Next Phase of Subscriber Reports

[savin]

The next PHase of the Subscriber only reports will be to allow subscriber to see stats on on which openings aplayer users as white and black and how they perform with these openings.

This is more complex that most of you might guess. there are currently almost 2,500 recognised lines and this will just lead to information overload.

We are thinking of tracking the stats based on the top 20 most used positions. Based on a sample 100,000 games at stans this is what we see as the 20 openings we would collect data on:

A0 Irregular Openings
A1, A2, A3 English
A4, A5, A6, A7, D0 Queen Pawn Games
A8, A9 Queens Pawn Dutch
B0 Alekhine and Pirc
B1 Karo Kann
B2, B3, B4, B5, B6, B7, B8, B8 Sicilian
C0 & C1 French Defense
C2 Vienna Game
C3 Kings Gambit
C4 Phildor and Pertrov
C5 Evans Gambit
C6, C7, C8 and C9 Ruy Lopez
D1, D2, D3, D4, D5, D6 Queens Gambit Declined
D7, D8, D9 Gruenfeld
E0 Catalan
E1 Queens Indian
E2, E3, E4 Nimzo Indian
E5, E6, E7, E8, E9 Kings Indian

We are willing to track up to 20 for player stat purposes so we could break one of these in two if a case can be made. As it is we are tracking 38 stats per player (white and black), And for each of these we know if the won, loss, drew, and the average rating of their opponents.

Thoughts?

BTW this will take a while to implement, but the first thing is to decide what to track by player. We will as part of this process be asigning each game with an opening code.

[savin]

Based on the first 100,000 games at the site these openings have the following percentages:

Phildor and Pertrov 14.44%
Queen Pawn Games 13.90%
Vienna Game 11.86%
Alekhine and Pirc 11.31%
Irregular Openings 10.51%
Sicilian 8.53%
Evans Gambit 5.81%
French Defense 5.70%
English 5.10%
Queens Gambit Declined 4.13%
Ruy Lopez 3.11%
Karo Kann 2.04%
Kings Gambit 1.43%
Kings Indian 0.62%
Queens Pawn Dutch 0.48%
Gruenfeld 0.31%
Nimzo Indian 0.30%
Queens Indian 0.24%
Catalan 0.21%

[Diet_Coke is better than chess]

Sounds great. ;D

I assume the percentages are whites advantage in each opening. ie. 10% would be white scoring 55%.

[Blockhead]

savin,

A formidable exercise!

I was wondering how the 100,000 sample of games chosen?

I suggest including:

1. Benoni Defence
2. Reti opening A4, A5, A6, A9
3. Kings Indian Attack A7, A8

I deliberately haven't suggested any particular ECO classification for the old/new/etc. Benoni as they are to numerous. Perhaps a search for the characteristic opening 3/4 moves sufficient.

[savin]

Blockhead,

The sample was actually from about the first 120,000 games to complete. I have a feeling that we may have less variation today.

I suggest including:

1. Benoni Defence
2. Reti opening A4, A5, A6, A9
3. Kings Indian Attack A7, A8

I assume you mean A04, A05, A06 and A09 for the Reti and A07 and A08 for the Kings Indian Attack.

The Benoni nis a bit of pain to gather up but modern Benoni is really A6.

However, I don't want to go oiver 20 of these. So I need to take something out. One idea is to collapse all of the E lines into a single group called Catalan and Indian Openings?

By the way, the Reti Opening is a little strange in that it can transpose very easily. I need to see what my engine is doing with the transposed games. I'm pretty sure that the following would not count as a Reti:

1. Nf3 d5
2. d4

or
1. Nf3 Nf6
2. d4

By the way, in the sample set of 118,900 games about 12,500 were A0 openings. Of these 2100 were Reti and 500 were Kings Indian Attack, Bird's Opening was 800, Nimzo 800 and the remaining 8200 were A00 which a bunch of other stuff.

[savin]

Diet,

The % were the number of games played at the site as a percentage of the total. I ordered them high to low, or most popular first.

[Blockhead]

savin said:

By the way, the Reti Opening is a little strange in that it can transpose very easily. I need to see what my engine is doing with the transposed games. I'm pretty sure that the following would not count as a Reti:

savin,

Yes, I'm aware of the transposition properties of the Reti. It's really a one move (1. Nf3) opening!

MCO14 offers the initial moves:
a) Nf3 d5 2 c4 (Page 706)
and
b) 1 Nf3 d5 (Page 708:)

I apologise if my suggestions make your task more difficult. I do understand that given limited resources some sort of compromise inevitable.

Even though I have a special affection for the Benoni (aka M.Tal!) I'll be more than content to accept your judgement in this matter

Thanks

[savin]

Blockhead,

All input is welcome. My only real limitation is that we don't want to collect more than 20 summary stats. If I compress all the E lines into one that means we can add
4.

By the way, we will assign the full ECo code to each game so it will track about 2700 variants. You will eventually be able to see games at this level. The only thing you won't be able to see quickly is stats by player below the 20 groupings we have. BUT, for a particular line you may be able to see which games a player has played with that line!!!

[shamou]

Sorry, I'm not exactly adding anything constructive - other than to prove how interesting it is to have these stats (and to show that I've worked out how to join the message board) - but I just find it amazing how high a percentage the Vienna Game has! Was the site originally full of romantics? What have we done to it?! Many years ago I did play two practice games against a player who specialised in this opening, but...

[Gritty]

One simple addition I'd quite like is on the "Active Games" page.

If I could see the number of moves made so far (as on the tournament pages), then I could select the games that are in the early stages to play when I haven't got long online, as I play my openings faster than my middlegame.

Also, if you could see the points score, then I could immediately see if someone has played an exchange, go straight into that game and recapture, rather than leaving my opponent waiting until i get to that game, and then making a move in 2 secs that perhaps I could have made days ago.

Incidentally, the reports suggested at the top of this thread look superb.