A new topic to attempt to centralize information, improvements and discussions about the most iconic Great Ball Contraption made by Kawaguchi Akiyuki,
The Ball Factory GBC Version 0, July 2008 by Superbird28. Akiyuki stated himself he was inspired by Superbird28 when designing and building the Ball Factory. Why can be clearly seen in this video dating back to the pre-digital age. Thanks @Jonas for noticing. Version 1, June 2011: with color sorter as input:
Version 2, november 2011: picker for 2 balls and optimized for higher speeds (1.3 balls/s)   Februari 2016, @Blakbird and @TheRebricker reverse engineered the Ball Factory over a period of two years and posted the below: Video instructions by The Rebricker PDF instructions by Blakbird 4428 parts MOC Review by Blakbird LEGO Digital Designer file by @TheMagician The Akiyuki Project topic by Blakbird April 2019: Ball Factory New Style GBC compliant, reliable and easy to move:   A big thank you to Blakbird and TheRebricker for all their hard work! This reverse engineered version adds the option to circle balls within the ball factory, here my build from December 2016
Many have build the Ball factory in various colors and versions. Below an overview of suggested and quoted improvements with pictures out of the 77 pages in The Akiyuki Project topic, up to you what changes you want to make. Feel free to add your builds, improvements or questions about The Ball Factory to this topic.
___________________________________________________ Spiral Lift platform improvement by Jonas: used a SNOT technique to get a stronger platform. It allowed me to remove the gap and to fix the superstructure to the studs. Moreover, I used the remaining stud of the 30414 brick to stabilize the tower column by a vertical 1x6 tile. Now, each column is fixed from 3 sides and the tower is really stable. It will survive even an accidental kick.  And, I added another (third) black pin to strengthen the vertical support (shown on page 186 in Blakbird's manual)
___________________________________________________ Ball picker improvement by Jonas:  I have also made a small improvement of the ball picker - the part that picks a pair of balls and puts them into a bucket. It happened quite often that the right ball dropped before it reached the bucket. When I analyzed the construction of the picker, I noticed that the rightmost axle is significantly looser than the other ones. While the other two rear axles are attached to the rail construction, this one is not and cannot be attached. In my version, I fastened this axle to the rest of the assembly by replacing a 3L liftarm by an L-shape liftarm (not necessarily in yellow color).  ->  Since that time I have not observed any earlier drop.
___________________________________________________ Back pivot point by @Ankoku I have used the 5.5L Axel with Stop just like the original. You could probably get away with a 5L, but the connection may be hanging off slightly. The main piston rod has been changed from 20L to 19L in complete length. I used the "Technic, Plate 1 x 5 with Smooth Ends, 4 Studs and Center Axle Hole" which Akiyuki used, just because I did have them kicking around spare. A 6L Plate would be fine. I have gone with a 11L and 7L lift arm. Akiyuki went for a 15L and 3L, thus far, I haven't seen any reason for using those particular lengths. One bonus with using the plate to connect the two lift arms, is that the connection is rock solid, so you no longer have the wobble of the altered version. As you can see on the right, the connect on the end of the 8L axel is flipped. In the ReBricker instruction video, he has it this way round and then adds a later correction to show it the other way around. This way round is the way it is in the Akiyuki video and it is required to make this pivot point work.
___________________________________________________ Bucket shift timing mechanism by Ankoku Here is the piston itself. I have changed it to the original format. This is obviously the left one of the two in the video, as you can tell by the connector end, but other than that, the two pistons are identical. The crucial change is Technic Beam 1x3 Thin between the two year 1x5 Technic Beam Thick. The reverse engineered solution doesn't have that. That provided the extra travel I was looking for. As you can see, the connections for the yellow beam on the right aren't as pretty and you lose the symmetry. That said, the new motion is as smooth as butter, which is immensely pleasing in comparison to trying to make the other method work.
___________________________________________________ Bob improvements by Ankoku The initial changes to the height didn't do much to improve smoothness and I found that the paddle seemed to now be an issue. It was also the source of quite a bit of friction. As you can see in the image, the paddle is 3 deep in total. You can tell this from the light grey 3L axel. The bottom part of the paddle is a Technic Beam 3 x 3 L-Shape Thin. Using that piece makes a lot of sense and seems far better than using 3x Technic Beam 1 x 3 Thin, where the 1st of those completes the light grey 3L axel length and the other two are at 90° to create the L shape. Unsurprisingly, this approach is smoother as there is less friction.
