What is the right way to use Fritz/Engines/databases to look at your games?

I've seen a lot of "but Fritz says" posts when people describe their games ... wanted to get some really good expert opinions on what should be the "correct" way for a constant improver to use Engines to go over one's games.

I'd like to open up my own method for critique:

Step 1. Play a slow + serious game against a stronger opponent ... hopefully get crushed or draw a won game. Not that I want to lose/draw,  but there's less skin invested in the game if I had won it. I'd take it a bit less seriously ...

Step 2. Go over the game on my own and identify critical positions and moments where the game swung from = to +- or -+. Remember ... DON'T TURN ON THE ENGINES immediately!!!

Step 3. Make my own notes (easier to do with a chess database)

Step 4. Start my engine/s and first do a tactical scan of the entire game to see if there were any gotchas. Probably won't let Rybka/Fritz/Engine sit on a position for more than 3-5 seconds a move here.

Step 5. Assuming I don't feel bad already at the tactical mistakes and inaccuracies, I then fire open the Opening Book pane and see if I deviated from theory or if my opponent did. I mark this point out in my notes.

Step 6. If it was me who deviated from theory, I then indicate what the proper line was.  I'll try really hard to remember this for future games. I'm certainly no Master-level player who can even claim to find novelties, so this is merely me trying to crawl before I learn to walk.

Step 7. So we've got tactics and opening deviations covered. I then try to walk through the game a lot slower, asking my engine to now list the top 3 or 5 lines in every position. I let it sit on each of the move until I reach 14-15 ply in ALL of my lines. I then see if the move I played (or my opponent played) was within the top 3/5 lines. If it wasn't even in the top 5 lines, then I SERIOUSLY misread the position. I save the position (FEN code would do) and discuss this with a stronger player offline (at the club, maybe a coach?)

Step 8. If I played a top 3 or top 5 move BUT

 a) If the "delta" / difference in evaluation scores between MY move and the best engine move was more than 0.4 (4/10ths of a pawn)  (I think this is decent for my level (USCF 1700) )

AND

 b) The evaluation of the position before my move is equal or not good for me (If I am winning, I'll ignore the quality of my moves as long as it is in the top 5 and all lines are still +- for me)

THEN I look really hard at the position and see if I considered the "better" move as one of my candidates. If I suddenly see the light, I record this position as one worth revising. Evidently, I'm seeing the better moves during the post-mortem but not while playing. I need to perhaps reinforce my knowledge of these positions better.

Step 9. If I still can't figure out why the engine says one move is better, I will save the position and take it to a stronger player / coach and ask him why this might be so. Note that I will NEVER blindly trust an engine unless it is a tactical sequence.

Step 10. Rinse and repeat for each game. This whole exercise takes easily an hour or more even for a 25-30 move game.

Step 11. If I am really curious, I will try to query the megabase and see what has been played historically in this opening line and if anybody played the deviations that occurred.

Please feel free to comment and suggest improvements to this approach! :)

Thanks in advance,

Shivsky

If I practiced this regimen for even a fraction of my serious game losses, I'd be 300-to-400 rating points stronger. Good analysis method, if one can stick to it.

If I practised this regemin for even a fraction of my games, I would be bored stiff, using up a lot of time and I doubt if it would improve my chess very much. If it whooshed me up to master level I might be tempted. I've realised that there are a lot of mediocre players like myself who will never achieve the great heights but can still play pretty well sometimes, improve gradually maybe, and most importantly, enjoy chess. Having time for other pursuits is a good idea.

Thank you for your thoughts ...

Well I'm not saying this is easy to follow. I get to do this just once a week given that work/home and other aspects of "reality" take precedence.   Just started on this regimen and trying to fight through the pain of making something I love resemble work ... wanted to know if I could skip a few steps or add a few useful ones to the mix.

Great post Shivsky! It's hard work to do as you say, but as the saying goes: in all work, there is profit.

I second that. Outstanding post Shivsky! I'm going to save your post into my notepad.

