3Champs reloaded

[ouachita]

I would be quiet in your place regarding bragging about importance of TCEC results in any possible sense.

My observations are:

1. your attack on Martin is not persuasive or admirable;

2. it is infinitely easier and quicker to run a dozen 1+1 matches and report the results than it is to run even one LTC match; and,

3. there is no known direct correlation between STC (1+1, 2+2. 3+2, etc.) match results and LTC (i.e., >60+) results. One can only hypothesize about any correlation.

[M ANSARI]

Nice tournament ... it pretty much confirms TCEC in that Houdini 4 does not scale as well as Komodo and Stockfish. Stockfish is quite a monster on big hardware, and Komodo seems to gain as well as more time and bigger hardware covers for weakness in tactics when compared to the other big guns. Quite a big gap between SF and H4 ... SF really seems to have Houdini's number.

That's completely false.
SF and K advantage over H is only due to RH's decision to pursue large contempt (around 15cp) in order to benefit more from rating lists (and justify more Elo gain between H3 and H4) assuming in his business model that potential customers rely more on rating lists than on some "geeky" tournaments.
Problem here is that there are many SF/K fans who simply don't want to run tournaments with H with contempt 0 even though this is advised model for the analysis by RH (i.e. for LTC and correspondence).
Replay the same tournament as this one with H4 contempt 0 and H4 would comfortably win it.

I do agree that contempt has some effect, but I think it is very minuscule and really it seems that all LTC results have been quite similar. Playing through the games it is very obvious that Houdini is not losing because it is pushing too hard or avoiding drawish positions and losing because of it. SF just seems to be Houdini's nemesis ... it is simply totally outplaying Houdini and seems to search important lines deeper quicker (especially in the endgame). I am not sure what contempt was used in this tournament, but I do know that in TCEC RH setup his engine in a way where he thought it would perform best.

Would be interesting to know what H4 setup was in this tourney, was contempt default?

[Milos]

So RH lied to me and everyone else about this then. Just a FYI, contempt is disabled in the 9601 version that played in Stage 4. Contempt is not even an UCI option.

Do you have a source code?
Did you run a test?
What does not having UCI option has to do with having or not having actual contempt implemented??? Straw man as usual...
I analyzed some of the positions from TCEC played by H with H4 trying to reach the same depth as it was in TCEC, and guess what there was a constant offset in eval of 7 to 11cp. But anyway you don't even know what that means so why bother...

I have never claimed it to be statistically relevant because I know it isn't.
It is you who are saying that 100 games for H4 with contempt 0 in this 3Champs would be statistically relevant because then H4 would have won comfortably.

You didn't claim it to be statistically relevant but you claim contempt doesn't work based on 48 games?
You can't know what is my set that I based my claims on, I certainly don't base it on 100 games of this tournament.
So you just try to shift argument even though you are obviously wrong, blaming me for something that you are doing. That's very sound...

[Milos]

I do agree that contempt has some effect, but I think it is very minuscule and really it seems that all LTC results have been quite similar. Playing through the games it is very obvious that Houdini is not losing because it is pushing too hard or avoiding drawish positions and losing because of it. SF just seems to be Houdini's nemesis ... it is simply totally outplaying Houdini and seems to search important lines deeper quicker (especially in the endgame). I am not sure what contempt was used in this tournament, but I do know that in TCEC RH setup his engine in a way where he thought it would perform best.

Would be interesting to know what H4 setup was in this tourney, was contempt default?

Contempt effect is a statistical category you can't judge it by looking at the games, no one can. You have to actually have a relevant data set. At this forum you can't get it because nobody wants to run H with contempt 0. I can't imagine why, the only explanation that comes to my mind is that most of the ppl are fans of SF or K so they are kind of afraid of what might come out.

Regarding TCEC, it is easy to check what was actual contempt of H in TCEC (I would not listen to Martin since he has no clue, and Robert H. is just deflecting as usual when details about his engine are in question), but it is not relevant at all since TCEC is not relevant at all. SF is clearly stronger than Komodo, but Komodo won. What does that prove? Nothing except that difference between them is not large and that TCEC results (sited so many times here) have 0 (zero) relevance.

[Milos]

I would be quiet in your place regarding bragging about importance of TCEC results in any possible sense.

