I love the visuals to show the functional probability. The "nested bags" was a very good implementation, I enjoyed the story. I come from a pure mathematics background, and probably would have done some weighted matrix, but the "how it works" isn't as clean as your visual. I enjoy the peaks behind the curtain that this series, and LD offer, to a Dev process that doesn't get a lot of coverage.

I'd love more game design videos like this! Your explanation and visuals made this concept so easy to understand! I can't wait to try it!

Great video and channel! In my work I prefer to work with floating probability: after each iteration, I increase the weights of not-chosen elements by a certain value, I decrease the weight of chosen elements by another value. The values for increasing and decreasing are set separately for each item in the set. Thus we get a self-balancing system, in which we can not only guarantee that the player will get M results in N trials, but we can also manage the number of identical results in a row.

here from the chatgpt video, i rly enjoy your clean style! it surprised me that NES tetris wasn’t brought up in this video, as its 100% random piece drops allow extreme behaviors to happen. this caused later tetris games to use a 7-bag, in the same way you described in this vid. this leads to competitive NES tetris having a nice risk/reward balance, as players need to play somewhat defensively against bad luck.

Informative video and really well explained, thanks :)

It sounds similar to the way old (mechanical) slot machines worked in a sense that you can control the chance using "par sheets", but on the other side it differs in the fact that distribution in not changing with each drawn marble, of course. In case of slots machine clearly each following attempt is decoupled from the previous ones. Thank you for making these videos!

I've got an idle crafting game called Craftsmith and I think this system would be helpful for my quest system. It could really help control the pool of quests, especially the rare ones.

WOOOOOW, the bag analogy is one of the most brilliant things i ever seen in my life!!! Such a GREAT way to express how probability actually work!!! Those who complain about 3 misses in a 97% chance would finally understand.

I enjoyed this topic a lot. I love RPGs and randomness is usually a big part of that. In the game I'm planning out, I feel this could be used for combat, loot, and power up distribution, and randomized event outcomes.

One idea that came to mind for reducing memory usage in the case of a large number of marbles but a limited number of outcomes is storing the number remaining of each “type” of marble, instead of creating a new entry for each marble. So for 3 in 100 chance of finding a rare item, store two numbers: the rare item marble count = 3, and the normal item marble count = 97. Then to pick one, choose a random number in the range 1-(total marbles remaining), if the random number is the rare item count, give a normal item and decrement the number of remaining normal items by 1. Memory use would then scale according to the number of outcome types rather than number of outcomes. A downside to this is that the memory required for a bool is in theory 1 bit, while the counters will require more than 1 bit, so memory use may be higher until a larger number of total marbles is reached.

I've done randomess the normal way for awhile math.random(1, 99) etc But I've never done it this way, and it seems perfect for your word game

have mixed feelings about this because some of the greatest moments in video games happened because of not controlled randomness for example if a person is lucky enough to land 5 low chance crits in a row, i dont want to take this joy and his amazing luck away from him other than that the marble bag is an amazing n useful tool

Where’s the link to Piggy Pileup? The only links in the description are to your Discord, Twitch, and Twitter pages.

What about a dynamic approach to solving too many items in a row - or never getting one of them? Let's say you had to create a decision between two items to award. You set a percentage chance. Let's say at default white is 80% and blue is 20%. You win and blue is awarded. Now we exchange 5% (or whatever grain you wanted). White is now 85% and blue is 15%. Leaves us a chance that blue could be awarded again, but weights it heavier to white. It would be interesting to run this through millions of iterations and see what the actual chance for white or blue became. Ever since I programmed a Commodore 64 and 128, I always thought that was the proper way to create 'fair randomness' ala Marble Bag. But I went web code and SQL instead of gaming. Cheers, thanks for the vid