Rebel wrote: ↑Tue Nov 03, 2020 5:34 pm
I get 2700 elo for Seer 1.1, can that be?
Thank you for testing! Could you elaborate on that time control a little? Was it 1s per move fixed? In that case, it's possible as I've never really tested at such a time control. Would you know if it was losing on time? It's been a long time since there was any reason to test against Fruit 2.1, but I'll run a short match now.
Eds test involves no test games at all. It is a test suite, which means it asks for solutions for single positions ( a lot of them!) and gives points to them.
Such test suites are attempts to get an emulated (kind of) rating and saves a lot of cpu time. I should add that it is a controversial theme, but I think
most people see it as a kind of first indication, besides real game testing.
It is quite possible (I even would say it is expected) that Seer needs more time here.
(I already mentioned the similarity to early LC0 behaviour, which only played exceptionally at higher depths/time,
especially in midgame and/or tactics)
Rebel wrote: ↑Tue Nov 03, 2020 5:34 pm
I get 2700 elo for Seer 1.1, can that be?
Thank you for testing! Could you elaborate on that time control a little? Was it 1s per move fixed? In that case, it's possible as I've never really tested at such a time control. Would you know if it was losing on time? It's been a long time since there was any reason to test against Fruit 2.1, but I'll run a short match now.
Eds test involves no test games at all. It is a test suite, which means it asks for solutions for single positions ( a lot of them!) and gives points to them.
Such test suites are attempts to get an emulated (kind of) rating and saves a lot of cpu time. I should add that it is a controversial theme, but I think
most people see it as a kind of first indication, besides real game testing.
It is quite possible (I even would say it is expected) that Seer needs more time here.
(I already mentioned the similarity to early LC0 behaviour, which only played exceptionally at higher depths/time,
especially in midgame and/or tactics)
Ahh. Thank you for the explanation. In that case, 2700 elo is very reasonable in my opinion. By virture of being so incredibly slow, I would expect it to be extremely tactically weak at 1s.
Rebel wrote: ↑Tue Nov 03, 2020 5:34 pm
I get 2700 elo for Seer 1.1, can that be?
Thank you for testing! Could you elaborate on that time control a little? Was it 1s per move fixed? In that case, it's possible as I've never really tested at such a time control. Would you know if it was losing on time? It's been a long time since there was any reason to test against Fruit 2.1, but I'll run a short match now.
Eds test involves no test games at all. It is a test suite, which means it asks for solutions for single positions ( a lot of them!) and gives points to them.
Such test suites are attempts to get an emulated (kind of) rating and saves a lot of cpu time. I should add that it is a controversial theme, but I think
most people see it as a kind of first indication, besides real game testing.
It is quite possible (I even would say it is expected) that Seer needs more time here.
(I already mentioned the similarity to early LC0 behaviour, which only played exceptionally at higher depths/time,
especially in midgame and/or tactics)
Ahh. Thank you for the explanation. In that case, 2700 elo is very reasonable in my opinion. By virture of being so incredibly slow, I would expect it to be extremely tactically weak at 1s.
Sylwy wrote: ↑Tue Nov 03, 2020 4:56 pmAn engine with a very interesting play. Is my favourite among all 12 engines having NNUE feature.
Thanks for the kind words and testing! For some reason, on my processor, Seer performs disproportionately well relative to other engines (I get 600-700knps locally which is about 50% to 60% the speed of engines such as Koivisto and FabChess). Probably some of the unique play can be attributed to Seer's implementation, from training code to inference code, being all original and written in large part without looking at any of the implementation found in Stockfish. As a consequence, the model architecture differs quite significantly from that used in Stockfish. Therefore, Seer is more akin to Halogen in that it uses some NNUE ideas while not relying on any Stockfish code.
Rebel wrote: ↑Tue Nov 03, 2020 5:34 pm
I get 2700 elo for Seer 1.1, can that be?
