Houdini 5 Dev - Stronger Than Ever.

The development version of Houdini 5 is currently taking part in the TCEC superfinal against Stockfish 8, it already has managed to send Komodo 10 out of the competition. It seems like the newer version of Houdini is clearly at least as strong as Stockfish 8 from the first few games it's played. You can follow the games of the super-final live here - http://tcec.chessdom.com/live.php

Here are a few excerpts from an interview with Robert Houdart (Houdini's creator). More details about Houdini 5 can be found here - http://www.chessdom.com/robert-houdart-on-houdini-5-dev-in-tcec/

• Chessdom: Is the performance of Houdini 5 dev matching your expectations? Have you found points where it still needs improvement?

  Robert Houdart: The performance is about as expected – competitive with the top of the rating lists. In the games I’ve seen Houdini has been playing very well in the opening and early middle game, but then sometimes failed to convert its advantage in the late middle game. We’ll definitely try to improve that aspect.

• Chessdom: Stockfish looks unstoppable in the last rounds. It is the only engine that defeated Houdini. You stated earlier that you “would not have Houdini 5 dev competing, if you do not believe there is a chance for first”. Has the situation changed in the past 32 rounds?

  Robert Houdart: From the games in Stage 3 you can see that the three top engines are close in strength – playing mostly draws between them and regularly beating the other engines. Any of the engines that qualify for the final will have a genuine shot at winning the tournament.

Interesting

100 games looks insane for human, but just another day in the office for engines. 

Interesting for sure. At least because it's about "recent engines".

What are you talking about? @MSC157

About schedule. You can see they have to play 100 games.

Stockfish in the early lead currently :)

You're right @MSC157. In fact Stockfish is currently leading the match by 2 games. Something tells me that it's going to be a walk in the park for Stockfish.

Sf is gaining 2 to 3 elo every week. I wonder how strong these engines would get say 10 years from now.

I recommend looking here https://chessprogramming.wikispaces.com

In future there will be engines that can reprogram themselves to improve.

But still gets destroyed by stockfish.

Thank you. Now I can start writing my own chess program :)

Ok, SF won another game. Looks like the winner is clear.

Houdini 5 is officially released too - http://www.cruxis.com/chess/houdini.htm

You can buy it if you want, it's just $60.

Has anyone successfully purchased Houdini 5 or 5.01 from the Houdini chess site (http://www.cruxis.com/chess/houdini.htm)?  I've tried twice and in neither case did it provide me a means to specify a method for delivery and in neither case was a link sent to me with a link where I could download it.  When I purchased previous versions this link was provided automatically.

Emails that I sent to Robert Houdard at houdini@cruxis.be were not answered.  I had to contact PayPal twice and arrange to get a refund.  Very annoying.  Hard to understand why a company that wants to sell a product makes it so difficult for customers to buy it.

If anyone has any suggestions I would greatly appreciate it since I would still like to buy the latest version IF I could be guaranteed that it would be delivered.

ChessKing's highest priced ($99?) version includes Houdini 5.  I have the Standard 2016 Version ($27.95 at Amazon) with Houdini 2 and I'm able to run the Houdini 2 64-bit engine with Lucas Chess, etc. as an ELO 3000 engine, though also since I have Stockfish 8 64-bit that Lucas tells me is ELO 3395, I tend to use Stockfish for analysis.

These engines are far more than I need for evaluating who is ahead in a position or what's the best move.

What I would like to see, and I think it will happen within 10 years, is for an engine to tell me, "Your Kingside attack was wrong because you didn't have enough excess power there. You should have moved your b3 Knight to the fine outpost at c5 and then stormed the center with your d, e, and f pawns."

Just a question, who are we to evaluate the strenght of 3000+ ELO players?

I love you chess.com

Actually it's not a matter of what is possible right now or not. I have no overview of such engines, if there is one fitting your needs or not. But from the viewpoint of developing such additional logic (not even neural networks are needed here) it's less difficult as you may think. The logic finding the patterns is nothing new and with an internal representation of the moves in a tree (because of the different outcomes one wants to know about), recalculating the nodes against each other ... then for any current move (represented by a position in the tree, Node A or B for example) the software could tell you exactly what you'd like to hear. Had you done B instead of A then ... because the software has attributes attached to each node which could tell you even more, any metric which can be calculated.

But imho such UI could be already present.

Save yourself money, Stockfish is free, and is a better engine too...

not only is Stockfish stronger, it just seems to run a whole lot better on my machine. Maybe it's better matched to my hardware, maybe it's something else... but i do know Houdini tends to use up more resources than Stockfish when used with the same settings...

Can someone tell me how to know which engine is stronger and more precise? I cant understand 3300 ELO games .