Could it be that the computer considers white's position to be a bit desperate?

_________________
Patience, grasshopper.

Good point - let me check on the exact percentage when I get home.

_________________
be immersed

No, LZ consider that white has the advantage (a little over 65% winrate). And it start a fight, but white don't die (or not easily/without compensation)

What Kirby saw is probably the result of a very low playout numbers (and that's unreliable when fighting).

I get the same results as Tryss (network #145, 50k playouts). Leela Zero considers lots of similar moves in the original position to be of roughly equal value, including both the actual move and the one Kirby's run found.

Is there any rule of thumb that says "X-many" playouts leads to a decision that's close to optimal? I guess it must depend on the board position, so I'm probably asking for something that's not feasible to be expected.

Given that, I'm not sure how much I can trust Leela Zero on my low-grade computer. Maybe I can use it to get some good first guesses on moves and/or intuition on areas to play. But it looks like actual fighting, etc., can't be trusted given that there's always a chance that I haven't done enough playouts to get close to an optimal result...

_________________
be immersed

Yeah, I think that's about right. Even just a few visits is good for positional intuition, and I believe just one visit is already well into amateur dan level, but fighting still depends on reading, which means visits.

I'd say, for non-complex fights, 3-5k playout is often enough. If the fight become complicated or if LZ consider seriously several different moves, the sky is the limit.

I'm certainly not a dan player, but I'm surprised at people showing such a preference for black here.

My first instinct was that I would much rather be white. It's not that I think that the position is clearly better for either player, but just because white has solid cash (except for the ko) and his goal is now clear (even if the means of achieving that goal are not.) black seems to have all the big calls to make about how fighty to be

My first instinct would be to play something solid from the top group, but on looking at the centre a bit more I think I would go deeper, but not as deep as Carlo (and would loose!)