At MGC I was talking to some folks about my custom games and was asked if there was a thread on pinside, so figure I should at least start one for the machine I am working on now. A lot of the content of the first few posts I will make are edited version of what was on the pinballcontrollers.com forum, so at first it will look like I am making rapid progress on this game, when in reality, not so much.

Lets start with some background. After moving my first custom game “the Kugler family” to a larger full color display (it started with regular DMD, then color the same resolution and then to large screen full 24 bit color), I decided to start a new project.  My brother kept pushing on doing a gambling theme. While at first I was lukewarm to the idea, after I came up with some cool ideas for toys, I became more interested and decided to give it a try.

The game was at expo in 2014, but believe it or not, not a single game was played on it before bringing it to expo. I had not put the playfield into the cabinet until about 3AM Thursday morning of expo. So late in the day on Thursday after I fixed some issues with it, it was played for the first time.  I was actually quite pleased with how it played as far as the number of shots, the feel of the shots, etc. but I was not happy with how the rules for the different casino games came together. After expo I was pretty burned out and did nothing on either game for months. Then I got back to work on adding artwork and some playfield changes (needed to do a new version of the playfield to get the inserts where I waned them to go with the art) for the Kugler’s and Casino sat idle and I did not even bring it to expo this past fall, since I had done nothing on it. But, a week or two ago, when I drove past the Westin, it struck me that expo is just six months away, and if Casino was going to make an appearance, it was time to get back to work on it.

The game is running with a resolution of 450x900  (my other machine is running 96x192), which is then doubled when pushed to the LCD.  So the LCD is running at HD resolution. The actual screen size that is visible is 13”x6.5” so it still fits into a standard height speaker panel and between standard size speakers.   The game does not use a dot look, since the dots become just way too small and really just seems to darken the screen.  The game is currently running on a low end pentium processor, which appears to have no issue providing the necessary loop rate, even with the high resolution.

The color support comes from using the pyprocgameHD version that MOcean and I have been working on for a while.  Our work on that evolved from the standard 16 shade dmd to an 8 bit, 256 indexed color approach and then to full 24bit color.  To be honest Michael did more of the heavy lifting than I did.

The game features the five most popular games you will find in a casino (slots, craps, poker, blackjack and roulette) and then also has a set of 8 modes to capture other things that often occur in a trip to Las Vegas (bachelor party, all you can eat buffet, drunk and can’t find my room, etc). The original rules had all five of the casino games active at the same time, well, sort of, but I now have some playfield changes and rule changes to reduce that so it is not so confusing to the player as to what is going on. So while you still may be building towards multiple games at a time, the play of those gambling games is more clear. I’ll do posts on each of the casino games to explain how they are being implemented.

I should comment, that I did a lot of work on this machine before I was aware of High Roller Casino despite knowing over a dozen guys with anywhere from 10 to 50+ machines, I had never seen or played one. From the couple of times played it at expo, there is not a lot of similarity, although both have a Roulette wheel, but as I told Keith Johnson, I think mine better (I’ll explain who it works in a later post).

Here are a couple of photos of the playfield and cabinet. The first is after I finished routing (by hand) and then the inserts and the cabinet with playfield at expo.
This is the playfield after being routed by hand using a 30 year old hand router, using templates for the square and triangle inserts and Forstner bits for the round ones.

IMG_0839_(resized).JPG

Here it is with the inserts added.

IMG_0845_(resized).JPG

and here is the machine at expo, funny thing is you can see the Kuglers machine, which was much, much further along in the background, without cabinet art (which it did have this past year at expo, this is from 2014), yet Casino has art . . . .

IMG_0881_(resized).JPG

I'll do some other posts shortly that will cover the features, rules and toys of a couple of the casino games.

Happy to answer any questions.

recently heard about that, but have not seen it yet, wondering if they are using rfid or some other approach vs machine vision. In my game not too concerned if occasionally it gets the split of the dice wrong, which can occur if they are very close tougher that would be a much bigger issue in vegas. They also could be doing pattern matching for dice identification vs clustering analysis. I looked at pattern matching, but would require a lot more work to implement and the results I am getting are certainly good enough.

detecting the dots themselves is not too hard it is the clustering that is the most challenging. After I capture the image, I convert to grey scale, and to black and white using a threshold that will clearly separate the white pips/contours and then using clustering on those contours to get to what is on each die, two things cause challenges, the pips on a six are so close together that the code that identifies the pips can lump them together and think they are a single pip, so I use some 'dilation' after thresholding to shrink the dots slightly to be sure there is spacing and I need to adjust that slightly to improve the six handling. The clustering can then occasionally cluster wrong and assign a pip to the wrong die. Recently changed the algorithm I was using for the clustering and it seems to be have a very high accuracy rate, as I indicated above.

no doubt opaque would be more legible, as far as not seeing pips from the other sides, and typically the pips are not as big as on vegas dice, but I wanted real vegas craps dice and once I got the lighting consistent, I've been able to get the needed accuracy. Did take time to tweak all of the parameters at each step of the process.

Here are some images of the results of each step in the process . . .

