Perfect play may be weaker against handicap stones than we think.
Against 5 stones for example it will see that winning is impossible(against other perfect player), and will start to surround less big territory instead. Later top human opponent will make mistakes and perfect player will start to surround bigger territory - but fuseki is already not optimal - and it will lose.
So I got and idea: maybe its really possible to create bot that plays better than “perfect play” (against humans with handicap stones).
Perfect player starts with smaller fuseki because surrounding bigger territory would be impossible in 100% cases.
Artificial Bot will start with big fuseki anyway - and this may compensate strength difference. It will try to win even if probability of it is 10%.
For example: Perfect player will win in 1% cases against 5 stones(when top human has bad day). Artificial Bot will win in 50% cases against 5 stones and in 10% cases against 6.
Artificial Bot will use trick moves - it has access to human games database.
I already talked about it in this topic. What I’m saying now - that its probably will be possible in some years later. If human games database is not enough, then it may use data from neuro implant like Neuralink - some more years later. And if that’s not enough - then it may scan brain of opponent and make a simulation - even more years later.
Bot that plays better than perfect play against handicap seems totally possible. But when it happens, we will not know about it - since perfect play on 19x19 is probably not possible in our galaxy.