OpeningTree released for first time on Android - has stockfish and open PGN and a book

[adams161]

https://play.google.com/store/apps/deta ... gtreechess

If you want to search then searching OpeningTree michael adams seems to work now though on one of my devies it was a few items down yet. WE just released so not well indexed.

Has engine analysis with stockfish 10 ( including multi line support), an opening book based on 240K games both players 2300 and up, and Open PGN as well as some installed PGN in Open App's PGN ( both items on actions menu lower left).

Nice to have maybe in the future is open more than stockfish but other UCI engines as the code is generic enough.

Let me know what anyone thinks if interested.

Mike

[Ferdy]

https://play.google.com/store/apps/deta ... gtreechess

If you want to search then searching OpeningTree michael adams seems to work now though on one of my devies it was a few items down yet. WE just released so not well indexed.

Has engine analysis with stockfish 10 ( including multi line support), an opening book based on 240K games both players 2300 and up, and Open PGN as well as some installed PGN in Open App's PGN ( both items on actions menu lower left).

Nice to have maybe in the future is open more than stockfish but other UCI engines as the code is generic enough.

Let me know what anyone thinks if interested.

Mike

Thanks for sharing. It is very responsive, making moves on the board, pressing buttons.
Suggestions.
1. Find a way to disable stockfish behavior of showing analysis line from more than 1 pv move to a single pv move. Example
1. e4 e5 2. Nf3 etc then suddenly it will show
1. e4 then back again to
1. e4 e5 2. Nf3 etc
Perhaps use a custom stockfish that is not using aspiration window. It is important for the user to see always the full pv. This can be further improved by also showing the old main line something that is scrollable. Example
+0.12/6 1. e4 e5 Nf3 etc
+0.06/7 1. e4 e6 d4 etc
This way the user will have an idea what is going on with analysis.
2. In multipv mode just use
+0.20/16 1. e4 e5 etc
Instead of D: 16 S: +0.20 1. e4 e5

There is a bug in analysis mode. In analysis you play e2e4 e7e5 g1f3 b8c6 then press the left most navigation arrow to return to start position. The engine analysis line does not update.

[adams161]

Thanks for taking a look at the App. Was pleased it performed well in responsiveness. I'm going to look into these single move analysis lines. Maybe not swapping them in if I have the same first move with more moves after. But for now I updated with the one bug fix and one suggestion.

Analysis will show in multi PV as .02/21 1 e4 e5 etc now not D: 21 S: .02 1 e4 e5
The bug on backing up to start position and analysis stopping is fixed. It was a problem of not hitting code to properly initialize analysis if no moves as the code to start it was encapsulated in the writing of moves.

One other thing done is I have a friend who's been trying test builds each day who is a programmer and sometimes contributes on my projects and he fixed the analysis line display for ambiguous moves which were not written out as cleanly, more long form, and added showing checks to analysis such as Bb5+ not just Bb5.

Really excited to have OpeningTree on the Android market and happy to see some people trying it and learning from their experiences for those things we can do. The App from it's name specializes in chess openings but I've wanted each of its three main components, engine analysis, chess openings, and PGN game reading to work independently for whatever people want to use it for.

Update is live now.

[adams161]

Quick update as this is probably important in an engine forum. I updated it to use stockfish-10-arm64-pgo build, which is the one optimized for speed, if it detects 64 bit support. Before it was just using 32 bit Stockfish for everyone. Update should be live in about an hour or so.

I did a speed test against 'Chess - Analyze This' with an opening position out of the Ruy Lopez around move 5. The 32 bit Stockfish 10 on my Nexus 9 was taking 10-11 Mississippi to reach depth 20 when 'Chess - Analyze This' took 5 Mississippi. Now we hit depth 20 in about the same time, 5 Mississippi. The default cores and ram are apparently the same for both apps. These can be adjusted in settings though a change wont happen till you stop and start the engine but after that its saved and will be there going forward and from app launch to app launch.

[adams161]

I had to do an update today to add back an option in settings to change to 32 bit stockfish. If the engine doesn't start i print i the analysis view to go into settings and try the 32 bit. The problem is devices can incorrectly report they support 64 bit. My one friend has a Samsung S8. You'd expect it to be 64 bit it's a modern phone top of the line. The device page does list the processor and say it's 64 bit. But there is a * and it says as a disclaimer can vary by carrier and country. He is from Australia and I guess his carrier went 32 bit. My guess is it was a hack and they installed the base OS etc and left some stuff that i look at that says its still 64 bit.

[Ferdy]

Thanks for taking a look at the App. Was pleased it performed well in responsiveness. I'm going to look into these single move analysis lines. Maybe not swapping them in if I have the same first move with more moves after. But for now I updated with the one bug fix and one suggestion.

Analysis will show in multi PV as .02/21 1 e4 e5 etc now not D: 21 S: .02 1 e4 e5
The bug on backing up to start position and analysis stopping is fixed. It was a problem of not hitting code to properly initialize analysis if no moves as the code to start it was encapsulated in the writing of moves.

One other thing done is I have a friend who's been trying test builds each day who is a programmer and sometimes contributes on my projects and he fixed the analysis line display for ambiguous moves which were not written out as cleanly, more long form, and added showing checks to analysis such as Bb5+ not just Bb5.

