Best program for automatic game analysis / annotation

After playing a game, I like to run it through a chess program to spot blunders and mistakes.

What do people think is the best program (GUI) for doing this?  I tried the demo for the Shredder Classic GUI, but found it clumsy for stepping through the game after the analysis was finished, and disliked a few other aspects of it.

I'm using ScidVsPC currently, but am finding myself running into the same minor nuisances over and over.  Plus, it analyses from start to end, instead of going from the end of the game backwards to the start, which means less effective use of information in hash tables.

Is there anything out there that's better?  I am tempted by the Hiarcs Chess Explorer, but they have no demo, so I would have to purchase that on faith and hope it doesn't come with any annoying quirks.

(For the actual AI engine, I've stuck to Houdini 1.5a so far - is there a good reason to use one of the other free programs instead?) 

the automatic analysis feature should be avoided really. you will get a lot more benefit going through your games yourself and only using the engine to spot any tactical shots you missed.

I have only recently returned to chess after retirement and I tried several UCI's which are so much nicer than setting up a board, constantly setting and resetting as I did 'back in the day'.  My freeware preference is Arena 3.0 on my desktop computer, and my commercial preference is Fritz 13.

I agree that entering your serious games into a UCI and re-playing them is one of the best ways to improve, but I think it works best if you slow down and study a position, noting alternate lines on a notepad, then turn the engine on to see what the computer comes up with and the thread related to that.  Although Fritz has teaching and suggestion functions, I think the real value comes from the positional study prior to unleashing the engine.  You get very little help from just racing through a game again.

great advice sir

That is probably true, but in practice, unlikely to happen.  The goal is to be able to have a very quick look over the game - if I were to spend significant time (of my own - if the computer takes time to produce annotation, that is fine) on it, I'd rather use that time to play another game.

If I ever play 40 moves in 120 minutes tournament games over-the-board, that might be a different matter... but with casual 15|10 games online, I definitely don't want to spend as long analysing as it took to play the actual game.

Then again, unlike some (most?) people, I am not viewing improvement as a big goal - my main aim is to have fun, with improvement a distant second to that.

So, to reiterate, what program is best to produce some automatic annotations that I can then quickly look through before moving on to the next game?

I would look at programs such as Fritz 13 which has built-in annotation capability and a teaching mode.  I suspect there would be few if any freeware programs capable of doing that type of analysis.  I don't know of any program that would start at the end and work backwards.

I just did a video overview of the analysis features of a program called Lucas Chess (which uses standard chess engines for analysis).  It has a unique presentation that you might like (plus it has an option to start at the end, as well as lots of other options).  Check it out and see what you think:

http://www.youtube.com/watch?v=_B3ISwYMIkc

Houdini. 

This post is what is actually utterly useless. You claim that automatic analysis is a bad thing, but what actually happens is you analyze more games in less time, and get what you want out of them. Manual analysis is exceptional and takes a while. Most people are looking to get a quick overview. 

I disagree with both of you :)  Or partially agree with both of you, depending upon how one views things.  I think neither automatic analysis, nor Pfren's post, are useless.

I think Pfren definitely has a point, in that in order to learn the most out of a seriously played game, one should analyse by hand, and only use the engine as assistance.  Same reason as when replaying a high-level game, I try to understand the moves with my own mind primarily... and also the reason why a well-commented game annotated by a human is much better than for learning from than a solely computer-commented one.

Still, the games I generally play aren't serious enough to deserve such effort - and time efficiency, as fianchette123 points out, becomes more important.  I think both manual and automatic analysis have their use and place.

Oh, and regarding analysing from end backwards to start - that is the way the Shredder GUI does it.  As for engine, I've been using Houdini so far... but I think at my level, pretty much any strong engine will do it - using a GUI that presents the results well and makes it easy to browse both the main game, the computer variations, and different variations of my own, becomes more important.

I use the analysis engine provided here on this site and find it very helpful. Im not a master just a mister but i thot i wud just throw that in. Also I have Lucas chess and will be sure to check out the vid mentioned above by MJ4H. Anyone who uses both letters and numbers to spell his name cant be all bad!

At lower rating/skill levels is that absolutely true? I don't usually have a much stronger opponent around to go over games with and if I didn't see something in the game, especially at longer time controls, I'm not likely to see anything different in the post-game analysis.

There are some exceptions to that, such as positional errors/blunders that quickly lead to a material gain for my opponent. Analyzing those positions manually are likely to result in some better lines in hind sight.

What I have been doing, is using the auto-analysis feature to just output the move score, not any variations, when it meets certain criteria (more details on what I actually do is in my most recent blog post) and then use that information to actual look at those positions.

Basically, I let the stronger player (the engine) tell me where the mistakes were made, by both sides, and analyze that to try and find the better variations. If I'm unable to find that after a period of time, a longer amount of time for a larger error, then I will let the engine show me what it thought was better at that point and I try to figure out why it was actually better.

As I improve, I'm sure this method will be less useful but for now I think it gives the most benefit for the time spent, helping me try and pinpoint errors in my thinking.

And to answer the OP question, I can't say what combo is best. I have been using SCID vs PC with Houdini most recently and while it has some quirks in the way it works (doesn't seem to catch all the blunders and doesn't flag faster mates) it is much better than SCID in how verbose it is (too many variations that have to be purged and scoring all moves).

I use Fritz 11 with Houdini 1.5 as engine. For 15/10 online games, I suggest you use blunder check at 5 second/ move and threshold at 0.8. If you want a deeper analysis, use full analysis at 30 second/ move and threshold at 0.3.

I am waiting for a engine which explain the failure.

A Fritz with a 3500 Elo I don't need to buy - has no sense for a weak player.

