REAL Elo Machine playing humans

[fern]

I understand that normally elo programs and machines are got from competitions between engines themselves. So the pool from which that measure comes is formed just by AI.
Nevertheless, from the point of view of a human playing a program, things can be very different. IF a human has not tried to keep in memory the weaknesses of an engine in order to repeat indefinitely a win, then chances are that the human side will have less opportunities to exploit those weaknesses when they appear at the game than the chances of the engine to exploit human weaknesses, that occurs in the realm of tactics.
I did an experiment yesterday putting a strong human player -not me- against old Chess Champion Champion by Fidelity, which in the ratings given by different sites appears as just a 1570 of so Elo machine.
The human player -as I said, a strong club player rated around 2200- played two games against the machine. The firs he lost just after losing one pawn first, then a second after sa tricky maneuver of Champion. The last game he got a draw after difficult fight and only because Champion dos not know that tou cannot win with King and Bishop against King.
All in all the human side said to be surprised and awed. Champion executed very clever threats, played a good ending -until that failure I mentioned- and my friend said he felt as playing a strong human player of around 2000 or more.
I have the feeling he is right and that many of the alleged wins of human in these confrontations are more the result of repeating an already won game and/or taking back moves and many things of the sort that my chess player friend did not.
If you play machines without any sort of pre--arranged tactic as if playing a human, chances are you will face LOT more problems that those suggested by the theoretical Elo.

My best
Fern

[Codesquid]

There's another reason that makes comparisons even more difficult, namely the hardware the engine runs on.

I suppose most engine-engine matches are played with both engines on the same computer. If engine A is 100 elo stronger than engine B on a very slow computer, it would likely be the same on a computer that's a hundred times faster.

Yet if you were to put engine A on the slow computer playing against engine B on the fast computer, results would probably be different.

The problem with humans is that they cannot run on a different machine, soft- and hardware are one inseparable unit. Thus when comparing the strength man vs. machine, one needs to include the hardware the engine runs on, only then are the numbers comparable.

[fern]

Certainly it is so. Nevertheless I concentrate here more in the subjective aspect of this confrontation. I have experimented this thousands of time. I am a decently strong chess player in a good day, that is, if I play putting myself totally in the game. If not, if I multitask, if a drink a beer as I play, if a lose perhaps just a 20% of concentration, If i analyze just 5 or 6 ply each game instead of seven, if I do all that I can easily lose even with a 1300 elo machine. In human Vs machine games, it happens that the human side lose a LOT of capabilities with just a scarce diminution of concentration power. I mean, you go straight from 2000 to 1100 or so. We behave not along a continuous, but with abysms from one point of the scale to the other.

Fern

[Don]

Certainly it is so. Nevertheless I concentrate here more in the subjective aspect of this confrontation. I have experimented this thousands of time. I am a decently strong chess player in a good day, that is, if I play putting myself totally in the game. If not, if I multitask, if a drink a beer as I play, if a lose perhaps just a 20% of concentration, If i analyze just 5 or 6 ply each game instead of seven, if I do all that I can easily lose even with a 1300 elo machine. In human Vs machine games, it happens that the human side lose a LOT of capabilities with just a scarce diminution of concentration power. I mean, you go straight from 2000 to 1100 or so. We behave not along a continuous, but with abysms from one point of the scale to the other.

Fern

In the old days it was possible for an engine to beat 2 or 3 experts or master in a row and then some 1600 player would beat it easily, because he really understood all the weakness of the program. And often the strong players would lose a game or two but once taking it seriously the machine was putty in his hands.

It's probably still like that, but on a whole different scale, where stronger players are required to pull it off and much less frequently. I think for strong GM's a draw against a computer is a very good result.

[fern]

BTW, even if it sound weird, I have always tried NOT TO learn the weaknesses of my dedicated units. In my logic, where fun is the thing to get, to do such a thing would be stupid as much I would lose at once the fun capabilities of the contrivance, not to mention I would make zero of my purchase.
I try always that each game be as the first I had against a machine. I try to win, but not because I know from the beginning how to do it.
The so called anti-computer chess is for me the the pinnacle of absurd.

