Cracking the slot reels - Here you go!

[KasinoKing]

Originally Posted by Rusty

What a fantastic thread, how did I miss it?
This maybe the most important thread ever posted here...

Absolutely agree (with all of your post - no need to quote the whole thing)!

The stuff these guys are producing is incredibly interesting.
If I only had half a clue what all those figures meant - it might even be useful!

Only kidding guys! Excellent work - keep it up!

KK

[gerilege]

Originally Posted by kimss

The reelpositions are the accual position on the reels, and for all three symbols not the middle symbol. A zero (0) would mean that the reel is in its first possition showing the three first symbols from the top. First number is reel 1 from top to bottom, next is second reel and lastly the third number is the final reel.

I also have the data for the symbols on each reel, however this data is meaningsless and why I dropped it in the text file. I can regenerate the text files with more data if someone would have interest in this. I can also set a flash casino to do 10.000 spins, however the data from my spins clearly shows that the reels are weighted.

I hope I got it right.

Reel2 is:
Pink,Fries,Red,Wheel,Pie,Coffee,Wild,Yellow,Hambur ger,Bottle,Dices,Scatter

Weights are:
0.0235,0.1819,0.033,0.0255,0.0465,0.0245,0.059,0.0 54,0.2219,0.2244,0.056,0.0495

I'm almost sure that this is for the middle symbol. Which means that 0.0495 means that the scatter is the middle symbol on reel1 with 0.0495 probability. Because if each number would mean the top symbol, it would mean that the scatter is visible on second reel with 0.2244+0.0056+0.0495 probability, which is too high I guess. 0.056 + 0.0495 + 0.0234 seems more reasonable for the real payout.

[kimss]

You are correct, I see it now. Since I have been working on the % of reel positions I havn't really needed to check this (I am not calculating % payout). All the numbers (reel positions and symbol layout) are served from the MG server, so it has to be correct. This in turn means:

Reel 2 has the following symbols from top to bottom:
reel symbol 0: Pink lady with car
reel symbol 1: French fries
reel symbol 2: Red man with truck
reel symbol 3: Burning wheel with start flag
reel symbol 4: Pie slice
reel symbol 5: Cup of coffe
reel symbol 6: Wild
reel symbol 7: Yellow man with car
reel symbol 8: Hamburger
reel symbol 9: bottle and cup
reel symbol 10: Dices
reel symbol 11: Scatter

Since the highest reel position on reel 2 is 10, this means that the reel would display the symbols from position: 9,10,11. This confirms that the position is for payline 1 (the middle symbol) and not for payline 3 which I assumed.

[Zoozie]

Originally Posted by gerilege

I guess I have it now. After experiencing problems with the scatter payout I realized that if the reels are weighted, winlines DO matter. You can't simplify the analysis to one winline any more.

Sorry, I should have told you this after our PM mails. But it raises an interesting question. Payout now depend on number of lines play if same weights are used. (for scatter win this does not matter though, only for line wins)

Can you make a payout% for 1,2,...,9 lines played?

But I ran your program and it say 101% payout for 1 line played. Since I do not believe this really, could it mean that other weights are used when playing 1 line? Or generally 9 different weights for depending on you play 1 or 9 lines.

MG really made it hard for themself by creating their rigged video slots.

So Kimms, I have a small wish. Maybe you could make another sample of "5 reel drive" with only 1 line played? We can then see if the weights are different.

Gerilege: You programming in java now, or are you still using a C++ to Java translater?

[gerilege]

Originally Posted by Zoozie

Sorry, I should have told you this after our PM mails. But it raises an interesting question. Payout now depend on number of lines play if same weights are used. (for scatter win this does not matter though, only for line wins)

Can you make a payout% for 1,2,...,9 lines played?

But I ran your program and it say 101% payout for 1 line played. Since I do not believe this really, could it mean that other weights are used when playing 1 line? Or generally 9 different weights for depending on you play 1 or 9 lines.

MG really made it hard for themself by creating their rigged video slots.

