This machine has been built almost entirely from the construction toy K'Nex. The only non-K'Nex components are the column labels and the signs. None of the K'Nex pieces has been modified in any way.
This Instructable covers various issues concerning the design and build rather than being a step-by-step guide on how to build it. If anyone wants to build something similar, or wants further information, further details can be given as required.
The machine was designed and built over a seven-week period, the time spent being around 200 hours altogether - this includes the insertion of thousands of balls for trials.
Type of Machine
It was decided to make a 'slot' machine which worked with K'Nex balls which would bounce down a grid and end in one of many columns at the bottom, some of which would pay out balls as a win.
Style of Machine
The machine's height would be dictated by the height of the grid, the height needed for winning balls to be released, and how high off the floor the pay-out tray should be for it to be easily accessible. There was to be a ball slot at the top of the machine, but for the vertically challenged there would be a motorised ball lift which would raise the balls to top - this would use the traditional method of a 12v motor and a chain.
The number of columns needed to be high enough to make the game interesting, but not so high that the left- and right-hand ones were virtually never visited.
It was felt that there needed to be some kind of jackpot feature. The size of the jackpot couldn't be too high because of the difficulty of handling so many balls. It was felt that a 1/100 to 1/50 chance of a jackpot which paid 10 balls or so would be about the right balance - this would correspond to 10% to 20% of the overall return (actually a bit less, because sometimes the reserve would not be full).
For every 100 balls inserted it was intended that, on average, around 80 would be returned, and that the chance of a win would be around 25%.
It was decided to use as much white, yellow, orange and red as possible, with black 16mm rods used instead of green ones where they were visible (but they aren't really 16mm, they're 17½mm, aren't they?).
The starting point was to construct a grid with about a dozen exit points (columns) at the bottom so that the distribution of where the balls landed could be produced. This would help decide the final number of columns.
The diameter of K'Nex balls is 46mm. The grid was to consist of two arrays of white connectors (joined by white rods) separated by transverse yellow rods on which the balls would bounce. This way, the closest distance between two yellow rods was 53mm - a wide enough gap for a ball to pass through. The ninth image shows the construction.
The first grid had 12 columns and a thousand balls were dropped into the top of it. The distribution was surprising - it was, as expected, a bell-shaped curve, but it was rather flat. This was good news, because it meant that a good proportion of the balls landed in or near the end columns. In fact, so many landed in the end columns that it was decided to increase the grid to 16 columns.
Things were getting interesting now: more balls than expected were heading towards the left-hand column, and also towards the right-hand one. Generally, it was as though once a ball had started falling it wanted to continue in a straight line rather than continually change direction. There was no reason to believe that a ball would have a 50-50 chance of going in each direction after it had hit a rod (that would depending of the relationship between the yellow rod spacings and the diameter of the ball), but even so the behaviour was strange.
Moreover, at times a ball would suddenly shoot in a straight line at speed for no apparent reason; at other times the ball appeared to have a think when it hit each rod as though it was deciding what to do.
A computer simulation was run which assumed that, when the ball was going in a certain direction, it would be slightly more likely to continue in that direction than change, but whatever the chance specified, that flat bell-shaped curve could not be reproduced.
Exactly what causes this behaviour has not yet been discovered (any ideas are welcome), but the good news is that the first and last columns were visited just enough to make the game interesting.
It was very important to ensure that there was no bias when a ball entered the grid, and so the ball enters either from behind (if the ball lift was used) or from the front (if inserted manually) rather than from one side.
In the original prototype, black rotational connectors were fixed to the internal ends of the transverse yellow rods so that a ball would roll off these instead of occasionally ending up lying on top of a rod. This worked fine most of the time, but about one ball in a thousand ended up with one of its holes stuck on one of these connectors. This was fixed by using tan clips instead, there being no bit which could jam in a hole. The 11th image shows some of these in place.
An indication of the column a ball entered was required so that, if the player wasn't looking at the time, it would be possible to see what column the ball landed in. This was achieved by a carefully-balanced lever which swings when a ball enters the column. The column label (which shows the amount of a win) was actually produced using a badge-making machine!
The Win Sizes
It was decided that Column 1 - on the far left - would trigger the large win (it ended up as 13 balls).
Since Column 1 was paying a high win, it was decided that Column 16 would too, but only 5 balls otherwise the number of small and large wins would be out of balance.
Using the sampled distributions, it was decided that Columns 2 and 15 would pay three balls, and Columns 3, 4, 13 and 14 would pay two balls.
Further trials took place (3,000 balls were inserted), but the average return was only around 67%. Including extra winning columns would make the return too high, and there was something pleasing about the wins of 2 being a consolation prize, and the wins of three more so - the sequence of 2 2 3 5 felt much better than 2 3 3 5 or 2 2 5 5.
