Fritz 18 Friend Mode

Sort:
Javolenu5

Has anyone tried the Friend mode in Fritz 18? Does it emulate human play realistically enough to be useful for training? I have Fritz 13, and have used Sparring mode which plays a game giving the user tactical opportunities along the way. Sparring mode was replaced with Friend mode in Fritz 15 onwards, and I was wondering if Friend mode provides tactical opportunities (as would arise in a human game) as opposed to the highly position perfect play followed by a random blunder that typically results from playing a detuned engine. If anyone has tried training against the weak engines in Lucas Chess, I would also be interested in your opinions about that.

EscherehcsE
Javolenu5 wrote:

Has anyone tried the Friend mode in Fritz 18? Does it emulate human play realistically enough to be useful for training? I have Fritz 13, and have used Sparring mode which plays a game giving the user tactical opportunities along the way. Sparring mode was replaced with Friend mode in Fritz 15 onwards, and I was wondering if Friend mode provides tactical opportunities (as would arise in a human game) as opposed to the highly position perfect play followed by a random blunder that typically results from playing a detuned engine. If anyone has tried training against the weak engines in Lucas Chess, I would also be interested in your opinions about that.

I can't give an opinion on Fritz, since the only Fritz GUI/engine I ever owned was Fritz 8.

Regarding the training issue, I think it's less important whether you're playing a person or an engine; It's more important that you're studying tactics, strategy, and general principles, and that you're using your games to see if you've learned what you've studied. And the only way that you know whether you're learning is to analyze your games after you've played them, regardless of whether it's a game against a person or engine. (Most people recommend that you first analyze a game by yourself without an engine, then a second time with an engine.)

Of course, my first preference is to play against people, but if I'm playing against an engine, I *do* prefer to play against an engine that plays human-like moves.

You gave your opinion that dumbed-down engines have "highly position perfect play followed by a random blunder that typically results from playing a detuned engine". Maybe it's possible that some engines do that, but the engines I enjoy playing against definitely *don't* do that. I like to play against the commercial engines Delfi Trainer 5.4 (no longer available except for the demo version), HIARCS, and to a lesser extent, Shredder. I also like to play against the freeware Maia/Lc0 engine. Imho, these engines have the most natural, human-like moves of any engines I've tried. And, these engines, on their weakest levels, don't play "position perfect moves", because they don't use any CPU resources. If you use the Windows Task Manager to watch the CPU power that the engine uses, these engines only look a couple of plies deep and use ZERO CPU power. Therefore, they can't be calculating anything of significance. (The Maia/Lc0 engine doesn't calculate anything; It only uses the Maia neural networks, which were created from games of real people.)

For information, the Lucas Chess GUI for Windows and Linux contains the Maia/Lc0 bots, and the Lucas Chess GUI for Windows also contains the Delfi Trainer 5.4 demo engine (1000 ELO).

 

Javolenu5
EscherehcsE wrote:
Javolenu5 wrote:

Has anyone tried the Friend mode in Fritz 18? Does it emulate human play realistically enough to be useful for training? I have Fritz 13, and have used Sparring mode which plays a game giving the user tactical opportunities along the way. Sparring mode was replaced with Friend mode in Fritz 15 onwards, and I was wondering if Friend mode provides tactical opportunities (as would arise in a human game) as opposed to the highly position perfect play followed by a random blunder that typically results from playing a detuned engine. If anyone has tried training against the weak engines in Lucas Chess, I would also be interested in your opinions about that.

I can't give an opinion on Fritz, since the only Fritz GUI/engine I ever owned was Fritz 8.

Regarding the training issue, I think it's less important whether you're playing a person or an engine; It's more important that you're studying tactics, strategy, and general principles, and that you're using your games to see if you've learned what you've studied. And the only way that you know whether you're learning is to analyze your games after you've played them, regardless of whether it's a game against a person or engine. (Most people recommend that you first analyze a game by yourself without an engine, then a second time with an engine.)

