Houdini 3 Pro vs Deep Fritz 13

Edit: Thank you to Firebrand for the clarification that I can use Deep Fritz 13 as a shell and Houdini 3 Pro as an engine, rather than just buying the latter packaged from Chessbase.

I am trying to make a repertoire on the chess section of my website so it's important that I find the best possible moves, and I was wondering if anyone thought there was a redeeming value or some Niche that Fritz might have over Houdini (aside from the fact that they share the same "Let's Check" cloud.

I'm wondering if I should cut my losses and just preorder Houdini 3 now (which will be released Nov. 15). 

Additional Info: I built a PC with heavy calculating power, so I was thinking that "maybe" there would be an argument for running Deep Fritz 13 and Houdini 3 Pro in parallel. That way if they both come up with the same result than it's a good sign, and if there are discrepencies then I can run a deeper analysis. That notion is cheering me up a bit, anyone agree? Their engines are supposed to be quite different.

Edit: I guess I  answered my own question but I still would be curious to hear what others have to say. I don't have much money but I don't want to get cheap because I will be spending years updating my chess repertoire and so I want to have the best possible tools (also to avoid rechecking all lines later on).

Err, isn't Houdini 3 from Cruxis?

Wow thanks! I just checked the Cruxis site and I can buy it for $75 instead of $120!

Also, since they won't be on different copies of the same GUI, I can use Chessbase's "compare engine results" feature. Now I don't feel so bad for buying Fritz.

I was a little annoyed at the overlaps between Deep Fritz 13 and (Chessbase's) Houdini 3 Pro since they said they would be using the Fritz interface. Also, Chessbase forces to buy a bundle package with Sun Tzu's "Art of War" and I already own three copies of that book lol. 

PS. Here are the instructions to install the Houdini 3 download into Fritz/Chessbase: http://www.cruxis.com/chess/manual/index.html

Houdini 3 still has many bugs. I would wait before ordering, and anyway for my needs (and hardware) it has no real advantages over the free 1.5 version.

I guess I'm just outta the loop here, but I'm  having a hard time imagining anyone below IM/GM level who really needs the very latest and greatest chess engine available...esp since IM pfren is still getting by with the apparently hopelessly antiquated Houdini 1.5, judging by all the hoopla over Houdini 3

Personally I don't see how you could use an engine to prepare a repertoire. The engine will give the best possible move FOR ITSELF to play, not for you to play: it will not see things like giving you an easy position to play, a position with less opponent's responses, and especially a position that is more consistent with the rest of your knowledge.

An engine is only useful if you're a GM that wants to prepare some novelty, or to show you why some dubious move (outside of theory) fails. IMO, of course.

The idea is to prepare myself (or anyone using the repertoire) not to make any critical mistakes up to about 7 moves deep. 

If I only have to memorize the best decision for each direction, then this is somewhat possible. With white you can close the lines off pretty well, but this becomes more complicated with black, as there are twenty positions to memorize alone for the first game.

And no I'm not a gm, but I have been casually playing since I was three years old. Part of what this project was about is preparing myself to play blind. So that wherever I am, I can imagine a starter game and at least play "correctly" up to three or so move deep in any position. 

Another perk is in actually creating the repertoire, in which case I am thinking along with the computer. I will let it select the four top choices and then ponder them over myself while I set it in infinite analysis mode for hours or even a full day. Thinking along with the computer as I create enables me to internalize the openings (mixing some of the "why" with the "what"). 

Further, it gives me a much stronger capacity to envision a board algebraicly. When I read chess books, for example, I am almost completely unable to follow PGN notation without pictures. But as I create these decision trees, the board is become much more clear to me algebraicly. Especially in terms of which square is which color. 

"It will not see things like giving you an easy position to play, a position with less opponent's responses, and especially a position that is more consistent with the rest of your knowledge."

I am also cross-referencing the decisions with the win/loss statistics of the opening book here on chess.com (master games) as well as the with the win/loss statistics of Fritz (gm games). I don't see how any harm can come from memorizing which positions have the highest success rate.

But as a main preperation, I believe that computer analysis is more helpful then just digging through opening books (which provide many options). Computer analysis will help to find the best possible move. Don't forget the most valuable aspect of Chessbase programs which is the "Let's Check" cloud.

Every time someone analyzes a position with a new record of depth, their result becomes available to everyone else who has access to "Let's Check". So for any position, you can see what the running best choice is, or try to solve for a deeper one yourself. Memorizing (and trying to understand) these results strikes me as much more effective than memorizing classical lines. 

One last argument I'd like to throw out is I feel like I am able to contribute something (to the GMs to whom it would matter) by offering my computing power. It's just like folding@home where you can donate your computing power to a supercomputer for medical research. I have already created new-depth analysis for hundreds of positions, and am calculating some position at any given second. Makes me feel like I am contributing to collective consciousness. *shrugs*

"The idea is to prepare myself (or anyone using the repertoire) not to make any critical mistakes up to about 7 moves deep"

The critical flaw here is you can only 'think' 7 moves deep, but ultimately you make 1 move and the game shifts. Knowing that you enjoy the blitz style games you don't even think farther than 1 -3  moves most of the time. ( with exception of the opening )

You're joking, right? What are the "many bugs" you are talking about?

Thank you for not spreading FUD about a product you don't use.

You cannot form an opening repertoire using an engine, period and fullstop.

