## So i have this weight problem…

At least, that’s how I call it. Let me go into details.

I have 5 jars, and each jar has a unique flavor that I want to taste. The jars are labeled from A to E.

What i’m doing at the moment: randomly picking a jar and displaying it. Only that randomly picking the jar is… random… and that means that I have a pretty big chance of getting the same jar two or three times while not getting another jar that may have not been opened.

Now, the jars have different flavors inside. If I like the taste, I should have a bigger chance of getting that jar, while if I didn’t like the taste, i should have a really small chance of getting it again.

## How can I do this?

I was thinking about a weight problem.

Let’s say that each jar has a ‘weight’ of 1000mg. If I have 5 jars, the weight of them will be 5000mg.

Jar | Jar weight | Total weight | Probability |

A | 1000mg | 1000mg | 20% |

B | 1000mg | 2000mg | 20% |

C | 1000mg | 3000mg | 20% |

D | 1000mg | 4000mg | 20% |

E | 1000mg | 5000mg | 20% |

Now I can do a random of 1000*N (N = number of jars). If i’ll get number 3302, I’ll taste jar D. If I get 4232 I’ll taste jar E and so on.

Now, i was thinking, if I pick jar D and I like it thus I want to have a bigger chance of getting it, I should increase it’s weight. To remain in the same total sum, I should substract from the other jars weight. Let’s say I’ll add 200mg if I like it and subtract 400mg if I don’t.

I pick jar D, i like it, I increase it’s weight to 1200. Those 200mg have to be taken from the other 4 jars. Thus i’ll have:

Jar | Jar weight | Total weight | Probability |

A | 950mg | 950mg | 19% |

B | 950mg | 1900mg | 19% |

C | 950mg | 2850mg | 19% |

D | 1200mg | 4050mg | 24% |

E | 950mg | 5000mg | 19% |

Now, because jar D occupies a larger space in the weight system, it has a bigger chance of being picked.

I take jar E, i don’t like it, so i substract from it 400mg and give to the others.

Jar | Jar weight | Total weight | Probability |

A | 1050mg | 1050mg | 21% |

B | 1050mg | 2100mg | 21% |

C | 1050mg | 3150mg | 21% |

D | 1300mg | 4350mg | 26% |

E | 550mg | 5000mg | 11% |

Let’s say, I took another taste of D, but now i don’t like it so I adjust the weight

Jar | Jar weight | Total weight | Probability |

A | 1150mg | 1150mg | 23% |

B | 1150mg | 2200mg | 23% |

C | 1150mg | 3450mg | 23% |

D | 900mg | 4350mg | 18% |

E | 650mg | 5000mg | 13% |

As you can see, the algorithm is very punishing if I don’t like the flavor and advantages the chance of me picking something that I either like or that I haven’t tried yet.

## Input

Do you have anything to add? Would you see another algorithm that can work better and achieve the same results?

By: Adrian Coman