ChessBase just posted a new article by A. Silver at https://en.chessbase.com/post/how-a-neural-network-is-made https://en.chessbase.com/post/how-a-neural-netwo

ChessBase just posted a new article by A. Silver at https://en.chessbase.com/post/how-a-neural-network-is-made https://en.chessbase.com/post/how-a-neural-network-is-made where the author tries to argue that doubling the neural network size in StockFish makes for an appreciable strength increase. However, only blitz results are shown. Looking at http://computerchess.org.uk/ccrl/4040/ http://computerchess.org.uk/ccrl/4040/ we see Fat Fritz being only 2 ELO stronger than SF at blitz while losing out on longer time controls. This questionable tradeoff makes for a very poor excuse to charge $100 versus the free SF. Public comments are once again blocked.