The Strongest Chess Engine of all time. Alpha Zero???

It is amazing to me that even today. People believe Alpha Zero is the best chess engine of all time, and is still the strongest chess engine today. A Chess.com poll also confirms this belief by the chess community that Alpha Zero is the best chess engine of all time.

This shows you what a good job DeepMind did in the press releases. By cherry picking stats to promote to the public. And by only counting wins and losses of a 1000 game match and ignoring the draws.

But even under the best results against Stockfish 8 playing without a book, or proper time controls, and sub optimal hash table settings. Alpha Zero was still less then 50 Elo better then Stockfish 8. With Stockfish 8 playing on weaker hardware.

When you look at the match played in 2018 were Stockfish 8 was able to play with a book against Alpha Zero. And this is important as Stockfish 8 is not a neural network engine. But a hand coded chess engine. So Stockfish 8 does not have the advantage of knowing how to play the openings like a neural network chess engine.

Alpha Zero results got even worse playing Stockfish 8. Here are the results and Elo of that 100 game match played with books, but still with Stockfish 8 playing Alpha Zero with sub optimal settings on weaker hardware.

Computer Match  2018

                     
1   AlphaZero     +31  +17/=75/-8 54.50%   54.5/100
2   Stockfish 8   -31  +8/=75/-17 45.50%   45.5/100

Now Alpha Zero in this more proper match with books playing Stockfish 8 lost 8 games to Stockfish 8. Not the widely reported and only remembered 6 losses in a 1000 games.

Now Alpha Zero beat Stockfish 8 by only 31 Elo points.

And this is confirmed by my centipawn analysis error rate of both Alpha Zero, and Stockfish 8.

Alpha Zero scored a error rate of .03 CP
Stockfish 8 scored a slightly worse error rate of .05 CP

So the conclusion that Alpha Zero is the best chess engine of all time is clearly wrong. And by the Elo Statistics Alpha Zero is weaker then today's Dragon, Stockfish, Lc0, Berserk 10, Koivisto 9, Revenge 3, RubiChess, Ethereal 13.75, Slowchess Blitz, Seer, Rebel 16, RofChade, Minic, Igel, Uralochka, Arasen, and Ginkgo Vienna.

Same story with the IBM 1997 match against Kasparov, media exposure, Deep Blue by far is the strongest. One year later Rebel played 2 tournament games against Anand back then no.2 after Kasparov, Rebel lost the first, the second it had a won position that ended in a draw. But with a somewhat better hardware it had won the game. First sign at the wall what was coming. 4 years later (2002) Fritz played 8 tournament games against Kramnik 4-4. The next year (2003) Junior played 6 tournament games against Kasparov 3-3. Hardly (if any) mainstream media attention.

I am pretty sure that because of the media exposure (much more than with AZ) anno 2023 many people think that Deep Blue is still the strongest.

Admin wrote:

Same story with the IBM 1997 match against Kasparov, media exposure, Deep Blue by far is the strongest. One year later Rebel played 2 tournament games against Anand back then no.2 after Kasparov, Rebel lost the first, the second it had a won position that ended in a draw. But with a somewhat better hardware it had won the game. First sign at the wall what was coming. 4 years later (2002) Fritz played 8 tournament games against Kramnik 4-4. The next year (2003) Junior played 6 tournament games against Kasparov 3-3. Hardly (if any) mainstream media attention.

I am pretty sure that because of the media exposure (much more than with AZ) anno 2023 many people think that Deep Blue is still the strongest.

You are correct! If it were not for Alpha Zero ALL people would still think Deep Blue is the best chess engine of all times.

And what is crazier. When you do centipawn analysis of the Deep Blue and GM Kasparov 1997 match. The results show GM Kasparov was the more accurate player in that match.

GM Kasparov just self destructed.

I will do a CP analysis of the Rebel vs GM Anand match. I would not be shocked if Rebel played better then Deep Blue.

mwyoung wrote:

Admin wrote:

Same story with the IBM 1997 match against Kasparov, media exposure, Deep Blue by far is the strongest. One year later Rebel played 2 tournament games against Anand back then no.2 after Kasparov, Rebel lost the first, the second it had a won position that ended in a draw. But with a somewhat better hardware it had won the game. First sign at the wall what was coming. 4 years later (2002) Fritz played 8 tournament games against Kramnik 4-4. The next year (2003) Junior played 6 tournament games against Kasparov 3-3. Hardly (if any) mainstream media attention.

I am pretty sure that because of the media exposure (much more than with AZ) anno 2023 many people think that Deep Blue is still the strongest.

You are correct! If it were not for Alpha Zero ALL people would still think Deep Blue is the best chess engine of all times.

And what is crazier. When you do centipawn analysis of the Deep Blue and GM Kasparov 1997 match. The results show GM Kasparov was the more accurate player in that match.

GM Kasparov just self destructed.

I will do a CP analysis of the Rebel vs GM Anand match. I would not be shocked if Rebel played better then Deep Blue.

I find it hard to believe that most people can believe that a computer from 1997 is still the best in 2023 in something.
If people say that they do not know what is the best chess engine of all time then it is more logical.

Uri Blass wrote:

mwyoung wrote:

Admin wrote:

Same story with the IBM 1997 match against Kasparov, media exposure, Deep Blue by far is the strongest. One year later Rebel played 2 tournament games against Anand back then no.2 after Kasparov, Rebel lost the first, the second it had a won position that ended in a draw. But with a somewhat better hardware it had won the game. First sign at the wall what was coming. 4 years later (2002) Fritz played 8 tournament games against Kramnik 4-4. The next year (2003) Junior played 6 tournament games against Kasparov 3-3. Hardly (if any) mainstream media attention.

I am pretty sure that because of the media exposure (much more than with AZ) anno 2023 many people think that Deep Blue is still the strongest.

You are correct! If it were not for Alpha Zero ALL people would still think Deep Blue is the best chess engine of all times.

And what is crazier. When you do centipawn analysis of the Deep Blue and GM Kasparov 1997 match. The results show GM Kasparov was the more accurate player in that match.

GM Kasparov just self destructed.

I will do a CP analysis of the Rebel vs GM Anand match. I would not be shocked if Rebel played better then Deep Blue.

I find it hard to believe that most people can believe that a computer from 1997 is still the best in 2023 in something.
If people say that they do not know what is the best chess engine of all time then it is more logical.

People think many things what are just wrong.

Look at the YouTube Video above were you have a Grandmaster, and a chess expert. Telling everyone that Alpha Zero is the best chess engine of all time.

And he is not the only expert spouting such nonsense.

Many people believe what the experts say, and what the media tells them to think.

And that is nothing new.