Original image taken by camera . .

original_(resized).png

converted to greyscale

grey_(resized).png

Thresholded and inverted

threshold_(resized).png

dilated to shrink dots and remove noise . . .

dilated_(resized).png

Original image with clustering results mapped on top of image for debugging.

result_(resized).png

I am working on a mini-playfield with five drop targets, with an LCD display mounted above it to show a card for each drop. I had looked at one small LCD per card/target, but the cost and complexity of that was greater and with the one display it will be easier to provide instructions, etc. It will be used for draw poker and black jack, and who knows what else. It will require me to do new ramps so it can fit where I want it, but I've got Matt lined up to help with that. I need MOcean to get me the code snippets he did for curbfeeler on pinbot2.0 for the support of a second display on the pyprocgameHD framework so I can start to play around with it, while I also build/modify the mech for the drop targets, which I want to have full up/down control but with a single solenoid to minimize the footprint.

Made some progress over the weekend on the roulette wheel. I have changed the way the optos are set up. Originally I had a horizontal set for the upper ring/magnet, and then a verical set to detect where the ball was and then a U shape opto under the 'spoked wheel' to detect the home position. The U shape opto had not been implemented due to challenges with getting that postioned, wired, etc. So decided to also make that a vertical pair. When I was looking at that change it struck me that I could also make the magnets optos vertical, which would allow me to make the bracket that holds them much smaller and less obtrusive.

IMG_1507-2.MOV.png_@_100%_(Layer_2,_RGB_8)_*-2_(resized).png

Now that I had the 'home position' opto implemented, I was able to finish up the number detection piece. The opto is looking down through the number cover, where there is just one opening which allows it to detect the home position. Once it finds that, it advances the wheel one section at time (36 or 37 pulses to the stepper) to determine where the ball is, it is timed such that it transparent that it is doing this a slot at a time since the wheel does not stop at each section. The timing is key so that a 'spoke' does not make it seem like a ball is in that slot.

Here is a video with the changes, and a test of the number detection. I've tested it about a half dozen times so far, and it has gotten the correct number each time. I've added chunks of code to handle various failure scenarios (ball falls into lower portion early, ball does not exit correctly, home position not found, ball slot not found, etc.). I'm sure there are a few others that will come to light when I do testing during normal game play.

No, I prefer to 'fit on site' for that, based on which type of post I end up using.

Heck, now it even has a finial.

I'm using pdb-led controller boards from Gerry (multimorphic) and then led boards that I designed that have a lot of flexibility to how you use them, which can also speed up wiring. I'll post more info on them when I start to install them.

I am not a huge fan of the serial type for inserts.

about 10 minutes into a 5 hour routing?

Yes I did it, with my limited self taught CAD skills. It is an updated version of the original cad done for the first white wood, which I used for the manual routing (printed to scale and spray adhesive to the wood). Big issue we ran into is the blocks for the inserts that someone else made and shared, were not the correct sizes and way off for the shape of the arrows (maybe Scott will post a photo of the wood from those test runs). Hopefully we now have them all correct (and properly updated in the CAD).

For the manual routing I had created templates for the routing that were basically perfect, I had found a great method for creating templates for different inserts by using the actual insert and popsicle sticks to get the spacing right for the bushing on the router, and then a flush bit to cut the template -- it worked great. In two cases I used some tape to tweak the sizing. So the reality is, it sets a pretty high bar relative to quality and fit for the CNC routed one.

I don't think you saw the manual one, it really was surprising good as far as fit and finish, the biggest drawback to the manual is there is less accuracy around positioning and alignment than with the CNC, since I have to position the template over the insert and hard to get that perfect. Other routes and cuts that were not the inserts, should be much better with the CNC, since I typically did not create templates when routing those and for various cuts, not like I am that perfect in cutting with a jigsaw. Also, a lot less stress relative to the routing, when doing it manually you know that any mistake of significance, could wipe our hours of hard work.

I see you grouped the craps targets to all be on the second half to be sure the spacing between them stays precise, glad you remembered that. So, looks like it is about half way there.

Just got back from vacation today, things will be picking up pace quite quickly. Picking up the routed playfield later today, first run had an issue, but sounds like the latest one is good to go.

you will find some info here . . .
http://www.pinballcontrollers.com/forum/index.php?topic=843.msg7492#msg7492

For creating the templates, I take the insert and then use 'popsicle' sticks as a spacer between that and strips of wood to create the outline. This gives just the right spacing that matches up with the size of the router bit and the collar.

IMG_1374_(resized).JPG

Then I take the 1/4" hardboard and use a flush bit to create the template. I first drill a hole in the template material, then center that over the insert shape I created the wood strips to create the shape. I clamp the template to the scrap wood with the wood strip shape.

IMG_1375_(resized).JPG

I believe, that for most of the templates I use a 1/4 inch bit with a 7/16 collar. I use 3/16 for slots for switches. Somewhere I have some notes on the different combinations.

IMG_1376_(resized).JPG

If you read the above thread, you will see that I drill out the inside of the insert before routing so there is less material to route out, but then use the route of the outside of the insert to act as the template for then removing the rest of the 'hole' and leaving the necessary lip. It may be the same 7/16 + 1/4 combo, but not sure, I'll see if I can find my notes, It may be a smaller difference between collar and bit.

IMG_1796_(resized).jpg

I've used tape on one or two to refine the template to give a nice tight fit.

Hope this helps, let me know what questions you have.

Picked up the playfield last night, here is a photo of before I started to removing the waste material (well a few of the inserts have been removed). I finished that up last night, so will now start to get the inserts in. I am missing the small white opaque arrow insert, and pinballlife is out of them, so I'll either be switching that to a larger arrow or making the insert myself. Also have to make the insert for the 'vegas sign'.

IMG_1798_(resized).JPG