My observations are:

1. your attack on Martin is not persuasive or admirable;

Why is he some kind of God just because he runs a tournament???
When I think someone is wrong I tell him/her directly no matter how "important" or "admired" that person is.
You seams to be different dosing your reaction based on level of admiration you have for that person.

2. it is infinitely easier and quicker to run a dozen 1+1 matches and report the results than it is to run even one LTC match; and,

Why? One match like one in TCEC requires 5k$ configuration (I am sure many ppl here have much more powerful equipment overall) and consumes 1kWh of power for 1 month and requires one setup only. Why is this hard?
To run tens of thousands of 1+1 matches takes even more time on the same configuration and results are orders of magnitude statistically more relevant than TCEC. TCEC is nothing but fun and fancy coin tossing experiment.

3. there is no known direct correlation between STC (1+1, 2+2. 3+2, etc.) match results and LTC (i.e., >60+) results. One can only hypothesize about any correlation.

This is just ridiculous. What you call there is no direct correlation means actual correlation factor is less than 0.95 or what?
You know that correlation is a measurable quantity?
Did you measure it?
How do you know there is none, you saw it in your dream or what???

[Martin Thoresen]

What does not having UCI option has to do with having or not having actual contempt implemented??? Straw man as usual...

Try to think for a second:

1) No option for contempt
2) Author telling me it is disabled.

In this case 1 + 1 = 2. If either part of this is not true, then obviously 1 + 1 != 2.

That is why I tell you to ask RH if you mean it played with contempt so he can clarify it - whether he actually forgot to disable it or what the case really was.

You didn't claim it to be statistically relevant but you claim contempt doesn't work based on 48 games?

I have not claimed such a thing. It is you who claims that H4 would have won comfortably with contempt disabled, meaning contempt is more powerful than the error margin for 100 games.

So you just try to shift argument even though you are obviously wrong, blaming me for something that you are doing. That's very sound...

I don't even know what you are talking about here, to me this looks like some poor attempt of deciding to bring blame into a discussion where none needs to be.

[Milos]

1) No option for contempt
2) Author telling me it is disabled.

In this case 1 + 1 = 2. If either part of this is not true, then obviously 1 + 1 != 2.

You seams to be really weak with logic.
1) has nothing to do with actual implementation. Grasping this somehow misses you. Following your logic the fact that SF has no option for adjusting bishop pair bonus (and it had it in the past) means there is none (or that it was removed), right?
2) depends exclusively how much you trust the particular person. Btw. what he told you implementation is disabled or UCI option is disabled?
And related to that:

That is why I tell you to ask RH if you mean it played with contempt so he can clarify it - whether he actually forgot to disable it or what the case really was.

you are telling me to ask person that claims there was not a single byte of Robbo code in Houdini and trust what he says???
Basically you are telling me to trust a liar more than my own eyes?!!!
Give me a break, I wouldn't trust him even if he told me grass is green...

[Martin Thoresen]

1) has nothing to do with actual implementation. Grasping this seams somehow misses you. Following your logic the fact that SF has no option for adjusting bishop pair bonus means there is none, right?

You fail to see my point: these two factors were the reason I believed contempt was disabled. Has nothing to do with contempt still being part of the Houdini code or not.

2) depends exclusively how much you trust the particular person. Btw. what he told you implementation is disabled or UCI option is disabled?

None. He told me it was playing without contempt, not that it was disabled. The UCI option removal was something I saw myself.

And related to that:
you are telling me to ask person that claims there was not a single byte of Robbo code in Houdini and trust what he says???

Why not, you are obviously not getting what you seek from me.

[Adam Hair]

[MODERATION]
Please refrain from making posts that question the mental health of another member. In general, let's try to conduct this disagreement with some civility.

[Milos]

You fail to see my point: these two factors were the reason I believed contempt was disabled. Has nothing to do with contempt still being part of the Houdini code or not.

You based your belief on
1) irrelevant fact
2) thrust in word of someone whom I personally take for a liar
I understand what you believe, I just don't share your believes and have arguments why.
But that is not the problem. The problem is that you present your believes as some kind of universal truth. It seams to me that you think because you are some kind of VIP thanks to running TCEC tournament your arguments hold better. I don't subscribe to this point of view.