Engines are useful to strong players who already have their own repertoire, and want to blunderproof their occasional new ideas. They are also useful to correspondence players, which use the engines in a totally different way.

They already have their own repertoire, but how did they get it? And since "strong" players are qualified to use the latest technology but intermediates aren't, then I don't see how intermediates are expected to be able to close that gap until they are qualified enough.

Note that I will be building this repertoire for the rest of my life (I know it's not a small task). And since I'm doing it online, it will be easy to organize.

I am assuming the criticism here is because of how ambitious the scope is (being a universal repertoire). But to be hypothetical, let's say I was just trying to establish repertoire lines for Alekhine's opening. It seems that using the "Let's Check" cloud as well as checking each position on strong hardware and the world's highest-performing engine would be better than memorizing classical lines from books.

When they built the Rybka cluster, for example, lots of old classical lines suddenly needed to become revisited. Likewise Larry Kaufman in his 2012 repertoire, he used computer analysis.

I understand the argument that the best possible move for a supercomputer running Houdini would not necessarily be the best move for me (performance wise) since I am only an intermediate player, but Houdini's choice wouldn't be a bad move! Also, don't forget that I am cross-referencing win-lose statistics (choosing what works). If I see the computer pick something that goes against the stats, then I will let it run overnight to doublecheck.

And if anything it's a way to learn openings (first-hand and empirically). Like I said above, the main flaw of learning from books is that they just provide a ton of options.

You are hopelessly confused.

Both "methods" you described (analysing with an engine and memorizing book lines) are utterly useless- probably even more useless than Kaufman's repertoire book.

Studying carefully 3 or 4 games by Capablanca or Petrosian (even ones you have casually already studied) will be WAY more beneficial to you than one month's opening work with your beloved engine. And of course, nobody gives the slightest @#$% about the engine's "revisions" over the classiccal games.

You have clearly stated your opinion, which is appreciated especially because of your IM status. However, you have still failed to produce reasons.

But the questions still remains unanswered about why computer analysis is a poor way to determine opening lines. Why not learn your openings from engine analysis and/or the metadatabase which records the deepest historic engine analysis from any player? As opposed to using an opening book. Perhaps what you mean to say is that opening preparation in general is useless unless you are a pro, but that seems like a separate issue.

What I think might be going on with these criticisms is that this is new technology and is not understood. Here is a screenshot of what Let's Check does:

This to try and determine how black should respond to the Zukertort (1.Nf3). On the right you see the three all-time deepest solutions for this position. That second one was actually mine while I was leaving Deep Fritz 13 on overnight, but now I'm trying it again since computer analysis is struggling with this particular decision.

But that right there is valuable data. Sometimes you will see that all engines clearly agree on a particular position, but where they disagree or there are close calls, then it is worth consulting win/loss statistics.

So are you just criticizing the value of working on openings at all if you are <2000? Or is this about openings in general? Yes I know tactics are more important than openings, and yes I do make an effort to study classic games. I have been playing for 25 years and it doesn't seem fair to conclude that I am wasting my time for trying to step up my game, using some very new chess technology which will likely have a strong impact on the future of chess.

But I don't want this to be about me. Please offer reasons on why engines (or the work of others done on engines) are a poor way to learn openings—if openings are what you are trying to work on. 

Oh, boy.

Unfortunately all my available reasons are for sale today.

Cheapest one goes for 60.95 euro, which is the price for Houdini 3 pro plus one euro.

PM me if interested.

First Opening Theory is very important: If you don't start out well the game is lost immediately thus white is the favorite at the start.

Second: All so far are correct in many ways. You can't just use a computer engine for opening theory but it can be useful to a lesser player even with my inept abilities. At my level (You do need to know a couple of openings very well) both learning an opening you like to play from everyman or quality chess and combine that with fritz 13 can be very productive.

To answer the opening question what I found Houdini to be good at is the accuracy of the position towards the end of the game in comparison to fritz. My fritz 13 just can't seem to recongize how far behind or ahead he is in a 7-10 piece position. Houdini seems more accurate. In the endgame fritz is even worse analysiing simple endgame theory compared to houdini. On chessbase you see houdini beating everybody.

The engines are also important in analyzing where you went wrong on a move and what move would have been better and by how much. Thats where lets check works well in the opening lines and a good engine is important towards the end of the game. A lot of red lights in fritz shows I played a pretty poor game. Many green lights showed I played a pretty impressive game given my level of play. Sometimes I can see moves that fritz has missed and improperly analyzed. I don't know that I could do that in the endgame with houdini. A good product but lacks the graphics that Fritz has.

I call Houdini a hard core engine program.

Will you charge Sales Tax/VAT on top?

Yes I will, although chessplayers with thinking disabilities will be excluded.

I already gave you the reasons in my first post.

I think you might be confusing chess with tic-tac-toe.

You somehow think you can memorize the whole game tree complexity of chess and then in any situation be like "what would Houdini do?".

Opening theory must be UNDERSTOOD, not memorized. You'll never be able to understand (let alone memorize) what an engine tells you to do because it does not base its analysis on principles and ideas, but only in checking a million moves ahead.

Houdini 3 is a good elo bump over Houdini 1.5a. My testing backs up the CCRL rating list. That's with the 'glitches'.

There's a slight problem here, though: woodpushers cannot test engines.

Contrary to your belief, Houdart has already confirmed a major bug with non-popcnt compliant machines, which can result in wild evaluation fluctuations, as well as several lesser ones.

He may be wrong though... please, educate him.