1024 ways slot game, chi-squared test question

Hi Wizard,

As far as I know, the 1024 slots games with 5x4 reels are volatile because there're 1024 possible combinations that could payout.
So I try to apply Chi-squared goodness of fit test to see if the the winning patterns are distributed correctly.

So here is the brief description of the game.

Assume that the reel layouts for the game is as follow:

Q, K, Q Q J
Q, FISH T WILD K
TURTLE, J LILY T LILY
T, 9 T FISH 9
A, WILD A Q J
J, J J A K
TURTLE, Q LILY K T
T, BONUS Q TURTLE
T Q A
LILY T Q
J


Assume that we have the following win pattern: 5 Q, 4 Q, 3 Q, 5 T, 4 T, 3 T and no hits

First, I go through a permutation test (go through 10x8x10x7x11 combinations, and apply 1024 ways of payout to each combination) to count the total possible combinations for each pattern 5Q, 4Q, 3Q, 5T, 4T, 3T
Here is the permutation results I got
Pattern Observed Hits
5 Q: 36864
4 Q: 64512
3 Q: 0
5 T: 18432
4 T: 32256
3 T: 12672

Then I try to run a chi-squared test to make sure that the actual number of hits for each pattern is not significantly different from the expected hits.
What I do with Chi-Squared test is that I randomly picked 5 stop number for 5 reels, and apply 1024 ways of payout and recorded the hits for each pattern (5Q, 4Q, 3Q, etc ...)
The sample of my Chi-Squared test is 1,000,000 and I ran the Chi-Squared test 10 times, with 1,000,000 run each time, and I got 3-4 fails out of 10 times.
The degree of freedom I calculated is 6.

So I have couple questions:
1) It seems to me that I used the wrong variables (5 Q, 4 Q, 3 Q, etc ...), It seems like I need to use the payout of the pattern as the variables.
2) Or If I used the variables correctly, then my degree of freedom is wrong because If I add up the total hits of 5 Q, 4 Q, ..., no hits (194150) it's not equal to 10x8x10x7x11

Cheers,
Tony