Really excited to have OpeningTree on the Android market and happy to see some people trying it and learning from their experiences for those things we can do. The App from it's name specializes in chess openings but I've wanted each of its three main components, engine analysis, chess openings, and PGN game reading to work independently for whatever people want to use it for.

Update is live now.

Multipv analysis is now showing <score>/<depth>.

There is still issue on pv line shown.


Also it cannot open the pgn file at
http://kirill-kryukov.com/chess/discuss ... 06c5d88f60

Action > Open Pgn

Suggestion
1. Modify display on pv analysis info
In multipv mode, align the multipv score and pv properly, as in the image. Fixed width for score and same starting distance from left for pv lines.

Code: Select all

[+6.0 ] 12. e4 ...
[+0.25] 12. e6 ...

Also remove the depth in the pv line, just the depth at the top is enough. Add time and speed info.



2. Make the score based from the pov (point of view) of the stm (side to move) as in uci protocol. If stm is black and position is better for black display a positive score not a negative score as in white pov. Or make this an option white pov / black pov. Maybe this is just me, but I prefer showing the analysis score using side pov.

3. Add position evaluation symbol after the score

Code: Select all

[+6.0 ] +- 12. e4 ...
[+0.25] =  12. e4 ...

+- White has a decisive advantage
-+ Black has a decisive advantage
https://en.wikipedia.org/wiki/Numeric_Annotation_Glyphs

[adams161]

I'll look into the 2 bugs and 3 suggestions.

The bug in PGN reading i figure will have to wait till I can get to it. I did save the file. It appears to be a heavily sort of formatted PGN file that may be confusing the parser.

In multipv mode, align the multipv score and pv properly, as in the image. Fixed width for score and same starting distance from left for pv lines.

This suggestion on justifying of analysis line seems reasonable. And yea i probably need to set the text view to use a fixed width font. Not sure what i got now.

2. Make the score based from the pov (point of view) of the stm (side to move) as in uci protocol. If stm is black and position is better for black display a positive score not a negative score as in white pov. Or make this an option white pov / black pov. Maybe this is just me, but I prefer showing the analysis score using side pov.

The score display, side to move vs whites score i'm always hearing different things preferred. Maybe that is something that can get added to settings, a choice.

3. Add position evaluation symbol after the score

I thought about it but you need to change between +- += + -= etc. I don't now the break downs of scores when to do that. Do you know any reference?

We did update earlier today and have all new screen shots as well. The old ones were all done on a 10 inch tablet. Now there are phone screenshots for phones and tablet in tablet area. One thing apparent from looking at the phone screen shots of analysis view is there is not so much extra space left. It fills up the analysis view more the analysis. I used a Pixel XL in emulator. We also now always display last move in standard notation not long algebraic.

The other major change is it's possible to choose now from three different skins for the bottom chess openings move area. We still default to the brown theme shown in images in this post but there is also a light blue and a gray/white. The analysis area text view colors are also settable with five choices. One of them is black text view and white text which looks pretty cool. Anyone interested to see the new screen shots or even try the app feel free. It's at https://play.google.com/store/apps/deta ... gtreechess or search OpeningTree all one word in PlayStore.

Thanks again Ferdy for the valuable input and for looking at the app.

But for now let me just share two updated pictures of analysis. Board and pieces are configurable. I have shrunk these pictures down a bit to be friendly to fit on page.



And this is the current look of Multi-PV with the black analysis background and white text theme chosen here.

[Ferdy]

3. Add position evaluation symbol after the score

I thought about it but you need to change between +- += + -= etc. I don't now the break downs of scores when to do that. Do you know any reference?

Analyze This Pro is terrible on this, perhaps this is a bug. Starting at +0.50 and up it already give +- or decisive advantage.

Chessbase online is more reasonable on position symbols. https://play.google.com/store/apps/deta ... ssdb&hl=en
I manually observed this on the tablet.

Code: Select all

-0.25 to +0.25, =   (equal)
+0.26 to +0.69, +=  (slight)
+0.70 to +1.39, +/= (moderate)
+1.40 and   up, +-  (decisive)

Calculating Stockfish 10 score cp vs win rate, against top 9 opponents in FGRL 60 minute games at rating list http://www.fastgm.de/60min.html



Chessbase online starts decisive at 140 cp. For Stockfish this translate to 56% win rate. For humans perhaps this is not enough, I would suggest to start desicive at 200 cp which is around 75% win rate from Stockfish, or even 300 cp.

You can scale up the cp in ending (as material on the board decreases) as win rate decreases even if cp is high. So perhaps in ending, decisive starts at 500 cp. But implement first the basic one, add scaling later.

[Ferdy]

Another bug.

[Ferdy]

3. Add position evaluation symbol after the score

I thought about it but you need to change between +- += + -= etc. I don't now the break downs of scores when to do that. Do you know any reference?

Analyze This Pro is terrible on this, perhaps this is a bug. Starting at +0.50 and up it already give +- or decisive advantage.

I have to correct myself regarding analyze this pro. At 2.0 pawn unit and up its decisive. At 0.5 to 1.99 it is moderate and at 0.01 to 0.49 it is slight.

It seems like the start of slight is too low at 0.01. This is only 1 cp.