Of course, my first preference is to play against people, but if I'm playing against an engine, I *do* prefer to play against an engine that plays human-like moves.

You gave your opinion that dumbed-down engines have "highly position perfect play followed by a random blunder that typically results from playing a detuned engine". Maybe it's possible that some engines do that, but the engines I enjoy playing against definitely *don't* do that. I like to play against the commercial engines Delfi Trainer 5.4 (no longer available except for the demo version), HIARCS, and to a lesser extent, Shredder. I also like to play against the freeware Maia/Lc0 engine. Imho, these engines have the most natural, human-like moves of any engines I've tried. And, these engines, on their weakest levels, don't play "position perfect moves", because they don't use any CPU resources. If you use the Windows Task Manager to watch the CPU power that the engine uses, these engines only look a couple of plies deep and use ZERO CPU power. Therefore, they can't be calculating anything of significance. (The Maia/Lc0 engine doesn't calculate anything; It only uses the Maia neural networks, which were created from games of real people.)

For information, the Lucas Chess GUI for Windows and Linux contains the Maia/Lc0 bots, and the Lucas Chess GUI for Windows also contains the Delfi Trainer 5.4 demo engine (1000 ELO).

 

Thank you for taking the time to give me an in depth answer, it is much appreciated! I read on the HIARCS forum that it plays the most human like, and am considering buying it.  Which version(s) have you used? I am not sure whether to buy HIARCS Chess Explorer with the HIARCS 14 engine, or HIARCS Chess Explorer Pro with the HIARCS 15 engine. Do you use HIARCS to analyse your games? If it is better than a free engine for analysis, I will buy Deep HIARCS Chess Explorer Pro, as I would like a multi-core engine to save time. How would you say Maia compares to HIARCS? I have played the Maia bots on the other chess website, and at 1500 it was beating me with an evaluation of +8 when it repeated moves for a draw. I wonder if the bot on that website is less realistic than the version in Lucas Chess. 

EscherehcsE
Javolenu5 wrote:
EscherehcsE wrote:
Javolenu5 wrote:

Has anyone tried the Friend mode in Fritz 18? Does it emulate human play realistically enough to be useful for training? I have Fritz 13, and have used Sparring mode which plays a game giving the user tactical opportunities along the way. Sparring mode was replaced with Friend mode in Fritz 15 onwards, and I was wondering if Friend mode provides tactical opportunities (as would arise in a human game) as opposed to the highly position perfect play followed by a random blunder that typically results from playing a detuned engine. If anyone has tried training against the weak engines in Lucas Chess, I would also be interested in your opinions about that.

I can't give an opinion on Fritz, since the only Fritz GUI/engine I ever owned was Fritz 8.

Regarding the training issue, I think it's less important whether you're playing a person or an engine; It's more important that you're studying tactics, strategy, and general principles, and that you're using your games to see if you've learned what you've studied. And the only way that you know whether you're learning is to analyze your games after you've played them, regardless of whether it's a game against a person or engine. (Most people recommend that you first analyze a game by yourself without an engine, then a second time with an engine.)

Of course, my first preference is to play against people, but if I'm playing against an engine, I *do* prefer to play against an engine that plays human-like moves.

You gave your opinion that dumbed-down engines have "highly position perfect play followed by a random blunder that typically results from playing a detuned engine". Maybe it's possible that some engines do that, but the engines I enjoy playing against definitely *don't* do that. I like to play against the commercial engines Delfi Trainer 5.4 (no longer available except for the demo version), HIARCS, and to a lesser extent, Shredder. I also like to play against the freeware Maia/Lc0 engine. Imho, these engines have the most natural, human-like moves of any engines I've tried. And, these engines, on their weakest levels, don't play "position perfect moves", because they don't use any CPU resources. If you use the Windows Task Manager to watch the CPU power that the engine uses, these engines only look a couple of plies deep and use ZERO CPU power. Therefore, they can't be calculating anything of significance. (The Maia/Lc0 engine doesn't calculate anything; It only uses the Maia neural networks, which were created from games of real people.)