___________________________________________________ Bucket unloader by Ankoku So I was playing with it and it didn't seem to work very well and certainly didn't reflect the motion I was seeing in the Akiyuki video. Having watched the video more times than I care to remember, I noticed that the lift motion seemed to be in a single motion, not two. I also noticed that the lift arm used seemed shorter than the one in the instructions. So in the end, I changed it from a Technic Beam 1 x 5 Thin with Axle Holes on Ends, to a Technic Beam 1 x 4 Thin with Axle Holes on Ends, using yellow ones to reflect those in the video. This seemed to work well and reflect the video. Then I checked on Bricklink and it seems that Technic Beam 1 x 5 Thin with Axle Holes on Ends don't come in Yellow. So yeah, that bit needs to be a 4L not a 5L.
___________________________________________________ Back slider by Ankoku, There is definitely these two beams there, length unknown. Below I am using 2x Technic Brick 1 x 15. You can see both in the Akiyuki video, although only 1 at any one time. This also explains why the bottom Technic Axle and Pin Connector Perpendicular is attached differently to the two above it. For the forward back motion you see in the video, this is not possible with the use of a pin instead of an axle. The axle you can see here, allows full motion in one direction, but slightly limited in the other, which is exactly what you see in the video. It is only possible to put the top Technic Brick 1 x 15 in there if you remove the plates from the bottom of the slider. If you look in the Akiyuki video, those plates are not there. I still don't know how either piston connects to the slider exactly. I am still hoping to hear back on that. That said, these two beams seem to be correct and answers a few questions. ___________________________________________________ Bucket slide by Ankoku The area that the buckets slide along for the shifter is actually offset by half a stud in the Akiyuki version. This is not reflected in the instructions. This answers why the shifter itself is half a stud over the sliding area at maximum backwards reach. I always thought the shifter didn't do full travel, but it does, it is just the sliding area which is misplaced. Here you can see how the sliding shelf is connected to the conveyor with a Technic Brick 1 x 2 [2 Holes]  Anyway, changed the length of the slider my one, so that is now correct. Removed the plate from the bucket stopper just before the slider, so that is now working correctly and is the same as the video. Bob now actually aligns correctly with the buckets at maximum reach. ___________________________________________________ Shifter changes by Ankoku This removes the need for any universal joints and thus removes the variable speed of the conveyor. ___________________________________________________ Shifter changes by Ankoku It seems Akiyuki has reworked the pivot mechanism and beam length. This new approach removes any potential bend in what would have been an 8L axle. The new approach requires a shorter piston beam, a longer beam under the shifter. The brace is 2x Technic Beam 1 x 6 Thin.  Here, you can see the positioning of the pivot point, the indentation in the conveyor tower, which is 2 bricks deep to allow the pivot arm to travel that far. You can also see the black axle instead of the grey one, as it requires that extra 1L for the back and forth arm to pivot one. At the top, you can see where the slider platform connects and how it is moved 0.5L across. Shifter Changes in action: ___________________________________________________   Ball dumper by Ankoku It sits at the right height, the ball router works well and the axle housing is aligned correctly. As you can see above and below, I have added blockers to either side, which means that a renegade ball can't escape up hill. Balls can only pass when the router is in the correct position for that route. Inside is only 2 wide now and doesn't have any pins etc. ___________________________________________________ Ball unloader by @smdzucladoc After seeing Ankoku's mods as well as his detail photos of the bucket shifter, My ball factory has been much improved.  I was still having a lot of friction in the ball loader until I add these mods.  Now the ball loaders is very smooth. I add a 1x3 liftarm in between the two Axle and Pin Connector Perpendicular Triple and it seems that the loader doesn't bind to the axles anymore and it is very smooth.   I also added some additional bracing to the back of the ball loader. I added two additional Technic, Brick 1 x 2 with Hole and a 1/2 bush to the bottom axle.  I also change the axle 3L to 4L and added 2 Axle and Pin Connector Angled #1 and then a 6L axle perpendicular to try to brace the 12T and 40T gear as perpendicular as possible.       ___________________________________________________ For inspiration, this extended version by @Jonas and black version by @Thunderthumbs