Team split is not mentioned in contest rules or FAQ, but I assume:

1. A member can leave the team at any time, and he doesn't need approval from other team members or Kaggle to leave the team (although other team members may beat him up with machine-learning books for quitting).

2. When a team splits, the new teams carry on with the submission count of the old team.

Correct ?

-------------------------------------------------------

Now there're official meanings of 'member' and 'team' according to the contest rules, and there're common sense understanding of these words.

For example, the 'one team per member' rule and the 'maximum 8 members per team' rule refer to the official meanings of member and team: a member is someone who registered, and a team consists of 2 or more registered members.

So let's say there's a group of 20 people working as a team (common sense meaning), and only one of them is a registered member, then no contest rule is broken, correct ?

What if 2 people out of the 20 registered as 2 separate teams ? Are they considered to be in violation of the '1 team per member' rule ?

And how far can 2 official members/teams collaborate before they're considered to be really just one team ? Sharing ideas is encouraged, but if I give my source code or prediction files to another team, am I considered a member of that team ?

Scenario: a 3-person team decides they need more submission slots to win, so one member split off into another team. His former team has full access to his source code and submission files, but he makes no claim to being a member of his old team. That is, he gives up his official position in a potential winning team for the benefit of the team. Is this kind of altruistic behavior allowed in this contest ?

And what if 2 teams decide to share algorithms, source code, and submission data without merging, is this OK ?

These are excellent questions. The rule of thumb is as follows: if a group of people are sharing code and algorithms at a detailed level, and as a result are getting very similar answers, they should form (or remain part of) a team.

They is a grey area here of course, so in practice we have to use judgement on a case by case basis. Our recommendation is that if you want to avoid problems, you should form a team with any people that you are working closely with.