For information, the Lucas Chess GUI for Windows and Linux contains the Maia/Lc0 bots, and the Lucas Chess GUI for Windows also contains the Delfi Trainer 5.4 demo engine (1000 ELO).

 

Thank you for taking the time to give me an in depth answer, it is much appreciated! I read on the HIARCS forum that it plays the most human like, and am considering buying it.  Which version(s) have you used? I am not sure whether to buy HIARCS Chess Explorer with the HIARCS 14 engine, or HIARCS Chess Explorer Pro with the HIARCS 15 engine. Do you use HIARCS to analyse your games? If it is better than a free engine for analysis, I will buy Deep HIARCS Chess Explorer Pro, as I would like a multi-core engine to save time. How would you say Maia compares to HIARCS? I have played the Maia bots on the other chess website, and at 1500 it was beating me with an evaluation of +8 when it repeated moves for a draw. I wonder if the bot on that website is less realistic than the version in Lucas Chess. 

I bought the HIARCS 13.2 engine only; I've never used the HIARCS GUIs. HIARCS 13.2 is fairly strong at its weakest level; At its 750 ELO setting, it's a bit stronger than Delfi Trainer 5.4 set to 1000 ELO. Unfortunately, I've never used the HIARCS 14 or 15 engines, so I don't know how they play.

For analyzing, I really just use any strong engine, since I'm only trying to spot my blunders and missed two or three mover tactics.

How does Maia compare to HIARCS? Maia blunders at times, just like people do. HIARCS seems to almost never blunder outright; Instead, HIARCS' mistakes are more subtle, occasionally not seeing tactics, etc. One disadvantage of Maia is that for any particular position, it will always play the same move - There is no variability. However, the Maia programmers have stated that they intend to fix this in a future version.

In case you'd like to look at some PGN games of HIARCS ELO 750 versus a few other dumbed-down engines, here's a PGN of an engine tournament I ran in Arena. The weird thing about this tournament is that Maia 1100 couldn't seem to win any games against HIARCS 750, but it had better luck with the other dumbed-down engines.

Maia.pgn download link:
https://www.sendspace.com/file/n82q34

(Link is good for at least 30 days.)

EscherehcsE

Oh, and the Maia bots on Lichess should be virtually identical to the Maia bots in Lucas Chess. (It's the same Maia neural weights.)

ChessElk

The Maia bot is great. 👍👍👍

I installed this engine on Fritz 18 GUI. Runs fine.

Javolenu5
EscherehcsE wrote:
Javolenu5 wrote:
EscherehcsE wrote:
Javolenu5 wrote:

Has anyone tried the Friend mode in Fritz 18? Does it emulate human play realistically enough to be useful for training? I have Fritz 13, and have used Sparring mode which plays a game giving the user tactical opportunities along the way. Sparring mode was replaced with Friend mode in Fritz 15 onwards, and I was wondering if Friend mode provides tactical opportunities (as would arise in a human game) as opposed to the highly position perfect play followed by a random blunder that typically results from playing a detuned engine. If anyone has tried training against the weak engines in Lucas Chess, I would also be interested in your opinions about that.

I can't give an opinion on Fritz, since the only Fritz GUI/engine I ever owned was Fritz 8.

Regarding the training issue, I think it's less important whether you're playing a person or an engine; It's more important that you're studying tactics, strategy, and general principles, and that you're using your games to see if you've learned what you've studied. And the only way that you know whether you're learning is to analyze your games after you've played them, regardless of whether it's a game against a person or engine. (Most people recommend that you first analyze a game by yourself without an engine, then a second time with an engine.)

Of course, my first preference is to play against people, but if I'm playing against an engine, I *do* prefer to play against an engine that plays human-like moves.

