Stockfish Deep Game Analysis Failure

In theory, newer versions of Stockfish play better than older versions, but the changes from one version to the next have been incremental. I still have a program on my phone that uses Stockfish 6, but for most positions, it gives the same analysis as newer versions.

Stockfish 12 started using AI for evaluation, but this was improved with Stockfish 16, when the older evaluation methods were completely abandoned. However, for most positions, it still chooses the same moves.

For game analysis, one doesn't have to look very deep with Stockfish to find the correct move, especially in the middlegame. Chess.com used an 18-ply search, which could analyze a whole game in seconds, but now they don't say how deep their analysis function looks. I suspect that it varies depending on the demand on their servers because occasionally Chess.com won't analyze a game, stating that its servers are busy. Hypothetically, there wouldn't be much difference with a 16-ply or a 17-ply search.

I notice very little difference between 20-, 30-, and 40-ply Stockfish analyses, except in the opening. The deeper analysis often prefers different opening moves. For this reason, I would analyze some of my games at 40 ply to see which opening moves it prefers. Rarely, I analyzed even deeper just to see if the opening choice changed.

However, I noticed a problem starting with Stockfish 17.1. According to Google AI, the difference between versions 17 and 17.1 is trivial. However, I noticed that during deep searches, version 17.1 would run into bottlenecks in some positions and get stuck, sometimes taking days to complete the analysis, if it could finish at all. This was especially true in endgames, but it could also happen in the middlegame.

Stockfish is optimized for normal time controls, but deeper searches are where I have been having this problem.

I tested Stockfish 19 to see if it could analyze a whole game at 40 ply. It looked like it might take days, so I switched to Stockfish 17, which analyzed the same game at 40 ply in roughly an hour. I then tested version 17.1 against version 17 with a couple of games, and version 17.1 took almost twice as long.

It seems almost impossible to analyze a whole game at 40 ply with newer versions of Stockfish, so I was using 30 ply instead, although that sometimes failed as well. For one game, I had to drop the depth to 20 ply.

There doesn't seem to be much difference between Stockfish versions 17 and 19 in terms of analysis. Version 19 is reportedly a little stronger. According to ChatGPT, Stockfish 19 is only about 8–15 Elo stronger than Stockfish 17, based on a CCRL rating list at a fast time control. According to the people who make Stockfish, version 19 is 90 points stronger.

For the deep analysis I like to do, I am stuck with version 17.

I was thinking that maybe I am the only person who likes to do deeper analysis, so this has never been an issue. However, I had ChatGPT research this, and it found reports of other people having the same problem.

--
Best wishes,

John Coffey


P.S. Although I use an old version of Fritz to analyze with the Stockfish engine, I sometimes also use this free program: https://www.triplehappy.com/

Testing old chess computers through emulation

In the days before everyone had computers, if you wanted to play chess, your only option was to play with another person. In the late 1970's Fidelity introduced a series of electronic computer chess games. These early models played poorly, but I knew people who bought them just to be able to play and practice whenever they wanted. I managed to borrow a few of these so that I could get a feel for how well they played.

Although the early machines did not play well, things started to improve in the 1980s. There was a golden age of dedicated chess computers that went from 1983 to about 1993.  In 1984, I purchased the Novag Super Constellation electronic chess game for what I think was $200, which was quite a bit of money in 1984. The U.S. Chess Federation had given it a rating of 2018, which is better than at least 90% of all adult tournament players. Any rating between 2000 and 2199 is considered to be the skill level of "Expert" and a higher rating of 2200 is considered to be "Master."

Although I am currently rated 2016, at the time I bought the Novag Super Constellation I was rated just a little over 1700. In a few months, I would reach a rating of 1800 which is considered to be "Class A." Nevertheless, what I remember about the Novag Super Constellation is that it played better than me, which is surprising since it only contains an 8-bit processor running at just 4 MHZ. That is not very fast compared to modern 64-bit processors with multiple cores running at gigahertz speeds.

Over time, I bought a couple of better chess-playing computers and I have fond memories of practicing with all of them. I sold all these machines when I got a desktop computer in the mid-'90s, but I kind of regret it because they all were fun to play with it.

This became an issue when I was researching these old chess-playing computers where I saw many online claims that these computers were not as good as the ratings that had been assigned to them. For example, I saw the claim that the Novag Super Constellation was only about 1750 strength, and two other computers that I owned rated 2100 and 2265 were also claimed to be weaker than their advertised ratings. None of these claims match my experience, since all of the computers played better than I did.

I was so curious about this that I wanted to get my hands on one of the old chess computers, assuming that one can be found, however unlikely, and see how it compares to my current chess ability. Fortunately, I found software that allows me to emulate dozens of old chess computers on my Windows PC.

In my first game against the emulated Novag Super Constellation on level 1, the lowest level, I was able to win by only the slimmest of margins. I tried the same thing on the Fidelity Designer 2100, a slightly better machine, and I lost. I have no doubt that the other computer I owned, the stronger Fidelity Designer 2265, would stomp me like it used to when I played it 30 years ago. I will confirm this eventually.

So I tested a variety of chess computers with a somewhat difficult chess problem..

Most serious chess players have seen this problem already and know the answer. However, if they were not familiar with it, the solution might be difficult for them to find in a real game. There is the more direct solution of 6. Nxe5 Bxd1 7. Bxf7+ Ke7 8. Nd5# (checkmate). However, for a computer to see the solution it also has to see 6... Nxe5 7. Qxh5 Nxc4 8. Qb5+ c6 9. Qxc4. There is also 7... Nf6 8. Qe2 Nxc4 9. Qxc4. Either way, that is 7 half-moves deep, which is pretty deep for ancient chess computers to look.