So Kimms, I have a small wish. Maybe you could make another sample of "5 reel drive" with only 1 line played? We can then see if the weights are different.

Gerilege: You programming in java now, or are you still using a C++ to Java translater?

I'm programming now in JAVA. I believe that the one line result calculationis incorrect in analyze. I will check it in a few minutes with analyze1.

[kimss]

Originally Posted by Zoozie

MG really made it hard for themself by creating their rigged video slots.

Assuming fair play that is... If you were to end the player at 95% overall payout the weighting becomes far simpler... there goes mr. conspiracy again,

Originally Posted by Zoozie

So Kimms, I have a small wish. Maybe you could make another sample of "5 reel drive" with only 1 line played? We can then see if the weights are different.

I'll do 1, 2, 3, 4, 5, 6, 7, 8 and 9 lines for 2500 spins.

[gerilege]

Originally Posted by gerilege

I'm programming now in JAVA. I believe that the one line result calculationis incorrect in analyze. I will check it in a few minutes with analyze1.

Analyze calculates the payout for line2, not line1. But you can only choose line1 first in the real game, and line2 only after. So it's possible that line2 alone would produce more than 100%. I'm checking this right now.

[kimss]

Originally Posted by gerilege

Analyze calculates the payout for line2, not line1. But you can only choose line1 first in the real game, and line2 only after. So it's possible that line2 alone would produce more than 100%. I'm checking this right now.

I could expand my data file, since the data served from MG also tells what line the win is on, and what the win was. This means I can give you excat figures on what line pays what I did this with the data for Thunderstruck, which you can see in the datafile (however i dropped the win sum, only added the win lines).

[kimss]

Here are the reelstats for 2500 spins on 1 line and 2 lines bet.

[gerilege]

Originally Posted by kimss

Here are the reelstats for 2500 spins on 1 line and 2 lines bet.

I don't think that different weightings exist. I think that the first sample was simply a good session, probably good line wins on winline1. The second sample was a bad session, but the first one was much better than average, the second was only slightly below average (I assume this because of the combined results, see below).

I made some corrections, and reran the extrapolated simulation with
-Original first sample only, 2000 spins (weightset 1)
-Second sample only, first recent sample of 2500 spins (weightset 2)
-First and second sample, 2000+2500 spins combined. (weightset 3)

Please find attached the results. Highlights:

Number of winlines: 1
Weightset used: 1
Payout ratio: 1.0137243209419724

Number of winlines: 4
Weightset used: 1
Payout ratio: 0.9151179228378415

Number of winlines: 5
Weightset used: 1
Payout ratio: 0.9513881379078132

Number of winlines: 9
Weightset used: 1
Payout ratio: 1.002396223151912
__________________________
Number of winlines: 1
Weightset used: 2
Payout ratio: 0.9118255004155771

Number of winlines: 4
Weightset used: 2
Payout ratio: 0.8568190079218757

Number of winlines: 5
Weightset used: 2
Payout ratio: 0.8894922862113461

Number of winlines: 9
Weightset used: 2
Payout ratio: 0.9482225993767085
____________________________

Number of winlines: 1
Weightset used: 3
Payout ratio: 0.9626296600537657

Number of winlines: 4
Weightset used: 3
Payout ratio: 0.885674539193263

Number of winlines: 5
Weightset used: 3
Payout ratio: 0.9204153397040318

Number of winlines: 9
Weightset used: 3
Payout ratio: 0.9750664675151187

I am yet to add the third sample (recent, second 2500 spins), we will see the combined results. I have also created a weightset 4 (First and second sample combined, but different sample size were considered,2000 spins for 1 and 2500 spins for 2), but whis is only slightly different from weightset 3, however I added it to the file.

I think that the payouts will be close in case of different number of lines played, but the maximum will probably be at 9 lines, the minimum will be at 4 lines. The current order is (based on the combined results):

9 lines, 8 lines, 1 lines, 7 lines, 2 lines, 6 lines, 3 lines, 5 lines, 4 lines. But this might very well change with the sample size.