Fern

[carldaman]

I think it may be safe to say it is 1570 rated machine that is also a lot stronger tactically than that rating suggests, while grossly lacking in strategic and positional strength.

CL

[fern]

That is a very good way to say what I said in less precise terms. Clearly most games are won or lost in the tactical realm, specially in the case of not too much strong players, so yes, Champion probably is a 2100 or so tactical player, but maybe lot less in endings, though not that bad as you can think.

Fern

[Don]

BTW, even if it sound weird, I have always tried NOT TO learn the weaknesses of my dedicated units. In my logic, where fun is the thing to get, to do such a thing would be stupid as much I would lose at once the fun capabilities of the contrivance, not to mention I would make zero of my purchase.
I try always that each game be as the first I had against a machine. I try to win, but not because I know from the beginning how to do it.
The so called anti-computer chess is for me the the pinnacle of absurd.

That must be why computers love to play you!

Seriously, I once had the beautiful wooden chess board ELITE fidelity machine and we were pretty evenly matched at the time. But for fun, almost as a puzzle to solve, I learned to beat it every time at several relatively fast levels by memorizing the moves I needed to play against every opening we would reach. I first learned to get it out of book quickly and then I had to learn to deal with the variety I would sometime get. Usually it played the same but in some position it would vary - so I had to learn to win from that position too. When I made mistakes I learned that I had to play something different. Over time I could beat it every time - and so I would move to the next level and start over. In a few case I could get a good start by playing what I had already learned as it would often make the same mistakes. But I never varied based on the level.

Strangely enough, I think this improved my play a great deal even though it was kind of a trick. It was a lot like doing opening preparation and in real tournament games I sometimes got to use that knowledge.

So I recommend this for opening preparation. Use your own openings and learn to deal with whatever the computer throws at you. If a human plays something unexpected ADD it to the book. After every loss you normally do that anyway, right?

Fern

[fern]

Well, no. I do not do such a thing or just occasionally if by chance I remember how I lost the precedent game. Rarely has been in my mind the goal to improve my game along a systematic path. If ever I have improved, it has been on the ground not of better knowledge, but of a deeper way of thinking. Even so, at my 64 years of age, I still cannot supersede my old time mistake of NOT taking into account parallel chances. I can go deep in a line forgetting that in the middle of the tree there is a node that takes the game, or can take it, to a very different result.
But maybe in any case, at this age, with less capability to concentrate really hard in a game all the time, I am already in the path of the deterioration of my game, not improving it.

Third age regards
Fern

[Henk]

BTW, even if it sound weird, I have always tried NOT TO learn the weaknesses of my dedicated units. In my logic, where fun is the thing to get, to do such a thing would be stupid as much I would lose at once the fun capabilities of the contrivance, not to mention I would make zero of my purchase.
I try always that each game be as the first I had against a machine. I try to win, but not because I know from the beginning how to do it.
The so called anti-computer chess is for me the the pinnacle of absurd.

That must be why computers love to play you!

Seriously, I once had the beautiful wooden chess board ELITE fidelity machine and we were pretty evenly matched at the time. But for fun, almost as a puzzle to solve, I learned to beat it every time at several relatively fast levels by memorizing the moves I needed to play against every opening we would reach. I first learned to get it out of book quickly and then I had to learn to deal with the variety I would sometime get. Usually it played the same but in some position it would vary - so I had to learn to win from that position too. When I made mistakes I learned that I had to play something different. Over time I could beat it every time - and so I would move to the next level and start over. In a few case I could get a good start by playing what I had already learned as it would often make the same mistakes. But I never varied based on the level.

Strangely enough, I think this improved my play a great deal even though it was kind of a trick. It was a lot like doing opening preparation and in real tournament games I sometimes got to use that knowledge.

So I recommend this for opening preparation. Use your own openings and learn to deal with whatever the computer throws at you. If a human plays something unexpected ADD it to the book. After every loss you normally do that anyway, right?

Fern

When I was 15 years old I bought a chess challenger von fidelity[Deutsche sprache]. It costed me 500 guilders and I only earned 5 guilders a week. After one year I almost always won the game. In the next one or two years I only used it for five or ten minutes games. I could also set the challenger on analysis level. So I could play games that took a few days or more. But the playing of the challenger did not improve very much.