What does "weighted error value" mean in Chessbase 16?

I splurged and upgraded from Chessbase 15 to 16. It has some cool features, but in its technical analysis, it took away the useful & obvious "percent accurate" and replaced it with the useless "weighted error value". Why is it useless? I don't know what it means and it does not appear in the manual. It sounds like a statistical analysis term, but I don't know how it applies to chess. Can anyone help, please? Thanks.

Example:  Weighted Error Value: White=1.59/Black=1.06

Hey Tom, I hope all is well. I wish I could help more, but nothing obvious is found online. I knew you checked this, but I verified your findings. So lets walk through this:

Let's presume the programmers didn't really change the algorithm (because they are lazy, and it works, and that's what programmers do), so the old 'percent accurate' calculations are probably very similar (or exactly the same) as the new 'error value' calculations.

'Error' can be defined as "1- 'accuracy' ". Is it as simple as that? If your accuracy was 75%, is your error now 25%? Check older games through both versions of the code and see if this correlates. If not, then we need to figure out how they apply "weighting".

That may, or may not be difficult. Let me know if this helps.

Example:  Weighted Error Value: White=1.59/Black=1.06

Ken, thanks, but that makes the above White = -41% and Black -94% accurate. I sometimes feel like I play that badly, but I don't think it's actually possible. BTW, did you have me (VBA) in mind with the lazy programming comments? :-) 

I tried redoing one that had been White 56% and Black 15% (fortunately, I was White); the weighted error value came out to Weight 0.38% and Black 0.87%. Those come pretty close to your estimate. If I hadn't deleted ChessBase 15 I could look at how it would rate the example I gave first. When the Weighted Error Value is over 1, I dunno what to make of it. 

Happy New Year. Hope 2021 will be better.

Tom, be careful not to confuse percentage with a 'weighted index".

For example, perhaps a "perfect" score is 2.0. Therefore, white would be 1.59/2=79.5% of "perfect" and black would be 1.06/2=53% of perfection.

They may have also incorporated a sliding scale, or perhaps a normal distribution, such that the difference between a 1.9 and a perfect 2.0 only applies a small 'penalty' similar to a z-score in statistics. This could be their 'weighting' factor.

Regarding VBA programming, there is laziness, and then there is "code efficiency". I'm trying to remember where you fell on that spectrum. 🤔

Take a look at the video here https://www.youtube.com/watch?v=xBzhdBJQ-g4&feature=emb_logo

At aproximately time 58:00 you will get your question correctly answered directly by one of the head Chessbase programmers.

@leifel Thanks! I had watched part of that video, but it's so long I ran out of patience. Not only does it answer my main question about Weighted Error Value, but right after that he expains a more minor function (for me), which is My Move. Terrific! Thanks, again!

It refers to a centipawn loss per movement compared to the engine movement.  So it indicates how much each of your movements lose compared to the engine moves. A perfect 0.0 would mean you played exactly what the engine would play and 1.0 would mean that on each movement you are losing a "pawn" compared to the engine

Thanks, all, for the responses. Now I wonder, is there a range of Weight Error Values that would correspond to a good score for a range of chess ratings? I've had a few great games where I scored under 0.20, and unfortunately several more where I scored over 1.00. I'm only a Class C USCF player, and am wondering what a reasonable goal should be. Guess I can go through a bunch of my games and make a spreadsheet from my scores, and just try to beat my average...

This is the question that has plagued me for YEARS and I think I finally have an answer, so here it is:

Average error value is, how much error you make per move on average, illustrated as average pawns lost per move. An Average Error Value of 1.59 would mean on average, the player is losing 1.59 pawns worth of advantage per move.

The way I understood it from the video @leifel posted, Weighted Error Value is a bit different; In that it weighs blunders more than the times you make a suboptimal move which is still winning. In other words, it seems to me that Weighted Error Value is how much of a "winning chess" you're playing instead of measuring the absolute similarity to "engine chess". 

We don't know how this is actually weighted of course. Knowing that would give one information on how often they can use engines and avoid cheat detection.

Accuracy is whole another thing though, and it depends on how each move affected your winning chances. That would require you having access to statistical data on how it would affect a large number of players, as explained by the lovable renegades at Lichess here:

https://lichess.org/page/accuracy

So okay, let's say you played a game and want to know how well. There are three ways: 

A-The best way: Just buy chess.com premium. They have the largest player base, hence largest statistical data, hence the most accurate number for accuracy.

B-The Free way: Copy the PGN file, go to Lichess, click on Import Game under Tools tab, import your game and run a computer analysis. It's easy and it's free. It seems to me the accuracy numbers are exaggerated though.

C-The Lazy Way: All you need is to multiply the Weighted Error Value by the number of moves played; Let's call that Total Error Value because I like making things up. Now you subtract that from 100 and you'll have your percentage(Which is not the same as Accuracy, but works well enough).

For an example, let's look at my game here:

https://www.chess.com/game/live/91210724681

where I respond to 1-C4 with my patented 1...Na6?! (The Outlander Defense I call it. Didn't make it up, just studied it a lot. The idea is c6 and Nc7 at some point. It works in Blitz, beyond that, well, it's horrible.)

Chess.com says the accuracy is 82.5%

Lichess number is 97%. I mean 1.C4 Na6 would make any CPU cry, fry then die...so no. Just no. Too high.

Fritz says my Weighted Error Value was 0.40. I played 51 moves. Total Error Value is 0.40* 51= 20.4. And 100 - 20.4 = 79.6. Done.

79.6% is pretty close to chess.com number so good enough. Again this is NOT "Accuracy", just an easy to understand and digest number in contrast to the tough concept of Weighted Error Value.