Based upon my testing, this is how long various chess computers take to solve this chess problem...

#	Model	Year	Processor	Speed	ROM	Time	Depth	Nodes/S
1.	Fidelity Chess Challenger 10	1978	Z80	4 MHZ	4K	Fails
2.	Fidelity Chess Challenger 7	1979	Z80	4 MHZ	4K	12 hours
3.	Novag Savant	1981	Z80	6 MHZ	24K	12:40m
4.	Novag Savant II	1982	Z80	6 MHZ	32K	12:33m
5.	Novag Constellation	1983	6502	2 MHZ	16K	6:46m
6.	Constellation 3.6 ROM set 1	1984	6502	3.6 MHZ	16K	3:30m*
7.	Constellation 3.6 ROM set 2	1986	6502	3.6 MHZ	16K	3:33m*	5 ply
8.	Novag Super Constellation	1984	6502	4 MHZ	56K	2:10m	5 ply	500?
9.	Constellation Expert	1985	65C02	5 MHZ	64K	1:54m	5 ply
10.	Novag Forte A	1986	65C02	5 MHZ	64K	2:15m	5 ply	~1000
11.	Novag Forte B	1986	65C02	5 MHZ	64K	1:58m	5 ply	~1000
12.	Novag Super Forte	1987	65C02	5 MHZ	64K	1:13m	5 ply	~1350
13.	Novag Super Expert A	1987	65C02	5 MHZ	64K	1:00m	5 ply	~1100
14.	Novag Super Forte B	1989	65C02	5 MHZ	64K	30s	5 ply	~1400
15.	Novag Super Expert B	1989	65C02	5 MHZ	64K	19s	5 ply	~1375
16.	Novag Super Forte C	1990	65C02	5 MHZ	64K	11s	5 ply	~1500
17.	Novag Super Nova	1990	HD6301Y	4 MHZ	32K	10s	4 ply
18.	Novag Super Expert C	1990	65C02	5 MHZ	64K	6s	5 ply	~1050
19.	Novag Scorpio 68000	1990	68000	16 MHZ	98K	9s
20.	Novag Diablo 68000	1990	68000	16 MHZ	98K	9s
21.	Fidelity Excellence	1985	65C02	3 MHZ	16K	2:16m	5 ply
22.	Fidelity Excellence	1985	65C02	4 MHZ	16K	2:00m	5 ply
23.	Fidelity Designer Display 2000	1989	65C02	3 MHZ	32K	1:45m	5 ply	~81
24.	Fidelity Par Excellence	1986	65C02	5 MHZ	32K	1:22m	5 ply
25.	Fidelity Designer Display 2100	1988	65C02	6 MHZ	64K	54s	5 ply	~180
26.	Fidelity Designer Display 2265	1989	68000	16 MHZ	64K	5s	3 ply
27.	Fidelity Designer Display 2325	1991	68020	20 MHZ	64K	3s	4 ply
28.	Chessmaster NES	1990	6502	1.79 MHZ	48K	7:00m	5 ply
29.	Chessmaster Super Nintendo	1991	65816	3.58 MHZ	110K	4:43m	5 ply
30.	Chessmaster 2000 (DOS)	1986	?	?	NA	1:33m
31.	Chessmaster 3000 (DOS)	1991	?	?	NA	4s	4 ply
32.	Stockfish 14.1 2017-iMac	2022	i5	3.4 GHZ One Core	NA	<1s	<12 ply	~880,000

It is noteworthy that the Super Constellation solved the problem in roughly 2 minutes, which is within tournament time controls. I am disappointed in Chessmaster on the Super Nintendo because it failed to achieve this. It is running on a similar processor, and it is a port of Chessmaster 2000 written by Dave Kittinger, who also wrote the Super Constellation program!

* The second version of the Constellation 3.6 solves this problem on its top two tournament levels, but the first version moves too quickly to see the answer.  It can only solve the problem on its infinite level, even though it takes about the same amount of time to see the solution. The second ROM set is based upon the Novag Expert program.

Super Constellation game #1.

Jumble Puzzle

The answer to the puzzle is ...

"Make a move"

White to Play and Mate in 2

This is an interesting problem. I found it easy, but there are five different continuations.

https://chesslevel1.blogspot.com/2026/10/white-to-play-and-mate-in-2_0698473086.html

White to Play and Mate in 5

https://chesslevel2.blogspot.com/2026/10/white-to-play-and-mate-in-5.html

Semi-Tarrasch Defense

I had some good games in the USCF Online Rated 15+10 tournament last night, winning 3 out of 4. I didn't make any obvious blunders, although I misstepped against a higher-rated player in a very complicated Pirc-Lisitsin Gambit.

According to the Chess.com analysis feature, this was my most accurate game last night, with a 98.5% accuracy. It helped that my opponent left his king in the middle of the board, which I took advantage of. There were some tactics, but mostly I just had to defend when necessary and make accurate moves. For this game only, the analysis feature assigned me a rating of 2350 and my opponent 2300, both of which seem too generous.

https://onethousandpositionstochessmastery.blogspot.com/2026/10/semi-tarrasch-defense.html

This may not be the most instructive game on my chess lesson website, but I think that it is interesting enough.

Levi Rozman Book

Is chess going mainstream?