Thank you for testing! Could you elaborate on that time control a little? Was it 1s per move fixed? In that case, it's possible as I've never really tested at such a time control. Would you know if it was losing on time? It's been a long time since there was any reason to test against Fruit 2.1, but I'll run a short match now.
Rebel wrote: ↑Tue Nov 03, 2020 8:38 pm
Have a look at SRL.
Yes. Haha. I derped out and totally misunderstood what exactly you were testing. Thanks for the link. I ran a short match between Seer and Fruit 2.1 at fixed 1s/move and Seer was about +380 elo to Fruit 2.1 on my hardware (Seer @ 700 Knps, Fruit at 3.5-4.5Mnps) which isn't necessarily inconsistent with your results on the position tests, but surprising to be sure.
Point 4 in NextGame().
Point 7 in NextGame().
================ C ================
... busy ... ... please, wait ...
Game Nr. 1 (Seer 1.1 64-bit <==*==> Beef 0.3.6 64-bit) in Tournament "Beef v Seer"
Seer 1.1 64-bit
Beef 0.3.6 64-bit
Pairings_Openings: nGameNr = 1 of 10, nOpening = 1, Round = 1.1 at Time0000000342505701
Now in GetInitialFen() at Time0000000342505701
Now in PickAtRandom(7815 of 46953, nType = 1) at Time0000000342505779
Now in Chess960_ID_To_Fen(nID=518) at Time0000000342505779
nMainPrefs in GetInitialFen() = 13, nInitialFenNr = 518, nPgnEpdNr = 7815, nCurrRound = 1, nCurrRndGame = 1
Cycle = 1
Now in DisplayGameSchedule() at Time0000000342505779
Now in DisplayRoundRobinTable() at Time0000000342505795
Got a RoundRobin pairing !
... now you can relax ...
Now in GetInitialFen() at Time0000000342505811
Now in PickAtRandom(10157 of 46953, nType = 1) at Time0000000342505873
Now in Chess960_ID_To_Fen(nID=518) at Time0000000342505873
nMainPrefs in GetInitialFen() = 13, nInitialFenNr = 518, nPgnEpdNr = 10157, nCurrRound = 1, nCurrRndGame = 1
================ EngineAgainstEngine(0000000342505951) =======================================
==== LaunchChessEngine(Seer 1.1 64-bit) ====
Processor is Intel.
Computer Speed = 2 MHz.
Total Physical Memory = 8183 MB.
Total Available Memory = 1951 MB.
Total Available Disk Space = 356189 MB.
=======================================================
Now in EraseInputBuffers(Seer 1.1 64-bit).
Now in MakeTimeControlString() at Time0000000342506357
LaunchChessEngine(Seer 1.1 64-bit) EngineCode = 144686951302168608
LaunchChessEngine() succeeded for Seer 1.1 64-bit
Seer 1.1 64-bit is playing white.
==== LaunchChessEngine(Beef 0.3.6 64-bit) ====
Processor is Intel.
Computer Speed = 2 MHz.
Total Physical Memory = 8183 MB.
Total Available Memory = 1846 MB.
Total Available Disk Space = 356189 MB.
=======================================================
Now in EraseInputBuffers(Beef 0.3.6 64-bit).
Now in MakeTimeControlString() at Time0000000342506981
LaunchChessEngine(Beef 0.3.6 64-bit) EngineCode = 144686951302168608
LaunchChessEngine() succeeded for Beef 0.3.6 64-bit
Beef 0.3.6 64-bit is playing black.
==============================================================================
EngineAgainstEngine() succeeded for Seer 1.1 64-bit
EngineAgainstEngine() succeeded for Beef 0.3.6 64-bit
Now in MakeTimeControlString() at Time0000000342507199
================ SendFirstCommands(Beef 0.3.6 64-bit), nProtocol = 1, EngineCodeHex = 00000020 ====
SendFirstCommands UCI
SendToEng2Time 0000000342507199 : Eng02 (Beef 0.3.6 64-bit) -> uci
.
