Hi Jason,
Thanks for that, it's really important to keep this kind of feedback coming in.
I touched on Bulk Upload + Multiples in the Cuesby thread, but just to reiterate I think mandating a particular file format that people can use to upload cubes containing multiples could work. Although it does slightly defeat the point of "change nothing" bulk uploading. I'm unsure how I could reasonably generalise and discover card quantities without mandating a format. Using a CSV you might have:
4, Birds of Paradise,
4 x Birds of Paradise,
Birds of Paradise, 4,
4 Birds of Paradise,
2, Birds of Paradise, Revised,
2, Birds of Paradise, Beta
Split Cards are a real pain generally. Wizards really broke the mold when all of a sudden one card could be two cards. It makes it difficult to build an application based on rules when there are so many exceptions! At the moment in order to get Fire // Ice into the site you have to specify Fire // Ice (Fire). This isn't ideal, and I will be removing the bit in brackets soon.
Autocorrect would be awesome. I could definitely put some work into this to make the process of bulk uploading easier.
How would you like to mentally organise the cube? In a similar fashion to the View Cube page, or something else?
The short version of the bot picking algorithm is that they use observed pick orders. The long version is as follows. I would describe it as a close approximation to real drafting rather than a full on simulation, but I want to be as open about stuff like this as possible in order to improve it.
1. The bots are assigned colours at the beginning of the draft. Each colour is assigned to a maximum of three bots. The bots force these colours.
2. Each bot receives a pack and orders it by pick percentage. For example if Sol Ring has been picked by a human 9 times and passed once it has a pick percentage of 90%.
3. With the whole pack ordered by pick percentage the bot examines them in order. If the card is colourless it will take it a percentage of the time equal to the pick percentage, i.e. a Black Lotus with a pick percentage of 100% will never get passed*. If the card is coloured and in the bots colours it will take it regardless of the percentage.
4. If the bot reaches the end of the pack without having picked a card, it will hate draft the card with the highest pick percentage.
5. The bot shuffles the pack in case it is passing to a human and passes the pack.
*a note on this point. When a card has been picked less than 5 times, one Pass is added to the percentage to ensure that the card does get passed by the bots. Without this as soon as a colourless card is picked once without having been passed, it would never be passed again by a bot. This would artificially affect the pick orders as human drafters would see the card a lot less. It just means in the first few drafts that you will see colourless cards slightly more regularly than the pick percentage dictates.
I will revisit this algorithm at some point and look to remove the forcing and also try to find a sensible way for bots to make associations between cards. 'I have Entomb so I should take Exhume over Dark Confidant'.