Post by steadyeddie on Dec 18, 2016 6:07:25 GMT -8
Having exceeded the result I felt SteadyEddie might be capable of, I feel it's time to splurge out the design and thinking. So here's a very long, and unless you are me, probably boring explanation of what I've done with the last year and a half of my spare time sitting in car parks waiting for my children to finish playing sport. It's embarrassing compared to the complexity available in the Sancho blog (http://sanchoggp.blogspot.co.uk/p/table-of-contents.html).
I'm hoping to achieve three things.
Firstly, if there are beginners out there looking for ideas on which to build their own framework, this might provide some clues. I've attempted to reveal all my algorithms. I'd like you to avoid the countless dead ends I went down.
Secondly, for the more experienced in the community there may be something in here that is new to you that you can use. Perhaps not, you guys are a very smart bunch. I guess it's even possible, however unlikely, there is some idea worthy of wider publication, in which please give me appropriate credit.
Finally, I expect the community at large to have better ideas than mine in many areas, and might share them, or just be able to correct my faulty thinking. That'd be much appreciated.
It's likely you are going to find the reason for SteadyEddie's success, mostly raw MCTS power, a little underwhelming; I certainly find it a disappointing state of affairs.
I'm hoping to achieve three things.
Firstly, if there are beginners out there looking for ideas on which to build their own framework, this might provide some clues. I've attempted to reveal all my algorithms. I'd like you to avoid the countless dead ends I went down.
Secondly, for the more experienced in the community there may be something in here that is new to you that you can use. Perhaps not, you guys are a very smart bunch. I guess it's even possible, however unlikely, there is some idea worthy of wider publication, in which please give me appropriate credit.
Finally, I expect the community at large to have better ideas than mine in many areas, and might share them, or just be able to correct my faulty thinking. That'd be much appreciated.
It's likely you are going to find the reason for SteadyEddie's success, mostly raw MCTS power, a little underwhelming; I certainly find it a disappointing state of affairs.
