If you came here for a brief (or long) history of chess computers, then you came to the wrong place. What I want is a chess computer history.
I've heard people say that Stockfish is the best and people say that AlphaZero is the best. I don't know which is the best when, because I think they were both at the top at some time.
You don't want a history of chess computers.
Instead, you want a chess computer history...
Perhaps a good night's sleep would help (you).
If you're reading this for an answer to your question, then you're reading the wrong post. What I'm giving is the question's answer.
https://en.wikipedia.org/wiki/Top_Chess_Engine_Championship#Tournament_results_(TCEC)
Alpha Zero beat Stockfish in 2018 in a closed door match (The games were not released, the match was not witnessed and the Stockfish team had no input). The match was actually not a chess engine match... at least not only. Alpha Zero also beat the best Go and Shogi engines.
It was a publicity stunt by Google to show off their neural network research, not to show off a chess playing apparatus.
There were more dubious aspects:
The version of Stockfish they used was old
The hardware setup was untested and strange
The time control was untested and strange (again, SF team had no input)
Even so, in 2018 Alpha Zero was probably very slightly better than Stockfish even though it never played a real match... but since then chess engines have continued to improve, and both Leela and Stockfish have closed that very small gap years ago.
This isn't to say Google couldn't flex its billion of dollars and thousands of engineers to one up Stockfish again some time in the future... but that was never what it was about. Google isn't in the business of making chess engines 
I'm even more confused than I was when I made this a few hours ago. So, before 2018 Stockfish was the best, then AlphaZero was the best for a while and now Stockfish again?
Follow the wiki link. Houdini, Komodo, Stockfish, and Leela have all be at the top at one point.
But yes, basically that's right.
2018 Stockfish
late 2018 / early 2019 Alpha Zero
2019 Leela (which is a neural network engine like alpha zero)
2020 Stockfish
2021 Stockfish
Well, I'm talking about recently, like maybe 5-3 years. What is the difference, I wonder, between Leela and AlphaZero?
And they're pretty close... looks SF scored 53 out of 100 vs Leela.
So it's about 20 Elo stronger.
Well, I'm talking about recently, like maybe 5-3 years. What is the difference, I wonder, between Leela and AlphaZero?
CCRL lists Stockfish 8 as 3377
AZ scored 64 out of 100 so its estimated Elo is 100 points higher, 3477.
CCRL lists Stocofish 12 as 3551
So about 74 Elo stronger.
(Remember ratings aren't absolute, what matters is the difference. It's not that SF12 is literally 3551, it's that SF12 is literally about 74 rating points higher than AZ)
Ok, so Stockfish ruled, AlphaZero taught itself chess and beat SF. Then SF got updated and is now better than AZ.
And... there's an issue with time controls and hardware, but if you want a simple answer, there you go
Ok, so Stockfish ruled, AlphaZero taught itself chess and beat SF. Then SF got updated and is now better than AZ.
Yes.
And even more interesting is SF and all engines before that used alpha-beta search. Alpha Zero is fundamentally different, it used a neural network.
So they were like, oh cool, let's copy that, and Leela was born, which is also NN instead of AB.
So Stockfish copied that, and (as far as I know) is now a hybrid AB and NN, google "Stockfish plus NNUE"
Stockfish is at least 3600 at 40 moves in 2 hours.
From our human point of view, all of the top engines are indistinguishably powerful. It takes long automated matches at accelerated time controls to filter out who is king of the hill.
Stockfish is at least 3600 at 40 moves in 2 hours.
From our human point of view, all of the top engines are indistinguishably powerful. It takes long automated matches at accelerated time controls to filter out who is king of the hill.
Yeah, they're barely better than each other.
And they're so monstrously better than humans it's hard to imagine. It's not that Carlsen would lose every game, it's that he'd lose every game even if he were given a huge handicap.
If you want to look at the very first US computer chess championship, see: The First Computer Chess Championship in the USA
If you came here for a brief (or long) history of chess computers, then you came to the wrong place. What I want is a chess computer history.
I've heard people say that Stockfish is the best and people say that AlphaZero is the best. I don't know which is the best when, because I think they were both at the top at some time.