This January, I look forward to investing in software for post mortem analysis and for use in a thread I wish to start on positional analysis for the beginner.

My questions:

1. Are both Rybka and Fritz equally accurate in reporting positional advantage, i.e., producing the same answer of +- 0.4 (4/10ths of a pawn)? 

---

"Step 9. If I still can't figure out why the engine says one move is better, I will save the position and take it to a stronger player / coach and ask him why this might be so. Note that I will NEVER blindly trust an engine unless it is a tactical sequence."

---

2. By tactical sequence, do you mean that engines understand basic material wins or exchanges (4/10ths of a pawn up), but sometimes miss the "+"  of an opening gambit or a sacrifice in the middle/endgame that results in a superior position?

Last, I don't feel the need to limit myself to one engine, if each has a unique strength.

Answer to #1 : Rybka trumps Fritz/Shredder and Co. by pretty good margins. I don't use the latest Rybka, but I notice that it tends to be more cynical of "theoretically winning" positions and will probably give you a +1.1  to +1.3  instead of a Fritz-ian +1.8 to 1.9 score for the same "good position". The more nuanced a position is, the more powerful your engine might need to be. Though I think it is fair to say that for 95% of the critical positions (that warrant deep analysis), any engine (including the free open source ones) would do for the amateur player. As long as it can get to 15 ply within a 20-30 seconds, you're golden. Once again => I don't trust them blindly ... that's why having a coach or a strong player +friend at a club helps ... he sorts through the nonsense for you and gives you his take.

Answer to #2 : When I meant tactical sequences, I mean positions that contain potential for checks/captures and threats.  You can have rich middlegames where the board is fairly locked up. 

My thumb-rule for tactical positions are when the top 3-5 lines contain eval-scores that look like:

1. +1.50 13.Nxd4
1. +0.50 13.h3
1. +0.45 13.h4
...

Notice the huge delta between the best and second best moves. This position demands accurate play. Nxd4 may or may not be obvious.

Consider a different spread ...

1. +0.70 13.exd5
1. +0.50 13.Bb3
1. +0.45 13.Nh5

The deltas between scores is not significant to suggest a tactical "Aha!" move now, is it?

You're spot on regarding gambits, openings (out of book) and for really rich quiescent/non-tactical positions  where there are several candidate moves ... I'd say a strong human has a better grip on things.  I play the Danish gambit from time to time and I enjoy watching the eval. score swing from -1.50 to +1.50 (on occasion!) when Black steps on one of the mines in the Danish minefield. That's also why my proposed Step #11 helps ... Let the history books (Mega databases) tell you how the majority of Master-level players played certain positions rather than a horizon-limited machine. :)

Another post I will paste into my notepad. Thanks for your time!

Too bad you can't inform Rybka in advance that you wish to play the Danish Gambit, so that Rybka can then pull from top GM play and provide a synthesized analysis of your play. It sounds like engines won't put chess teachers out of work anytime soon. I kinda like that, since I'm a teacher, of music.

Interesting post, I would like to add another method. A master once told me he was a bullet game addict, and what he did to get better was he did an overnight analysis of all his bullet games. I've been doing this with my 5-minute blitz games using the 'blunder check' option and can get through a handful of games fairly quickly. It turns out my opponent and I are making mistakes every other move that could cost us the game. There's also positional value from Fritz's analysis and points out your weaknesses over many games. 

this is an awesome post, but honestly i dont think that i can practise the way you suggested, i hardly solve 4 tactics a day and analyse a game upto a whole week, i get frusturated too fast

whenever i play otb tournaments, i am okay with my opening and middlegame but as soon as the endgame phase is reached my concentration slowly gets vanish and my board vision too and finally i ends up poorly, any suggestion??

I learned just reading this thread so far...

Great post, I will be following this now :) I played chess religiously about 4 years ago, 5-6 hours a day at the city square. Sadly, havn't played since then, and naturally my play has suffered... greatly lol. Anyone out there far stronger than I that wants to help me out, by all means message me :)

Again. great post!