You gave your opinion that dumbed-down engines have "highly position perfect play followed by a random blunder that typically results from playing a detuned engine". Maybe it's possible that some engines do that, but the engines I enjoy playing against definitely *don't* do that. I like to play against the commercial engines Delfi Trainer 5.4 (no longer available except for the demo version), HIARCS, and to a lesser extent, Shredder. I also like to play against the freeware Maia/Lc0 engine. Imho, these engines have the most natural, human-like moves of any engines I've tried. And, these engines, on their weakest levels, don't play "position perfect moves", because they don't use any CPU resources. If you use the Windows Task Manager to watch the CPU power that the engine uses, these engines only look a couple of plies deep and use ZERO CPU power. Therefore, they can't be calculating anything of significance. (The Maia/Lc0 engine doesn't calculate anything; It only uses the Maia neural networks, which were created from games of real people.)

For information, the Lucas Chess GUI for Windows and Linux contains the Maia/Lc0 bots, and the Lucas Chess GUI for Windows also contains the Delfi Trainer 5.4 demo engine (1000 ELO).

 

Thank you for taking the time to give me an in depth answer, it is much appreciated! I read on the HIARCS forum that it plays the most human like, and am considering buying it.  Which version(s) have you used? I am not sure whether to buy HIARCS Chess Explorer with the HIARCS 14 engine, or HIARCS Chess Explorer Pro with the HIARCS 15 engine. Do you use HIARCS to analyse your games? If it is better than a free engine for analysis, I will buy Deep HIARCS Chess Explorer Pro, as I would like a multi-core engine to save time. How would you say Maia compares to HIARCS? I have played the Maia bots on the other chess website, and at 1500 it was beating me with an evaluation of +8 when it repeated moves for a draw. I wonder if the bot on that website is less realistic than the version in Lucas Chess. 

I bought the HIARCS 13.2 engine only; I've never used the HIARCS GUIs. HIARCS 13.2 is fairly strong at its weakest level; At its 750 ELO setting, it's a bit stronger than Delfi Trainer 5.4 set to 1000 ELO. Unfortunately, I've never used the HIARCS 14 or 15 engines, so I don't know how they play.

For analyzing, I really just use any strong engine, since I'm only trying to spot my blunders and missed two or three mover tactics.

How does Maia compare to HIARCS? Maia blunders at times, just like people do. HIARCS seems to almost never blunder outright; Instead, HIARCS' mistakes are more subtle, occasionally not seeing tactics, etc. One disadvantage of Maia is that for any particular position, it will always play the same move - There is no variability. However, the Maia programmers have stated that they intend to fix this in a future version.

In case you'd like to look at some PGN games of HIARCS ELO 750 versus a few other dumbed-down engines, here's a PGN of an engine tournament I ran in Arena. The weird thing about this tournament is that Maia 1100 couldn't seem to win any games against HIARCS 750, but it had better luck with the other dumbed-down engines.

Maia.pgn download link:
https://www.sendspace.com/file/n82q34

(Link is good for at least 30 days.)

Hi. Sorry for the late reply! Thank you for making the tournament, I am working my way through looking at the games. What software did you use to run the tournament, and how did you get it to weaken the engines? I

EscherehcsE
Javolenu5 wrote:

Hi. Sorry for the late reply! Thank you for making the tournament, I am working my way through looking at the games. What software did you use to run the tournament, and how did you get it to weaken the engines? I

I just used the Arena GUI to run the tournament. It's maybe not the best Chess GUI around, but it's pretty good at running engine tournaments.

http://www.playwitharena.de/

As is the case with most Chess GUIs, if the engine supports weakened settings, the GUI will allow you to set the weak levels of the engines.

foobarred1

Fritz friend mode is unplayable. It takes forever to finish a game. The engine itself sucks to play against. This review nails it: https://www.chessengeria.eu/post/fritz-18-review

Apparently easy mode is pretty good according to the review, but I haven’t tried it myself.

In use HIARCs instead. The engine is much better to play against.