al

@sdesrozis Thanks for the response. Yes exactly. I'd imagine `EarlyStopping` would be a specific case of the more general `NoImprovmentHandler` by passing in the `Stop Policy Function`. So `NoImprovement` determines...

Not sure which is more suitable, `no-improvement` may even be thought of as an scenario-based event, and actions like `stop training` can then register as an event handler when `no-improvement`...

Looks great for both current need and flexible. What is `stuck_function`? Is it just typo of `pass_function`?

I'm a bit confused on the expected performance for the small network. How is it that training on 100 games can give 95% accuracy on validation set, yet for 15k...

After applying the above fix, the error still appears. Should we make it return pass, if no more legal moves? PS, It seems after adding pass turn as legal moves,...

For me, the network doesn't train most of the time. 

Shallower and smaller network architecture, and dense weight initial value scaled with 0.1 worked better for me. 

I'm also confused by this line `while (not node.can_add_child()) and (not node.is_terminal()):`. At the beginning of the round, node is root, `node.can_add_child()` will be true, and `not node.can_add_child()` will be...

Regarding the uneven tree expansion part, how likely `select_child` returns different child? e.g. if it returns child `child-1` at some round, will it tend to still return `child-1` or more...

Oh on another look, since `select_child` is in a while loop, even if it returns same child under the root, once exhausted the child's legal moves, it will select a...