Scheduling StartEngineTimerEvent() for Beef 0.3.6 64-bit at 0000000342507199.
Waiting for Beef 0.3.6 64-bit.
Eng02 (Beef 0.3.6 64-bit) <- Beef 0.3.6 by Jonathan Tseng
Eng02 (Beef 0.3.6 64-bit) <- Using POPCOUNT
Eng02 (Beef 0.3.6 64-bit) <- id name Beef 0.3.6
nChessVariant(1) = 1, id name.
Cancelling StartEngineTimerEvent() for Beef 0.3.6 64-bit at 0000000342509851.
Eng02 (Beef 0.3.6 64-bit) <- id author Jonathan Tseng
Eng02 (Beef 0.3.6 64-bit) <- option name Hash type spin default 64 min 1 max 65536
Eng02 (Beef 0.3.6 64-bit) <- option name ClearHash type button
Eng02 (Beef 0.3.6 64-bit) <- option name Threads type spin default 1 min 1 max 32
Eng02 (Beef 0.3.6 64-bit) <- option name MoveOverhead type spin default 100 min 0 max 5000
Eng02 (Beef 0.3.6 64-bit) <- option name Ponder type check default false
Eng02 (Beef 0.3.6 64-bit) <- option name BookFile type string default <empty>
Eng02 (Beef 0.3.6 64-bit) <- option name BestBookLine type check default true
Eng02 (Beef 0.3.6 64-bit) <- option name MaxBookDepth type spin default 255 min 1 max 255
Eng02 (Beef 0.3.6 64-bit) <- option name SyzygyPath type string default <empty>
Eng02 (Beef 0.3.6 64-bit) <- option name SyzygyProbeDepth type spin default 0 min 0 max 127
Eng02 (Beef 0.3.6 64-bit) <- option name AnalysisMode type check default false
Eng02 (Beef 0.3.6 64-bit) <- uciok
nChessVariant(1) = 1, uciok.
SendToEng2Time 0000000342509882 : Eng02 (Beef 0.3.6 64-bit) -> setoption name Hash value 256
.
SendToEng2Time 0000000342509913 : Eng02 (Beef 0.3.6 64-bit) -> setoption name Threads value 1
.
SendToEng2Time 0000000342509913 : Eng02 (Beef 0.3.6 64-bit) -> setoption name MoveOverhead value 100
.
SendToEng2Time 0000000342509929 : Eng02 (Beef 0.3.6 64-bit) -> setoption name Ponder value false
.
SendToEng2Time 0000000342509929 : Eng02 (Beef 0.3.6 64-bit) -> setoption name BookFile value <empty>
.
SendToEng2Time 0000000342509945 : Eng02 (Beef 0.3.6 64-bit) -> setoption name BestBookLine value false
.
SendToEng2Time 0000000342509945 : Eng02 (Beef 0.3.6 64-bit) -> setoption name MaxBookDepth value 8
.
SendToEng2Time 0000000342509945 : Eng02 (Beef 0.3.6 64-bit) -> setoption name SyzygyPath value C:\SyzygyTablebases
.
SendToEng2Time 0000000342509945 : Eng02 (Beef 0.3.6 64-bit) -> setoption name SyzygyProbeDepth value 0
.
SendToEng2Time 0000000342509945 : Eng02 (Beef 0.3.6 64-bit) -> setoption name AnalysisMode value false
.
Total Physical Memory = 8183 MB.
Total Available Memory = 1730 MB.
Hash + TbCache to allocate for Beef 0.3.6 64-bit = 288 MB.
SendToEng2Time 0000000342509960 : Eng02 (Beef 0.3.6 64-bit) -> isready
.
Scheduling StartEngineTimerEvent() for Beef 0.3.6 64-bit at 0000000342509960.
