New Feature for subs: Post game engine analysis

How about a new feature for the site which is to show a computer analysis of a completed game so that you can learn from any straightforward mistakes you might have made. Allow the analysis to be viewed with the game at any time from then on

(similar to lichess.org analysis)

How the feature would work:
1. When a subscriber finishes a game (either player B or W), there is an option on the finished game page to "request computer analysis".
The user can press this, and then the site responds something like "Analysis in progress - check back later".

2. In the background the site runs some sort of queue of games requested for analysis and runs them through an engine (hand waving) which stores the compiled analysis against the game

3. The next time the subscriber checks the game (eg a few minutes later), the page now has an option which says "View analysis" or similar (because there is now a stored analysis for that game)

4. Open up the analysis to see list of moves and alternates with blunders and mistakes highlighted.


This implies a bit of computing resource, so maybe there is a limit on how many games you can request per day or something like that.

I think it could be a great training aid.

Here is an example of an old RHP game imported into lichess
http://en.lichess.org/31P1PtzE


Some objections are:

a) Everyone has a chess engine already. That's true for me, but not sure if its everyone - i have chessmaster but I think it would be really useful on RHP and so easy to use. I have been using lichess as i like the visual display.
It creates a compelling reason for being a subscriber

b) It would be difficult to implement.
no argument there - it would be a bit of work.

other disadvantages:

* server space (we have been told that we do have the major preformance limitation there. If people would begin to do the analysis (some just because they can) the database would grow very much indeed

* choice of engine. Probably not everybody could agree on what engine with what settings a good analysis would be performed)

* the "see for yourself" effect. If you look at the analysis and find a weak move you nod and go on, if you look at your game to find out where you went wrong you could (possibly) learn mre.

Originally posted by Ponderable
other disadvantages:

* server space (we have been told that we do have the major preformance limitation there. If people would begin to do the analysis (some just because they can) the database would grow very much indeed

Originally posted by delnz
From my understanding, the process of analysis consumes a lot of CPU but the storage of the results is quite small.

At its simplest, there is a board score for each half move, showing the state of play. Where alternates are provided (which is much more useful), its a list of moves from that point.

I don't see that text and those numbers being a lot of storage.
The graphing is on the fly from the numbers.

Good luck on analysing 10M games - that's a lot of pressing the button!
For fun work out how long it would take to do them all if they were done sequentially from a queue @ 30s/analysis (about 0.5s per half move)....*

I had a look at the lichess markup when it makes an observation:

(0.52 - 1.06) Inaccuracy. The best move was Qf6.
[7...Qf68.Be3Qxd49.Bxd4Ne710.Nc3Nc611.Be3O-O-O12.Rfd1Rde813.Bc4Ne514.Bb3Be715.a3]

You can get rid of the text as it can be auto-generated from the alternate which is just

[7...Qf68.Be3Qxd49.Bxd4Ne710.Nc3Nc611.Be3O-O-O12.Rfd1Rde813.Bc4Ne514.Bb3Be715.a3]

about 80 chars

and you could have one of those for nearly every half move I guess (blunder city!)

which would mean about 5-6k of alternate text for a 40 move blunder fest
+ however you store the scoring info



* nearly longer than it took to play them all ;-)