Kepler recommended to me the book How to Choose a Chess Move by Andrew Soltis and I brought it and it is incredibly helpful (thanks Kepler). In the Chapter "Reality Check" he brings up a fascinating point and some very interesting data from two detailed studies of GM games published in Chess Informant.
First, the point:
The greatest limitation we regularly face in finding the best move is not lack of time. Nor is it our insufficient powers of visualization. It is the absence of a best move.
Next the data. He made two studies of 50 random games in separate Chess Informant dividing the moves into one of four categories:
Book - established opening or endgame theory
Forced - routine recaptures, mandatory responses to checks or other one move threats.
Best - Non-forced move situations in which there was one move which post-mortem analysis proved to be best.
Discreationary - Remaining positions; none of the choices are clearly "best".
The percentages of each type of move were (combining the two studies 😵:
Book - 29%
Forced - 6%
Best - 30%
Discretionary - 35%
Musing on these figures they certainly show that forced moves are relatively rare. What light does this info throw on previous threads regarding the differences between the play of an engine and that of strong human players?
BTW, the book has a Fischer quote I am quite fond of and consider to be some of the best advice I've ever read: Don't worry about finding the best move. Just try to find a good move.
Glad you like it. As I recall it wasn't the forced moves that were causing the trouble, they are always going to be a tiny percentage. It is those situations where the best move is clearly best, the sort of move that an engine will find and so will a good human, that worried me. I wasn't worried because including them might indicate engine use, that situation would be picked up by the game mods examining the games in question. What worried me was that the average number of "best" moves is so high and could be used to justify high match up rates.
Originally posted by KeplerForced and best combined are approximately equal to discretionary. Even assuming that a strong player could always find the "best" move (a dubious premise), he'd also have to agree with the engine's discretionary choice 80% of the time to reach a 90% match up rate.
Glad you like it. As I recall it wasn't the forced moves that were causing the trouble, they are always going to be a tiny percentage. It is those situations where the best move is clearly best, the sort of move that an engine will find and so will a good human, that worried me. I wasn't worried because including them might indicate engine use, that situation ...[text shortened]... average number of "best" moves is so high and could be used to justify high match up rates.
How likely is that?
Originally posted by KatastroofHopefully. We'll see in the next few weeks; I'm playing in a Quad next Sunday, a tournament in either Utica, NY or the Bradley Open in Connecticut in the 2nd weekend in August and the New York State Championships on Labor Day weekend.
Apart from intresting and what you can make of it for the engine usage issue.What I want to know: did it help your chess any?
Originally posted by no1marauderI don't want to jinx you, but I find when I study anything that deeply impacts my chess in some way, or if I take lessons, I play worse in the short time period after. I think this was discussed in another thread, and is probably quite common. Read the book, but then give yourself enough time after to flush everything out until you're just playing like yourself again.
Hopefully. We'll see in the next few weeks; I'm playing in a Quad next Sunday, a tournament in either Utica, NY or the Bradley Open in Connecticut in the 2nd weekend in August and the New York State Championships on Labor Day weekend.
Originally posted by no1marauderIt would justify the idea that someone routinely cranking out a 20-50 move forced draw or win is possibly using a chess engine.
Musing on these figures they certainly show that forced moves are relatively rare. What light does this info throw on previous threads regarding the differences between the play of an engine and that of strong human players?
Originally posted by no1marauderVery unlikely.
Forced and best combined are approximately equal to discretionary. Even assuming that a strong player could always find the "best" move (a dubious premise), he'd also have to agree with the engine's discretionary choice 80% of the time to reach a 90% match up rate.
How likely is that?
Originally posted by no1marauderInteresting stats.
The percentages of each type of move were (combining the two studies 😵:
Book - 29%
Forced - 6%
Best - 30%
Discretionary - 35%
This would mean 71% of moves are non book and about 50% (36% of the total) are either forced or clearly best.
If we put this into engine terms what would constitute clearly best? I would have thought (and this is really just a guess) something in the region of the following
below 1400 probably 0.80 to 1.20;
1400-1700 say 0.60 to 0.80;
1700-2000 say 0.40 to 0.60;
2000-2300 say 0.20 to 0.40;
above 2300 possibly 0.10 to 0.20.
So if my assumptions are reasonable more and more of the best will become discretionary as ratings decrease. If you are graded 2150 then anything between 0.10 and about 0.30 would count as discretonary but forced or best would reduce slightly, say to 48%.
If the average number of moves between 0.10 and 0.30 was 5 on the remaining moves the chance of picking 1 would be 20% of the 52% of moves in this category or 10.4% of the total. So the odds of matching an engines 1st choice would become only 58.4%, the odds of matching 1st or 2nd choise 68.6% and the odds of matching choices 1-3 would become 79%.
Of course the odds of matchingan engines 1st 5 choices in these circumstances would approach 100% as any other move would in theory be a blunder.
Anything over these figures over a sufficient number of games / moves could not be down to normal play.
Of course I have used guesstimates as my percentages just to illustrate the idea but if realistic percentages can be determined then this becomes a perfectly reasonable way to conclude someone is not real
Originally posted by Dragon Firethat's true, and I think it's a very good idea, but for that, we need a sample of real and strong correspondence chess players (a collection of games) from this site. (to determine those realistic percentages.)
[b]Of course I have used guesstimates as my percentages just to illustrate the idea but if realistic percentages can be determined then this becomes a perfectly reasonable way to conclude someone is not real[/b]
I am intrigued by this quote of Bobby Fischer's: "Don't worry about finding the best move. Just try to find a good move."
Some time ago I heard contrary advice, which was to, after finding a good move, try to find a better one. You can imagine what taking this maxim to heart did to my playing time here at RHP with it's unlimited playing time (virtually unlimited). I typically spend 20 to 30 minutes on most moves, looking for that "better move." I might do well to consider backing off and just looking for a good move . . . but I must play the best game I can!
Originally posted by bassoBobby wasn't talking about CC.In OTB if you found a good move play it or the 17th piece will kill you.
I am intrigued by this quote of Bobby Fischer's: "Don't worry about finding the best move. Just try to find a good move."
Some time ago I heard contrary advice, which was to, after finding a good move, try to find a better one. You can imagine what taking this maxim to heart did to my playing time here at RHP with it's unlimited playing time (virtually unl ...[text shortened]... ng off and just looking for a good move . . . but I must play the [b]best game I can![/b]