I'll be happy for help in settling a dispute.
In an online discussion I participated in, I mentioned the following:
By "lossy" I mean that it could miss relevant variants within the specified search depth; Compared to vanilla Alpha-Beta, which only prunes irrelevant possibilities and is thus lossless.Modern chess software is "lossy" - it aggressively prunes possibilities that potentially could be relevant, but statistically are unlikely to be so, in order to reach such high depth that the tactics become akin to strategy.
This is what I gathered from my reading of resources like CPW and other references. My understanding is that reaching high depth is key, and this cannot be achieved without some sacrifices. The vast improvement in modern software like Stockfish compared to older software is precisely that it manages to make these sacrifices in an optimized way.
However, an interlocutor replied with the following:
He also referred to the writings of one Hans Berliner as proof of his claim that chess software is lossless.this simply isn't true. This approach was abandoned decades ago.
So, which is true? And is there a clear, concise, reliable reference which unambiguously answers this question?