Waiting for Beef 0.3.6 64-bit.
Eng02 (Beef 0.3.6 64-bit) <- info string Found 145 WDL, 0 DTM and 145 DTZ tablebase files. Largest 5-men
Eng02 (Beef 0.3.6 64-bit) <- readyok
nChessVariant(1) = 1, readyok.
Cancelling StartEngineTimerEvent() for Beef 0.3.6 64-bit at 0000000342510272.
SendToEng2Time 0000000342510272 : Eng02 (Beef 0.3.6 64-bit) -> ucinewgame
.
SendToEng2Time 0000000342510272 : Eng02 (Beef 0.3.6 64-bit) -> isready
.
Scheduling StartEngineTimerEvent() for Beef 0.3.6 64-bit at 0000000342510272.
Waiting for Beef 0.3.6 64-bit.
Eng02 (Beef 0.3.6 64-bit) <- readyok
nChessVariant(1) = 1, readyok.
Cancelling StartEngineTimerEvent() for Beef 0.3.6 64-bit at 0000000342510319.
nChessVariant(1) = 1.
================ SendFirstCommands(Seer 1.1 64-bit), nProtocol = 1, EngineCodeHex = 00000020 ====
SendFirstCommands UCI
SendToEng1Time 0000000342510319 : Eng01 (Seer 1.1 64-bit) -> uci
.
Scheduling StartEngineTimerEvent() for Seer 1.1 64-bit at 0000000342510335.
Waiting for Seer 1.1 64-bit.
Eng01 (Seer 1.1 64-bit) <- id name Seer 1.1
nChessVariant(0) = 1, id name.
Cancelling StartEngineTimerEvent() for Seer 1.1 64-bit at 0000000342510350.
Eng01 (Seer 1.1 64-bit) <- id author C. McMonigle
Eng01 (Seer 1.1 64-bit) <- option name Clear Hash type button
Eng01 (Seer 1.1 64-bit) <- option name Threads type spin default 1 min 1 max 512
Eng01 (Seer 1.1 64-bit) <- option name Hash type spin default 128 min 1 max 65536
Eng01 (Seer 1.1 64-bit) <- option name Weights type string
Eng01 (Seer 1.1 64-bit) <- uciok
nChessVariant(0) = 1, uciok.
SendToEng1Time 0000000342510413 : Eng01 (Seer 1.1 64-bit) -> setoption name Threads value 1
.
SendToEng1Time 0000000342510413 : Eng01 (Seer 1.1 64-bit) -> setoption name Hash value 256
.
Total Physical Memory = 8183 MB.
Total Available Memory = 1533 MB.
Hash + TbCache to allocate for Seer 1.1 64-bit = 288 MB.
SendToEng1Time 0000000342510413 : Eng01 (Seer 1.1 64-bit) -> isready
.
Scheduling StartEngineTimerEvent() for Seer 1.1 64-bit at 0000000342510413.
Waiting for Seer 1.1 64-bit.
Eng01 (Seer 1.1 64-bit) <- readyok
nChessVariant(0) = 1, readyok.
Cancelling StartEngineTimerEvent() for Seer 1.1 64-bit at 0000000342510615.
SendToEng1Time 0000000342510615 : Eng01 (Seer 1.1 64-bit) -> ucinewgame
.
SendToEng1Time 0000000342510631 : Eng01 (Seer 1.1 64-bit) -> isready
.
Scheduling StartEngineTimerEvent() for Seer 1.1 64-bit at 0000000342510631.
Waiting for Seer 1.1 64-bit.
Eng01 (Seer 1.1 64-bit) <- readyok
nChessVariant(0) = 1, readyok.
Cancelling StartEngineTimerEvent() for Seer 1.1 64-bit at 0000000342510693.
nChessVariant(0) = 1.