The final pay-out averages 84%. This was achieved by simply adding two extra yellow rods near the top of the grid which added about 8% (see the ninth image), and one rod near the bottom right which added another 9% or so (see the tenth image).
The chance of a win for this final version of the grid is about 0.28, resulting in a generally steady flow of wins, not many long losing streaks, and the occasional cascade of 13 balls into the pay-out tray. It is quite possible to insert 25 balls and get 30 or 40 back, but quite possible too to only get 10 or so. The balance is about right for an amusement machine.
The Pay-out Reserves
There are three main pay-out reserves - one for a two-ball win, one for three balls, and one for five. The number of balls paid out from the reserve is one less than the win, the winning ball being returned.
There is also a Bonanza reserve which is emptied when a ball enters Column 1, and again the winning ball is returned. On the rare occasion that two Bonanza wins happen one after the other, only one ball (the winning ball) will be paid out the second time.
The reserves are topped up by losing balls. Since wins of 2 are the most common, that reserve gets topped up first. When that is full, the ball attempts to top up the 5-win reserve, and if that is full, the 3-win one. If all three reserves are full, the ball tops up the Bonanza reserve, and if that too is full, the losing ball experiences a staggered fall into the 'cash-box' underneath the machine (the fall is staggered so that the ball doesn't split into its component halves through shock).
The Pay-out Mechanism
This needed some thought: how could a ball weighing only 24g be used to release the winning balls? The solution was to minimise the force required for the balls' release by lining them up on a gentle slope and letting the required number of balls roll into the pay-out tray.
It was decided that the winning ball would form part of the win so that the pay-out reserves were less likely to run out, and wins of 1, 2, 4 and 12 were therefore required.
At first, the balls rolled down white 8-way connectors, but every now and then a ball would stay on a white connector's hole instead of rolling away from it, and so lengthwise orange connectors were used instead. Many more pieces were required, and the build took longer, but the balls flowed much more smoothly. This had its own problem though: when orange connectors are connected with green rods, there is a slight bulge where the join is. If a ball was sitting on a slope just before a bulge, it sometimes wouldn't roll when the win was released. The solution was simple: the slope was eight connectors wide (i.e. 48mm, just enough for the balls) and the middle two connectors - the ones the ball was resting on - were not joined by green rods at that point (but they were at their other end), removing the bulge. Very few connectors needed this treatment.
The two- and three-win reserves hold a maximum of 18 balls, and the 5-win reserve, 19. The Bonanza reserve holds up to 12 balls. The machine can therefore retain up to 67 balls.
The pay-out mechanism can be seen working for a three-ball win at 4 minutes 44 seconds of the video - if you have a look at it now the following notes will make more sense.
For a two-ball win, only one ball is released, the winning ball being returned. There are two blue rods above the point between the first two balls in the slope. There is a release lever which prevents the balls from running from the slope into the pay-out tray. The winning ball falls onto a triggering lever which a) lowers the two blue rods so that the second and higher balls are prevented from rolling down the slope, and then b) raises the release lever so that the first ball can fall into the pay-out tray. All the levers have been carefully balanced using different type of wheels as weights. No rubber bands have been used anywhere.
A three-ball win works similarly, the blue rods falling between the second and third balls.
For a five-ball win, the blue rods fall between the fourth and fifth balls, but there was a problem: the slope is gentle so that the releasing lever has little friction when raised (the leading ball is resting against it), and when the winning ball rolled off the end of the triggering lever not all of the won balls had time to roll out. The solution was to make the triggering lever longer so that the winning ball had to travel further before it fell off the end of the lever en route to the pay-out tray, thus lengthening the time that the release lever was raised.
The triggering levers won't cope with one win immediately followed by another identical one, and so the ball lift releases balls at 2¼-second intervals - just enough time for the pay-out mechanism to handle close wins.
The Bonanza Pay-out
This uses a different method. The balls still queue on a gentle slope, but all the balls - up to 12 of them - have to be released in one go. There is a release lever, but it needed to be prevented from releasing just a few balls and then blocking the rest after it came back down after it had been raised by the winning ball.
A couple of green connectors on the release lever are used to block the flow of the balls, and when they are lifted the flow of balls hits the (raised) angled part of the connectors thus keeping the lever raised! The slope is actually slightly steeper than those for the other pay-outs so that there is enough momentum there. Have a look very closely at the video at around 5 minutes 2 seconds.
The Base of the Machine
It was obvious that the machine would be quite heavy and that it would need to be based on a rigid frame.
The base was constructed mainly from red-rod-sided cubes with the grey diagonal rods on opposite faces at right angles to each other - see Making Strong K'Nex Structures for the construction method.
The grid, being 16 columns wide, fits into 8 red-rod lengths, making its attachment straightforward.
... will be supplied on request - just ask!