Engine losing due to tactical positions

[connor_mcmonigle]

...
Also, saying that evaluation has great impact on move order is partially true. It will determine what the first move to search will be, but the order of remaining moves mostly relies on other heuristics.

Through the history heuristic, the order of the remaining moves is largely determined by the evaluation function. Better evaluation function -> better move ordering.

[jtwright]

I think there's a slight misunderstanding on the depth comment. I could be wrong, but my impression of what Mike was trying to get at was not that more depth > less depth all the time, but rather that it may be meaningful if in position X, stockfish finds the right move by nominal depth 11 and engine B takes until depth 30. SF is certainly not shy about reductions/pruning, so this may be indicative of some blindness or overly aggressive or badly applied pruning in engine B. Though I agree in general that the concept of "depth" as reported by an engine is in general nebulous.

Also, I agree that evaluation can have a pretty profound impact even on how quickly tactics are found, as some above me already mentioned. During the iterative deepening loop, the shallow searches inform how the history tables and other similar structures get filled in. A "good" evaluator here will be able to ideally sniff out fruitful lines of play early, giving better guidance on subsequent iterations of the loop. It may also allow for recognition of fruitful positions even before something like an exchange of material has taken place.

As for which is at play in this particular instance, I cannot say, unfortunately.

[Mike Sherwin]

I think there's a slight misunderstanding on the depth comment. I could be wrong, but my impression of what Mike was trying to get at was not that more depth > less depth all the time, but rather that it may be meaningful if in position X, stockfish finds the right move by nominal depth 11 and engine B takes until depth 30. SF is certainly not shy about reductions/pruning, so this may be indicative of some blindness or overly aggressive or badly applied pruning in engine B. Though I agree in general that the concept of "depth" as reported by an engine is in general nebulous.

Also, I agree that evaluation can have a pretty profound impact even on how quickly tactics are found, as some above me already mentioned. During the iterative deepening loop, the shallow searches inform how the history tables and other similar structures get filled in. A "good" evaluator here will be able to ideally sniff out fruitful lines of play early, giving better guidance on subsequent iterations of the loop. It may also allow for recognition of fruitful positions even before something like an exchange of material has taken place.

As for which is at play in this particular instance, I cannot say, unfortunately.

Basically true. However, the RC vs SF comparison was just to highlight what I was trying to point out. The real comparison would be engine A with a weak evaluation to engine A with an improved evaluation.

[jdart]

Reasonably good king safety evalution is a must - if you do not have this and your opponent does, you will lose regularly to those opponents.