I would like an engine to tell me what the name is of the tactic I should've executed OR the strategy I should follow. It should then explain me why this tactic/strategy should be executed based on board characteristics. These characteristics should be in natural language, not just telling me Bf3 because Kg2 and Qh1.

Then and only then, I would use automatic analysis!

Engines nowadays only show missed tactics with the sequences of moves missed. Not bad! But it could be better as explained above.

In the analysis output of the analysis engine on this site question marks appear after a bad moves. These are the moves that should be looked at and the variations given for these moves are the ones that I pay the most attention to. I believe this does help me since I do not have a master level coach to look at my games.

Before I moved to my current method, I tried just going through the game again, trying to find improvements in my play. I won't say I never found improved moves but I rarely found most of the errors in the game. Spending an hour or more on a game in analysis, after I had already spent a hour to two in some cases (OTB games) and coming away with little improvement, pushed me to try the method I described.

I'm not going to say that manual analysis isn't useful and isn't something that needs to be done by players of lower skill levels. My only issue with what you said is that auto analysis is "completely useless." I think that it can be incredibly useful, if you don't allow it to just feed you lines. If you do that, then I agree, it isn't very useful.

I know I'm using my method to find some of my weaknesses in thought processes and in tactical vision. As I find good positions, I'm marking them for inclusion in a database that I'll use to drill the places where I made major mistakes and some of the positions where I actually got it right.

I probably should do manual analysis on some of my games, prior to running it through the automatic process (that I mentioned) and start trying to exercise that part of my brain some more. Though, with limited time for chess, my current method is going to take up the lion's share of analysis time and I really believe it will benefit my play.

Especially since I have at least 500 games that I still need to analyse, though I'm trying to keep up with my current ones and work on the backlog as time permits.

Hi guys,

I'd like to add my 2 cents to this topic :)

I stopped running full automatic analysis on Fritz and instead i do as follows:

I add the game i played move by move into Fritz with just the "explain all moves" and i also write it down on my chess journal.

When Fritz suggests that the move i played it's really weak i add his variation and on the paper i try to explain in words the whys.

I find it more useful and I have also more fun.

That sounds really interesting.  

I will try it.  

An interesting thread.  As a former teacher, I know that different people learn in different ways - visual, auditory, tactile, etc.  I also know it depends on the attention span of the student as to how long (s)he can stay "on task" before losing concentration and becoming bored, so that factors in as to what works best.  My personal method of improving means spending time re-playing the game without time pressure.

As Martin_Stahl outlines above, using a stronger player or chess engine can give you alternate variations that you may have completely missed.  I use the UCI to input the moves quickly, generating a PGN of the game from my scoresheet with no analysis.  Then I set up a chessboard on a small table along side of my computer and replay the game on the board, following along on the computer until I reach a position of question.  I study the board while the computer is turned on to analyze with no time limit.  When I have spent what is a reasonable amount of time over the board, I then turn to check the computer analysis and variation which I may or may not play through on-screen, depending on its inherent interest.  The board stays set up according to the original game position and I can return the on-screen position to match the board with a single click of the mouse.  For me, for some reason, I seem to see things differently (and better?) using a board than I do on-screen.  Perhaps this will change with more practice, but it is a consideration.  If on the other hand, your play is online and on-screen, then you may not profit from board analysis.

When I played tournament chess years ago, I used to make notes in the margin on my scorepad of alternative moves that looked interesting that I passed up in favor of the move actually made.  I even had a personal code I used that the opponent looking at my pad wouldn't be able to use to discern my thought process.  Then, after the game, I would play the game over and analyze on a board.  Admittedly, this was in the days of Chessmaster and Sargon II.  Today's UCI's are wonderful for analysis.

You might consider looking at one of the Chessbase products such as Fritz 13 which I just purchased.  In addition to the common UCI features such as automatic move recording and analysis of "best move", it also contains features such as suggesting a move, explanation of all possible moves, brief strategic description of the position, and various other tutoring functions.  These can be turned on/off of depending on what you're trying to learn.  I think these features are much stronger in the commercial versions I looked at as compared to the freeware.  Ultimately, however, I just use the power of the engine to check my analysis and look for the best move(s) in a position; and any good combination such as SCID vs. PC and Houdini 1.5a or Fritz 13 allow you to install and use multiple engines simultaneously and compare results. 

Also of value is playing over games from players of different styles.  This can give you a look into differences in positional vs. attack lines.  I play over annotated master games and compare the annotations to the chess engine in tricky positions.  The level of annotation varies greatly from very verbose basic theory to subtle alternative lines that may be above my level of understanding, so find several authors who write at your level and seek them out.  Remember that today's chess engines are stronger than any author writing annotations, and there may well be flaws in their understanding and analysis and/or in the engine (though not likely).

From your initial post, you seem to be playing quick games online rather than tournament play.  I think there are two ways to improve under those conditions:  First, I would adapt several openings for white and black and stick with lines you know to avoid blunders.  If the Ruy (for example) becomes boring after a while, learn one new opening or a variation you haven't played.  Learn what the initial tactical intent of the opening is and play over master games to see how that is implemented.

Second, play over lots and lots of games including your own.  If you don't spend a lot of time in real time analysis, you need to learn positions and combinations that recur and learn to recognize threats quickly.

Ultimately, however, it is not the power of the engine, as virtually all chess engines today are stronger than we are and provide better moves in most positions (endgames being a potential exception depending on the tablebase being used).  The value of a commercial UCI such as Fritz is that it can be tailored to tutor according to your needs and contains features mostly (but not completely) lacking in freeware UCI's which are predominately PGN readers and positional analysis programs.  But they cost money and may not do what you want, so be careful.