Yuki Kobayashi

[sim_log.zip](https://github.com/koban6/Ray/files/783294/sim_log.zip) Ray sometimes plays white N1. But it's difficult for Ray to throw in L2 or L3 after capturing stones (L2 and L3). The file has 10 Ray's simulation samples....

@CGLemon Probably my implementation is essentially the same as Remi's implementation. ```Ej``` in my implementation holds the sum of $\gamma$ of all legal moves, so ```Ej``` is following equation, $E_j...

@CGLemon We need only following 3 parameters for updating gammas. - $W_i$ : The number of winning count for feature $i$. - $C_{ij}$ : Sum of features' gamma related to...

I'm glad you can train patterns well. Here is my paper's link. http://id.nii.ac.jp/1438/00001972/ I'm a little glad someone gets my paper from this page, because we can see the stats...

Length of PV depends on the number of expanded nodes. If your program expands a node when a leaf node visit is only one time, length of PV will be...

When we used only Monte-Carlo simulation for evaluating positions, evaluations are unstable. Each simulation has a mix of wins and loses. This is because Monte-Carlo simulation uses random number for...

Thank you for information. I tried to test current Ray in my PC, I can confirm current Ray is getting weaker. | options | ver 10.0.1 (commit hash #a914a70) |...

I already found training program's bug. But this issue is not solved yet. I think bugs are in move evaluations. Please wait for a while.

I'm sorry to keep you waiting. I finally fixed bugs and released [v10.1.0](https://github.com/kobanium/Ray/releases/tag/v10.1.0). I tested v10.1.0, results as follow; | options | v10.1.0 (commit hash #fe615d6) | ver 9.0.1 (commit...

Now I released v10.3.0. This is little stronger than v10.2.0. | options | v10.3.0 | v9.0.1 | Elo rating diff | | --- | --- | --- | --- |...