Playing strength of Houdini

What is the general and approximate playing strength (ELO) of the chess software "Houdini"

What's a google?

Google gives tons of information in mathematical scale about the ELO rating of Houdini which is complicated to understand.

Okay, thats pretty high :)

Is that a fact or a guess?

[sorry for my crappy english]

houdini is a chess software and that means its strength will depend in the hardware you’re using...
noone knows the engine's rating because super GMs don't report their results against them when they’re training...
even if they told us so we would need to know the HW they were using

pocket fritz 4 (hiarcs 13 running on a mobile phone) played in Copa Mercosur tournament 2009 and "achieved a performance rating 2898"
needless to say it was probably lucky. but even 2750 would be extremely strong for hiarcs 13 running on a mobile...

engine's rating by IPON-Rating list (single CPU)
rank-name-rating:

1.   Houdini 3     3081
66.  HIARCS 13.1   2728

3081 doesn't mean ANYTHING but is saying it's about 350 elo stronger than hiarcs 13

"Pocket Fritz 4 searches less than 20,000 positions per second."
so let say that mobile phone was about 10 times slower than a single CPU.
when you double the CPU you get about +50 elo

do the math... it sounds safe to say houdini should be 3300-3500 on a single CPU.
but engines are not programmed to speculate so there is a paranoid effect that will make the engine play unnecessary moves
(like anticipating clever attacks that will never happen) or not to play moves that will put the greatly weaker opponent in trouble...
basically "3000" engines vs "2200" ones drops the same score than "2700" engines vs "2200" ones.
this doesn’t exactly happen with humans vs humans

so basically we will never know if the GMs don’t give us a detailed information

Yes but that's a human and engine playing vs. an engine. From what you stated in the previous post, correct me if I'm wrong, you would be saying that if a human FM (or above) had several weeks to play the game vs houdini, they would win...?

nilshero, you seem so poorly informed that I wonder if I am feeding a troll.

"Disagree. More thinking time makes human stronger and stronger. Ok, you need a good motivation for human to dig deep. Eg the motivation to never again have to care for money in his life if he wins. That human outplays every computer if thinking time is big enough."

No. Completely and utterly wrong. You're pulling idealistic falsehoods out from nothing. It doesn't matter how much motivation you give the human, they will not be able to look as deeply or thoroughly as the computer. If I tell you to turn your computer off and then try and turn it back on by pressing the letter "q" again and again, and say that if you succeed you will never have to care for money or motivation again in your life, will that enable you to turn the computer on just by pressing "q"? No. Computers have tactical styles which humans cannot match without computer assistance. This is a modern-day fact.

If I give you a random position, will you be able to search it at 14 ply to a depth of 6? My Houdini can handle that. We're talking about 14^12=5.7*10^13 moves you have to look at. The computer will do it in less than an hour, but how long will a human take? The time limit we set is a week, but let's say the human can eliminate 1000 moves per second (and they can't - you might be able to eliminate 10-20 if you were Carlsen or Kasparov); then it will take the human ... 1,800 years. Now do you realize the difference in playing standard? I wonder when you will find me a human who has the motivation and endurance to calculate variations 24/7 for 1,800 years for a single move?

"wonder which GMs have said this. statement is not very precise"

Maybe I can make it more precise for you. In 2003, Garry Kasparov played a 6-game match with Deep Junior, the best program in the world at the time, and the final result was a 3-3 draw. Later the same year, Kasparov played another match, now against X3D Fritz - maybe the 2nd best in the world - and was lucky to get away with a draw. In 2004, a team of the world's best Grand-Masters, all in the World Top 5, including World No 1 Topalov, played against a team of computers Hydra, Fritz 8 and Deep Junior. The result? 8.5-3.5 to the computers, with positive individual scores against every single player. In early 2005, the same test was repeated; 8-4. In late 2005, Michael Adams, the best GM of the UK, was crushed 5.5-0.5 in a match. And in early 2006, Kramnik, the World Champion, was defeated 5.0-1.0 by Deep Fritz. I hope this shows you how much the computers improved by in just 2.5 years and helps you grasp how large the gap has become since.

"look what comps play in closed structures"

What time period is your information from? Rybka and Houdini - the topic of this thread - are both absolutely fine, superior to humans, in most closed positions.

"Yeah, but also human will have the necessary time to try out ideas and check them for holes. Also exponential growth of move possibilities is a bigger problem for comp than for human, especially in quiet positions"

Trying out ideas and checking them for holes is not nearly enough, as I showed above. And where on earth do you get the logic that exponential growth of move possibilities is a bigger problem for computers than humans? Computers can handle calculation far better, and always will. The tactical difference in ability is too big for humans' strategic superiority to make much of a difference; the playing style of Houdini is tactical, and it will just keep the game alive with tactical possibilities until you can't keep up.

"yes, human+comp is harder to win/draw against than just comp. But it is possible if the motivation+time is there. Those players you met may not have had the motivation, time or skills."

So there is someone out there with the motivation and time to meet the 1,800-year-per-move demand of what you're asking? Please point them out. It's a shame they're probably not sane.

I think Chessgod123 is mostly correct though. I don't think being a good chess player with a lot of motivation and time will be good enough to beat houdini. The time required might be more than a lifetime or the strength might require to be better than Carlsen. I don't know since I'm simply guessing here. I also think houdini is better at using a lot of time effectively than a human, but I could be wrong. Its search algorithm is really good (or so I've heard).

Deep blue is around 2700 strength I believe, and look at how well it did against Kasparov, who is arguably the best player ever.

Houdini is estimated at 3300 rating.

No human could possibly beat it unassisted in a match (much less an FM).

But after the game a human can go and get something nice to eat, maybe see a show...

Humans: 1

Computers: 0

You win this time, human. Beep boop.

Evidently you won't argue with me at all. Let me just say that my title comes from am impulsive youth, and my computing chess expertise is not particularly low ... certainly higher than yours, by the look of things. As to 3) and 4), I am perfectly sane, and I was answering your statements.

Sorry for the impoliteness. I do accept that I should have been nicer. It was late at night, sorry again on that particular account. But also, you are posting large posts and not that well-informed so please do actually read my post. You will get some statistics, and not have to listen to others calling you "frankly idiotic", because you'll stop making claims that go against evidence.

"Would a Strong Chessplayer + Houdini beat 10 Houdinis, sharing 'knowledge'?"

Why not try it yourself? If you see yourself as a decent Chess player, download Houdini and play, partnering Houdini on 1 computer at a certain processing power p, and letting Houdini alone play on the other computer at a processing power 2p. Let's see. This is only a factor of 2 so you might get away with a draw. Then try it at 10, as many times as it takes for you to feel the difference.