ability
to deal damage only increases by a factor of about 1.33.
The reason it's possible to get halfway through the board waves without much
difficulty is because you start with a tremendous advantage over the monsters
in terms of power level. By the 160'th wave, that advantage is whittled down
to virtually nothing. At the 230'th wave, the monsters have 250 times the
hitpoints of wave 160, while your ability to deal damage will have only
increased approximately 10-fold. It's possible that with optimization and
careful planning this factor of 25 difference could be cut in half,
but this still leaves the player at a hefty 10:1 or worse disadvantage.
To write Dentin a letter, send mail to: firstname.lastname@example.org
Looking for my
Alter Aeon main page
Connect to Alter Aeon