Chess for Android Release (API30 complaint)

[abik]

use engines like before Google went crazy.

I must admit that I feel a bit double on this. On one hand, I highly appreciate the efforts of Google Play to keep that a safe place, where apps only get the permissions they need and users are comfortable with, but no more. But as a developer, it is hard to keep features working (and in some cases, like third party engine install from SD, keeping it working has become even impossible). Getting bombarded with users that give one star ratings on my app solely because I try to be a good Google Play citizen does not help either.....

[Ras]

On one hand, I highly appreciate the efforts of Google Play to keep that a safe place, where apps only get the permissions they need and users are comfortable with, but no more.

F-Droid looks for anti-features anyway. The problem is that previously, an Android compile was just a Linux compile using the NDK. Now you have to go into a lot of Android fuss to make an APK, which many engine devs have no interest in. Which means, no Android version anymore. That will hit the mid-range in particular.

If Google chooses to ignore the needs of the chess community, I think it is time to do away with Google, and F-Droid offers that option. Droidfish there is up to date, and there are not ever more hoops to jump through. I have already updated my documentation to point users into that direction. Btw., the F-Droid store doesn't require an account, unlike Google's store, so that's a privacy plus for F-Droid.

[Chessqueen]

On one hand, I highly appreciate the efforts of Google Play to keep that a safe place, where apps only get the permissions they need and users are comfortable with, but no more.

F-Droid looks for anti-features anyway. The problem is that previously, an Android compile was just a Linux compile using the NDK. Now you have to go into a lot of Android fuss to make an APK, which many engine devs have no interest in. Which means, no Android version anymore. That will hit the mid-range in particular.

If Google chooses to ignore the needs of the chess community, I think it is time to do away with Google, and F-Droid offers that option. Droidfish there is up to date, and there are not ever more hoops to jump through. I have already updated my documentation to point users into that direction. Btw., the F-Droid store doesn't require an account, unlike Google's store, so that's a privacy plus for F-Droid.

What Chess engine for Android offer UCI Elo Settings ?

[abik]

What Chess engine for Android offer UCI Elo Settings ?

With the very small set of engine packages I have installed, I see the following list. Two have UCI_ELO.

Code: Select all

Chess for Android
Andscacs 0.921
BikJump 2.5
Deep Fritz 15
Gamit Fruit 2.2 beta 4bx
Hakkapeliitta 3.0
Komodo  8 - 13   (stops working at 11, but only for gui apps that have API30)
Laser 1.8 beta
Maia
OpenTal 1.1
Rodent III 0.171
Rodent IV 0.32  (with UCI_ELO)
Senpai 2.0
Stockfish 14 (with UCI_ELO)
Toga II 4.01
asmFish 20181204

[Chessqueen]

What Chess engine for Android offer UCI Elo Settings ?

With the very small set of engine packages I have installed, I see the following list. Two have UCI_ELO.

Code: Select all

Chess for Android
Andscacs 0.921
BikJump 2.5
Deep Fritz 15
Gamit Fruit 2.2 beta 4bx
Hakkapeliitta 3.0
Komodo  8 - 13   (stops working at 11, but only for gui apps that have API30)
Laser 1.8 beta
Maia
OpenTal 1.1
Rodent III 0.171
Rodent IV 0.32  (with UCI_ELO)
Senpai 2.0
Stockfish 14 (with UCI_ELO)
Toga II 4.01
asmFish 20181204

Under Stockfish 14, the UCI_ELO are not calibrated properly to equal human ELO, or close to it within 100 ELO rating. I matched it against Komodo Dragon UCI_ ELO rating and Komodo come closer when paired against Fide rated humans.

[Chessqueen]

Komodo Dragon Skill = 17 is closer to human rated 1700 to 1750, but Stockfish14 UCI_ELO 1700 is more like 1950 Human FIDE. Under Stockfish 14, the UCI_ELO are not calibrated properly to equal human ELO, or close to it within 100 ELO rating. I matched it against Komodo Dragon UCI_ ELO rating and Komodo come closer when paired against Fide rated humans.

[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.12.21"]
[Round "?"]
[White "Dragon-2.5-64bit-avx2"]
[Black "Stockfish 14"]
[Result "0-1"]
[BlackElo "1700"]
[ECO "C42"]
[Opening "Russian Game"]
[Time "19:57:30"]
[Variation "Classical, 6.Bd3"]
[WhiteElo "1700"]
[TimeControl "300+3"]
[Termination "normal"]
[PlyCount "54"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e5 2. Nf3 Nf6 3. Nxe5 d6 4. Nf3 Nxe4 5. d4 d5 6. Bd3 Nc6 7. O-O Bg4
8. Re1 f5 9. c4 Bb4 10. cxd5 Bxf3 11. gxf3 Bxe1 12. fxe4 Bxf2+ 13. Kxf2
fxe4 14. Qh5+ g6 15. Qg5 exd3 16. dxc6 O-O+ 17. Kg1 Qxd4+ 18. Be3 Qxb2 19.
Qd5+ Rf7 20. cxb7 Rf8 21. Qb3 Qe2 22. Bf4 Qe1+ 23. Kg2 Qe4+ 24. Kg1 Qxf4
25. Qxd3 Qg4+ 26. Qg3 Rf1+ 27. Kg2 R8f2# 0-1[/pgn]

[purechess]

UCI_ELO doesn't mean Elo like in Fide Elo ....it's basically just a level of strength you set for the engine.

[abik]

UCI_ELO doesn't mean Elo like in Fide Elo ....it's basically just a level of strength you set for the engine.

Indeed! And even if there was a formal requirement in the UCI protocol to keep UCI_ELO calibrated, that seems a daunting task to me. Unless you want to come up with some fixed-depth solution, any time-based solution would have to account for a vast amount of different target architectures and CPUs on desktops, tablets, phones, etc.

As an anecdotal example, I noticed that more recently, some users observed that the built-in engine of Chess for Android had become too strong. This really surprised me, since I have not made any changes to the core engine for a while. So, my conclusion was that over time, the mobile CPUs have become so fast that even my super modest bottom of the pile built-in engine starts to play a very decent game

[Ras]

Unless you want to come up with some fixed-depth solution, any time-based solution would have to account for a vast amount of different target architectures and CPUs on desktops, tablets, phones, etc.

In my CT800 engine, I'm throttling UCI_Elo via absolute node rate, and that works regardless of the hardware, also under Android. As added bonus, the throttling is done in a battery-saving way. On the low end, I also add in noise to the eval function on top of the node rate throttling.

[abik]

Now you have to go into a lot of Android fuss to make an APK, which many engine devs have no interest in.

On that note, and reading the room completely wrong of course , would you be willing to package the CT800 engine into an OEX or CB compatible app? I actually get a lot of requests from users for this engine. If licensing allows, I could also package the engine in my own engine package, which used to have many engines, but due to lack of third-party author enthusiasm, eventually just became my own BikJump engine.