I just spent some time analysing a recent correspondence game of mine and it is a good illustration of the limitations of the tool. The computer analysis only gives me 1 inaccuracy and 1 mistake and 12 ACL. For my opponent it gives 4 inaccuracies, 1 mistake 1 blunder and 30 ACL. In the opening and middle game, none of us were making any threat so it was impossible for us to make a significant mistake.
Then in the endgame I was lost several time but the computer analysis is not deep enough to see it. In fact three times the position went from drawn to losing and back but it takes time for the computer to realise it. So in truth, I blundered three times and my opponent also.
So it is important to not take this automatic analysis for granted and really look into our games to understand what the computer saw and learn from it.
Of course, I forgot the example of correspondence chess. Sorry :)
@Wolfram_EP slightly off-topic this, but how can one set of tablebases be more accurate that another? I thought that tablebases are complete solutions of positions with a certain number of pieces, so they are either completely right or wrong. Is the one tablebase just calculated with a higher search depth, and if so, how often does it give different results to Syzygy?
Lomonosov has 7 pieces, Syzygy only 6.
The analysis can also be made more accurate by augmenting the ply depth, but at the cost of more calculation time.
#11 With regard to engine evaluations (which are simply its opinions, unless it solves the position) if someone could test and manage the deployment of 4-man syzygybases for the entire fishnet cluster (a nontrivial task I failed at) that could help and maybe even reduce server costs.
@mCoombes314 There are no Syzygy bases for 7-men positions. But Stockfish and other top engines have an option to connect to 6-men (or less) Syzygy when analysing 7-men (or more) positions so that they can determine the precise assessment of 6-men whenever it occurs in the search tree. Of course, there are 7-men positions with so deep search trees that this doesn't help an engine to determine the precise value.
Also note that, unlike Syzygy, Nalimov (Lomonosov) bases are not precise under FIDE rules, because they ignore 50-move rule. This can lead to funny situations when engine games are adjudicated by Nalimov bases but engines themselves use Syzygy. For example, this position (http://tcec.chessdom.com/archive.php?se=9&sf&ga=17 ) is drawn under FIDE rules but is a win for white in "absolute chess" without 50-move rule.
I was looking at the computer analysis and I got that question. As several times on an old account which I closed down. I once had the computer analysis and it didnt seem to be accurate. I guess we can just say it is fairly accurate for mistakes and blunders.
This topic has been archived and can no longer be replied to.