Completely free experimental branch

[Tantali]

Currently the experimental branch is completely missing its purpose. It's intended for very broad searching. So for changes which might not necessarily make sense, and/or which don't have a real consensus.

master branch is 'better safe than sorry' experimental branch is 'lemme try this'

So in order to give experimental branch its real purpose, and to keep everyone motivated to come up with ideas, instead of hitting a demotivating no around every corner, I suggest these rules:

• master branch: consensus required, meaning 'convince the others it is good'
• experimental branch: automatic approval, meaning 'convince the creator it is bad'

[nymanjens]

I guess you could see this as a trade-off between frustration vs. more time spent in testplay (or slower progress) and I prefer less frustration. Making experimental free won't solve everything because I expect only a minority of discussion points will be resolved after testplay, the other ones may be slightly easier to resolve with some data points.

Anyway, I suggest that in the experimental branch, we follow the standard rules (1 approval required) and nobody blocks PRs with the "Requesting changes" flag (I dismissed mine). That way, there are always 2 people in favor of the change, but there is no veto.

[317070]

more time spent in testplay (or slower progress)

I think I see this differently. Test-play is the fast way to settle issues. Because so few ideas are tried, progress is getting really slow. If in every test game we would try 2-3 ideas, progress would be faster than when we try 1 new idea every 2-3 games.

So, I would mainly look at the experimental branch with the proof of burden on the one vetoing, rather than the one proposing. It is not up to the one proposing to prove it will work (because that is super difficult and an endless discussion) but on the one vetoing (because if he is right, test games should show anyway, nothing is lost).

So, in my view, experimental branch should be reviewed on a willing-to-try basis. If you are 100% sure it will not work, you will probably be able to prove it to the one submitting anyway.

[Tantali]

in the experimental branch, we follow the standard rules (1 approval required) and nobody blocks PRs with the "Requesting changes" flag

Sounds good.