Strange Statistics in ChessBase

Can someone tell me what in the world my database is doing?  In the two screenshots below, the positions are from the Lopez.  In the first one, the database says that 10 games have been played, and that white scores only 35% when continuing Ba4, which has got to be wrong.  Then, if you look at the next shot, the score completely changes when you make that move.  There are lots of great continuations after that, and the collective score sure as heck isn't 35%.  What's going on?

I realize that what could be happening is that the database is determining that the games that reached this position specifically (shot one) score 35%, but I don't know why that would matter on lick, considering they transpose into other branches (of a branch diagram) that score much better in the resulting position after Ba4.  In fact, I don't even know what anyone would want with the score 35%, if it refers to the outcomes of games that specifically reached that position in the first screenshot, considering the resulting position after Ba4 scores much better than 35%. 

This is the position with >10,000 games (the regular Ruy move order)

The position in your first picture doesn't allow the database to see this though because white has already castled. However after you play Ba4, the position has transposed into the Ruy mainline (white will play 5.0-0 in my first diagram and now it matches your 2nd picture whose last move was 5.Ba4).

---

I suspect the mover order from the first picture was something like a Berlin variation... but then black played a6 (diagram below). However this move order by black lets white win a pawn.

I answered your question in the OTB private forum.

The 35% represents White's score. Your position is the first diagram is a rare one, but after 5...Ba4, you have transposed to the most popular line of the Ruy Lopez. My screenshot was produced using PowerBook 2016 as the reference database.

I would guess the % for Ba4 is so low in that particular position (even though the move transposes into a normal position) because the players who missed the chance to win a pawn with Bxc6 followed by Nxe5 are either very low rated, or were having a bad day.

Pretty much THE explanation. OP, what's the move order for your first picture?

The previous posts basically answer it, ...a6 is a losing blunder and you can't read anything into a % based on 10 games (too easy for one or two freak results to distort it). That being said, a similar search gives me 400+ games. Which database are you using?

It's just strange that the percentage would be so low for Ba4, considering the very next position has a collection of moves for black, and none of them are lower than 50%.  That absolutely makes no sense, does it?

a6 isn't a losing blunder.  The percentage of 35% is white's score when playing Ba4 after a6.  The scores you show in your screenshot are more like what I'd expect to see.

I'm still not sure I know what this "35%" refers to, because the only thing I can figure is thatit refers to the outcomes of games that reached this specific position.  And that's useless, considering this position immediately transposes into lines that does well for white.  

i'm guessing you haven't yet read Ziryab's post above, which makes perfect sense to me.

Also note that Chessbase is looking only in your reference database for the given position. If you have a very small database set as your reference database, your results will be a much smaller sample of games

Of course ...a6 is a losing blunder as it loses a pawn for nothing after Bxc6. The 35 % is meaningless if it's only isolating a small number of games. 

Say for example there were 10 games with 1.e3 and white scored 0% (just using these numbers for an example). Then after you enter the moves 1.e3 e6 2.e4 e5 you transpose back into a main position where White obviously scores decently.

It does sound like you need to reset your search boosters on chessbase if your database is a decent size (several millon games) and that's all you're getting. (reference database, right click/delete search booster, right click again/create search booster)

Yeah, makes sense. I've had problems like this before where I see a GM game and a certain move is played... and I thought my database said it was not good, so I check again, and it was like your example (tons of games pop up and the move is fine).

From Ziryab's picture it looks like he opened two panes, one for notation and one for notation + book. Hopefully the  program I have has a similar feature (I'll have to look around in the options).

LOL...  I obviously didn't spend any time looking at the position.  I was looking at the statistics.  Oops.

Yeah, I'm good.  I read his response in another forum after commenting here.

Percentages can be tricky.

Even when there are a lot of games I (maybe incorrectly) try to imagine what the players have in mind. Sometimes I find, for example, a solid drawing line where most games it's some FM or IM trying to draw a 2600+ so the % suggests the line is a little dubious, but the line itself is super solid.

Or you know, a line the elite are using to push hard for a win. The % may suggest it's better or worse than what it really is... which is just a way to play for a win in an unbalanced position.

Makes complete sense.

That book is garbage.

Do you have it? Can you say more?

a lot of folks struggle with Chessbase, up until v7 there was a user's manual came with it, about 200 pages IIRC. v8 then had a much smaller booklet, then each version since relies more and more on the help files.

I'm a little rusty with some functions that i don't use often, but otherwise comfortable everywhere else, but i know many people really aren't getting the best from it, the above book is highly recommended for those people...;-)

When I was new to ChessBase fourteen years ago, I read a lot of articles by Steve Lopez. Back then, they were accessible through the ChessBase website.

James, I reviewed the book, but only spoke generally about the problems with it. If I were to expand the review, I would include the sorts of continued faults in my knowledge that the book didn't address, like the way the stats are set up, etc.

If it helps, here is the link: https://www.amazon.com/review/RYS8CK1N10A8Y/ref=cm_cr_rdp_perm?ie=UTF8&ASIN=1936490544