@sheckley666 said in #4:
> Sure?
en.wikipedia.org/wiki/Birthday_problem ;)
Well you're talking about the games you played, not everyone, so yes, I'm sure you'll need at least to play a big number of games before this is useful.
For instance, I did play a lot (nearly 6K games), so... I downloaded all of my games, extracted the FEN for each chess960 game I played and got these results:
- 5696 games analyzed
- 947 different starting positions (so I didn't cover all of them yet!)
- position appearing the most: 19 times
- no positions appearing 18 times, 1 position 17 times, 2 positions 16 times, etc.
- 34 positions appearing only 1 time, 73 positions appearing twice (many more than 1 time, normal because I always offer rematch)
So I do agree it would be "meaningful" (not necessarily useful) for me (but again ~6K games)
You have ~1K chess960 games played. The same analysis for you shows:
- 1016 games analyzed
- 614 different starting positions
- position appearing the most: 7 times
- 1 position 7 times, 3 positions 6 times, etc.
- 337 positions appear only 1 time, 193 positions appear twice
In fact, for you I can share the whole counts:
1 appearance: 337 games
2 app.: 193 games
3 app.: 59 games
4 app.: 14 games
5 app.: 7 games
6 app.: 3 games
7 app.: 1 game
We can argue whether that's meaningful enough or not, but that's already after ~1K games so I stand by my observation of "unless you play thousands of games..."
I think for most players, 1 occurrence or 2 (if you always offer rematch) will be by far the most frequent datapoints, as it's in your case.
So I think that using this to see
> which one do I have to improve
is a bit optimistic IMHO. The results of your games will probably have very little to do with the starting position (one of the perks of chess960) and by design I don't think it can be approached in the theoretical way you approach the study of chess. Can I really retain in my memory any studies I do on my "least successful" of those 947 positions I played? And more important, can I recall that when that position happens again in a few months from now?
Again if I were to apply some serious study to openings in 960 I would focus, as I proposed, in features of the starting position. But that's just me...
--
Oh, and I think the pigeonhole principle applies better than the birthday problem for your argument. ;-)