This is like debating with my youngest son. I admire his age appropriate bravado
I want to see a rematch of p8q vs Stockfish 11 using contempt 100, standard time controls and a decent laptop to run SF (less than 1 Mnps is pretty slow), see if you can draw that.
About time controls, I'm very bad at fast games. Anyone could beat me at fast games, we don't need stockfish to prove that, hahahah
I don't know what's "contempt 100".
It's true, 1Mnps is slow. That's the reason I let the engine run enough time to reach a decent number of nodes. It's explained at the begining of the forum, where I explain the conditions.
After letting the engine to think for a long time, it doesn't matter for how long I let it think, because the chosen move stabilizes after just some minutes thinking. So, in the end it's the equivalent as letting a strong computer to think.
When I analyze the game afterwards, I set the analysis with stockfish 11 at 6 secs, which is what I read in forums to be a very good realiable analysis time. And the analysis shows that stockfish was thinking correctly, not making blunders or anything wrong.
Here is the analysis settings:
I don't know what's "contempt 100".
It's relatively easy to draw a standard stockfish with limited hardware, even eaiser to beat it with handicap odds.
But by playing with the contempt value a human will always lose. I don't know the specifics, but it will avoid trades and go for lines that are objectively bad, but via "contempt" it expects you to play badly and blunder. In other words it plays more like a human would play you. It will set problems for you instead of naively pretending you're a super GM.
Isn't it nice though? To set a goal where you're guaranteed to fail?
In other words you risk nothing while pretending to be brave.
Why not set a goal of getting to 1400 blitz? Because that's not only possible to achieve, but also possible to fail? Heh.
"I'm trying to beat stockfish!"
That's nice kid. No one cares, and that's the truth.
I care, and I'm someone, so your last statement is false
For example you could probably draw Stockfish with knight odds a certain percentage of time.
But playing Nakamura at queen odds in a rapid game you might never win. That's because he knows you suck, and he wont give you an easy draw, unlike an engine.
Isn't it nice though? To set a goal where you're guaranteed to fail?
In other words you risk nothing while pretending to be brave.
Why not set a goal of getting to 1400 blitz? Because that's not only possible to achieve, but also possible to fail? Heh.
"I'm trying to beat stockfish!"
That's nice kid. No one cares, and that's the truth.
I care, and I'm someone, so your last statement is false
Of course you care, you made the topic.
The implication is no one cares but you.
I don't know what's "contempt 100".
It's relatively easy to draw a standard stockfish with limited hardware, even eaiser to beat it with handicap odds.
But by playing with the contempt value a human will always lose. I don't know the specifics, but it will avoid trades and go for lines that are objectively bad, but via "contempt" it expects you to play badly and blunder. In other words it plays more like a human would play you. It will set problems for you instead of naively pretending you're a super GM.
Well, if you call "limited hardware" to a processor Intel core i7, 2.8GHz (8GB RAM), what kind of super computer do you have at home? Deep Blue was much worst than your i5 laptop.
6 secs? So that's 6 million nodes? No that's far from a point in which SF will stabilize and even if it does stabilize, its PV will be used in the next moves to aid the new search, also, standard time controls aren't fast by any means.
Contempt is an UCI setting that determines how much SF will try to win, 100 is the highest possible setting while -100 is the lowest, the default is set at 24 and it's generally set to 0 or 10 when SF is pitted against neural networks, since you are a very weak oponent from SF's perspective, it's worth to increase the contempt.
I don't know what's "contempt 100".
It's relatively easy to draw a standard stockfish with limited hardware, even eaiser to beat it with handicap odds.
But by playing with the contempt value a human will always lose. I don't know the specifics, but it will avoid trades and go for lines that are objectively bad, but via "contempt" it expects you to play badly and blunder. In other words it plays more like a human would play you. It will set problems for you instead of naively pretending you're a super GM.
Well, if you call "limited hardware" to a processor Intel core i7, 2.8GHz (8GB RAM), what kind of super computer do you have at home? Deep Blue was much worst than your i5 laptop.
I dunno, I thought you were playing your cell phone or something.
It's far too easy to do things like limit its thinking time and give yourself takebacks. Posting a game where you supposedly drew stockfish isn't very interesting.
If you set up a game where it's not only recorded, but outside of your control, and THEN you draw or beat stockfish, then you'll generate some attention.
Wooow!! The "contempt" thing you said I like it!! How do I set that?
I don't know your software.
For me it's:
Add kibitzer -> advanced -> engine parameters -> contempt
Wooow!! The "contempt" thing you said I like it!! How do I set that?
IDK how to do so with the GUI you use but you can set* it up when launching SF from the console (which for your purposes is fine), you can google the commands to do so, IIRC they are in reddit.
*AFAIK the settings are NOT saved when they are changed from console, some GUIs can save the settings and set them as default.
6 secs? So that's 6 million nodes? No that's far from a point in which SF will stabilize and even if it does stabilize, its PV will be used in the next moves to aid the new search, also, standard time controls aren't fast by any means.
Contempt is an UCI setting that determines how much SF will try to win, 100 is the highest possible setting while -100 is the lowest, the default is set at 24 and it's generally set to 0 or 10 when SF is pitted against neural networks, since you are a very weak oponent from SF's perspective, it's worth to increase the contempt.
The 6 secs is the settings for the posteriori analysis after the match was played. While playing I let stockfish thinking for minutes, around 5 minutes per move.
Thank you guys for discussing about the "contempt". I'll search through the web more deeply what it is, what's the most competitive setting and how to adjust it.
I checked it out. Contempt value was 24.
I didn't touch any engine settings, I played with default settings. The only thing I did was to rise hash table size to 128 MB, a recommended value in forums to make it stronger. They said higher values were counterproductive.
I'm reading in forums that the higher the contempt value, the lower will be the strength of the engine:
"The default contempt setting for Stockfish 8 was zero, Stockfish 9 was 20 and for Stockfish 10 it is 24. Programmers did this because in some computer tournaments Stockfish was playing too many draws and by raising the contempt setting, the engine was able to score more points against weaker engines without losing too much objective strength against the stronger engines."
It looks like the best contempt-Strength equilibrium is 24
http://tartajubow.blogspot.com/2019/03/stockfish-10-contempt-settings.html
And the proof is in the post number. Serious topics don't get 100 posts.
Serious topics get less than 10
People don't come to the forums for anything meaningful. At least, they mostly come here for entertainment.