What you say makes sense. Danny Rensch said that a CAPS-2 was in the works in the recent "State of Chess" broadcast. Perhaps someone knows if any members are included.
Such questions deserve a look.
The analysis tool can't count
Sort:
Apr 13, 2021
0
#3
While this game, which features a potentially unsound gambit, but follows the great originator Gioachino Greco 100% netted a 99.6!
http://chessskill.blogspot.com/2020/01/near-perfect.html
Weird.
Apr 13, 2021
0
#4
Indeed. I have seen a >99% accuracy game that had a crucial blunder in it, which was a surprise.
Note that the "accuracy" stat discussed in the last two posts is not reduced by playing book moves in the way I have realised the "best moves" percentage is. Although the exact calculation is very obscure, "accuracy" appears to ignore book moves.
Apr 19, 2021
0
#5
For convenience I am going to post an image here of an unrelated problem. In this position, the slower 49. ...b4 was played. The fastest mate is 49. ... Rb3. When you try 49. ...Rd8+ as an alternative (via "Moments"), the tool says it is the best move (which is sort of true if you ignore the time to mate), but then confusingly gives the line after the faster move Rb3 as analysis for this best move.
As I see it, the reason for this is that it is getting confused which "best move" it is talking about.
Apr 19, 2021
0
#6
MelvinGarvey wrote:
Oh, and I'll unfollow this thread before you block me, since you go block what you don't understand.
Well that was rude. He hadn't even provoked you. But anyways
Apr 20, 2021
0
#7
I don't see why he would think I would block him, nor is there a basis for his claim. I would say he was exaggerating: they do care - just not enough to fix things efficiently, in my experience.
Apr 20, 2021
0
#9
To get back to the point, best move percentage would seem to be a pretty straight forward calculation.
Apr 20, 2021
0
#10
Yes. The definition used is, however, clearly not ideal. It would be fine to ignore opening moves but counting then all as failures to match the engine (even when they happen to do so) is absurd.
Apr 20, 2021
0
#11
MelvinGarvey wrote:
On here since 2008 and only realize now cc doesn't care about anything such as "truth", "accuracy", justice", "math", "vocabulary" (etc, etc...), but only about what makes their site look good or look bad in the eyes of potential pay account buyers, a very large majority of them being clueless about all or most of the notions I listed above.
Bah, the apple never falls far from the apple tree.
Welcome back.
As for the math created by the site’s analysis, it is obviously way ahead of its time.
In this game, played this morning, I had two blunders, one mistake, one missed win, and other inaccuracies in 23 non-book moves, yet still managed a CAPs above 93%.
The first blunder dropped me by five pawns, but I retained an advantage. The second blunder set up my opponent’s missed win. The mistake dropped me from a decisive advantage to slightly worse.
https://www.chess.com/analysis/game/live/12657421639?tab=report
Apr 20, 2021
0
#13
99.7% for two best moves of two possibilities. But, every move is in books. Shouldn’t it be 7 book moves?
https://www.chess.com/analysis/game/live/12671759393
Apr 20, 2021
0
#16
s7silver wrote:
How was White able to make 9 more moves than black?
Chess.com has an advanced math system.
Apr 20, 2021
0
#19
MARattigan wrote:
No wonder he won.
I think the dikkin’ around confused the computer. Sometimes when a player cannot find the resign button, I’ll mate him with a specific piece or on a specific square. It may be that moves in which there was only one legal move are not listed in the counts. That would give White more moves.
Apr 20, 2021
0
#20
Elroch wrote:
There are no provided definitions for the various statistics but one would expect the best move percentage to have a close relationship to how many best moves were played. Does it? Not as you might expect.
https://www.chess.com/live#g=12061959371
In this game white played 11 moves, the first 5 of which were the mainline Ruy Lopez. Of the remaining 6 moves, the first is rated by the tool as "Excellent" and the other 5 as "Best move". Yet it also says that the best move percentage was 45.4%.
This percentage clearly corresponds to 5/11 so I infer it includes the book moves in the denominator (11) but not the numerator (5). Unfortunately for this justification, some of the moves in this opening line happen to be clear engine top choices. This is most clear in the case of move 5, where there is only one move that does not lose, 5. Bb3, retreating the attacked bishop.
So, what you appear to have is a percentage that counts all opening theory as inferior moves. Games where you follow theory longer have lower best move percentages as a consequence.
I suspect this was the opposite of what was intended. It seems plausible that the intention was to count all opening theory as "best moves", while instead they all get counted as "non-best".
Mathematics aren't marketable, and CEOs need a 4 year degree to learn how to use a calculator.
I think that about sums it up.
There are no provided definitions for the various statistics but one would expect the best move percentage to have a close relationship to how many best moves were played. Does it? Not as you might expect.
https://www.chess.com/live#g=12061959371
In this game white played 11 moves, the first 5 of which were the mainline Ruy Lopez. Of the remaining 6 moves, the first is rated by the tool as "Excellent" and the other 5 as "Best move". Yet it also says that the best move percentage was 45.4%.
This percentage clearly corresponds to 5/11 so I infer it includes the book moves in the denominator (11) but not the numerator (5). Unfortunately for this justification, some of the moves in this opening line happen to be clear engine top choices. This is most clear in the case of move 5, where there is only one move that does not lose, 5. Bb3, retreating the attacked bishop.
So, what you appear to have is a percentage that counts all opening theory as inferior moves. Games where you follow theory longer have lower best move percentages as a consequence.
I suspect this was the opposite of what was intended. It seems plausible that the intention was to count all opening theory as "best moves", while instead they all get counted as "non-best".