Practice makes patzer - how I went 116 games without improving.

Practice makes patzer - how I went 116 games without improving.

Jan 15, 2017, 3:05 PM |

I'm a bit of a data junky, so understandably I got excited about the announcement of's Computer Aggregated Precision Score system. The CAPS system applies a score from zero to 100 percent to a given game, or a set of the player's past games. A CAPS of zero percent means a game played with one of the worst moves on every turn, and 100 percent is a game where only the top choice of the world's strongest chess computers was played on every move. 

This weekend I finally got around to analysing my CAPS score to see how much I had improved since I began playing in July.

This involved "analysing" (first rule of analysing: it's not analysing if you have six tabs open and don't look at any of the games or moves - it's data mining) all my daily correspondence games and rapid games that I'd played to date. I removed any daily games where I won/lost on time unless it was clear I was obviously winning or losing at that point. I have saved all the CAPS scores for myself and my opponents, along with both players ratings at the time of the match, the number of moves, the outcome, colour of pieces I had, and the date the game started.

It was very obvious that my CAPS scores jumped around from game to game, so I decided my primary CAPS metric would be a 50 game rolling average I call CAPS50. Short enough to indicate recent form, but long enough to take out the noise of a handful of exceptional/exceptionally poor games.

My expectation had been that I would see a slow gradual improvement. Instead what I saw horrified me. I went over 100 games without improving my "CAPS50" score at all. I knew my rating hadn't improved much in the first few months but I assumed that somewhere, in the background, my Chess was improving through the practice. It wasn't.




After my first 50 games I had a CAPS50 average of 49. After game 116 I had a CAPS50 average of 48.8!? I think this is clear evidence that to improve at Chess you cannot just play and get better. That might help a little, but to really improve you have to actively learn. I think this is why some of the players whose profiles have viewed have been playing for hundreds, or thousands of games with a similar rating. To improve you need to strengthen some part of your play actively through study or focused practice. I started doing this at the end of November (around game 105) and immediately saw more progress than I'd had in the previous hundred games.

Table below shows a month on month average of my CAPS since July.

Month My CAPS Opponent CAPS Sample Size
Jul-16 69.45 35.76 3
Aug-16 51.21 50.66 16
Sep-16 24.42 59.19 9
Oct-16 57.54 49.67 29
Nov-16 46.14 57.98 58
Dec-16 58.42 55.12 41
Jan-17 63.11 47.66



Whilst I've chosen CAPS50 as my primary CAPS measure to use alongside my rating I have looked at other rolling averages. The CAPS7 metric is useful to show exactly how I have improved. Whilst I'm definitely having more good games, I've (mostly) just improved enough to remove the absolutely rotten games. It's been over 70 games since my CAPS7 was below 40 which is a good indicator I have removed some of my worst games from my repertoire. I might still be losing, but I'm putting up a fight.



It is easy to play well if an opponent gifts you opportunities (taking a hanging queen with your knight is the 100% perfect move, finding that move in a positional deadlock is much trickier). I'll need more data but am provisionally assessing which are my best games using myCAPS+(OpponentCAPS/4). My "best" game currently is one I lost with a CAPS of 85 v an opponent who had a CAPS in the mid-nineties!



Chess is a 2 player game. How my opponents playing strength interacted with my own.

Opponent CAPS My CAPS average v opponents who scored in this CAPS range Sample Size
0-9 48.53 12
10-19 54.66 17
20-29 54.27 18
30-39 53.73 22
40-49 57.24 19
50-59 61.40 19
60-69 54.64 18
70-79 51.94 20
80-89 50.30 15
90-100 45.36



All very interesting (to me) but key lessons I have taken away are a) to improve you have to study and learn not just play, b) you can improve your CAPS/rating a lot by ensuring your bad games are not very bad games c) I suck! but it is nice to suck a little less than I did in September...

Finally, the graphic below shows I have a long way to go before I venture out over the board...