UCI time 18000 cs
UCI otim 18000 cs
Now in StartGame() at Time0000000342510709
Now in GUI_StabilityCheck() at Time0000000342510709
ChessGUI 0.245 has been running for 0 days 00 hrs 00 mins 19 secs
Longest run of ChessGUI 0.245 so far was 16 days 22 hrs 40 mins 43 secs
Longest ChessGUI run ever was 16 days 22 hrs 40 mins 43 secs (ChessGUI 0.245)
LaunchChessEngine() succeeded for TimeLagEngine
SendToEng9Time 0000000342511895 : Eng09 (TimeLagEngine) -> uci
isready
.
nMainPrefs=119021732, nPrefs=2222592, nExclusions=131072, nExtraPrefs=0, nCodeEng1=144686951302168608, nCodeEng2=144686951302168608
Now in MakeTimeControlString() at Time0000000342511910
Now in StartGame() for Seer 1.1 64-bit
Now in StartGame() for Beef 0.3.6 64-bit
====================== Beginning of Book moves ========================
Now in SendBookMoves() at Time0000000342511910
Pgn Opening PgnOpening00000001 not found.
Pgn Bookline : 1.d4 e6 2.Nf3 d5 3.g3 Nf6 4.Bg2 g6 5.O-O Bg7 6.Nc3 O-O 7.Bf4 b6 8.Qd2 Ng4 9.Rad1 Ba6 10.a4 Nc6 11.Nb5 Rc8 12.h3 Nf6 1/2-1/2.
Now in ProbeGuiBook() at Time0000000342511926
Now in MakeChessMove() at Time0000000342511926, nChsProgNr = 0, ChsStrct->PlyPtr = 0
Now in MakeChessMove() at Time0000000342511941, nChsProgNr = 1, ChsStrct->PlyPtr = 0
Now in CheckForEndOfGame() at Time0000000342511941
EcoStr = d2d4 d4 => 1.d4
SideToMove = Black.
Now in ProbeGuiBook() at Time0000000342511957
Now in MakeChessMove() at Time0000000342511957, nChsProgNr = 1, ChsStrct->PlyPtr = 1
Now in MakeChessMove() at Time0000000342511957, nChsProgNr = 0, ChsStrct->PlyPtr = 1
Now in CheckForEndOfGame() at Time0000000342511957
EcoStr = e7e6 e6 => 1.d4 e6
SideToMove = White.
Now in ProbeGuiBook() at Time0000000342511988
Now in MakeChessMove() at Time0000000342511988, nChsProgNr = 0, ChsStrct->PlyPtr = 2
Now in MakeChessMove() at Time0000000342511988, nChsProgNr = 1, ChsStrct->PlyPtr = 2
Now in CheckForEndOfGame() at Time0000000342511988
EcoStr = g1f3 Nf3 => 1.d4 e6 2.Nf3
SideToMove = Black.
Now in ProbeGuiBook() at Time0000000342512019
Now in MakeChessMove() at Time0000000342512019, nChsProgNr = 1, ChsStrct->PlyPtr = 3
Now in MakeChessMove() at Time0000000342512019, nChsProgNr = 0, ChsStrct->PlyPtr = 3
Now in CheckForEndOfGame() at Time0000000342512019
EcoStr = d7d5 d5 => 1.d4 e6 2.Nf3 d5
SideToMove = White.
Now in ProbeGuiBook() at Time0000000342512035
Now in MakeChessMove() at Time0000000342512051, nChsProgNr = 0, ChsStrct->PlyPtr = 4
Now in MakeChessMove() at Time0000000342512051, nChsProgNr = 1, ChsStrct->PlyPtr = 4
Now in CheckForEndOfGame() at Time0000000342512051
EcoStr = g2g3 g3 => 1.d4 e6 2.Nf3 d5 3.g3
SideToMove = Black.
Now in ProbeGuiBook() at Time0000000342512066
Now in MakeChessMove() at Time0000000342512066, nChsProgNr = 1, ChsStrct->PlyPtr = 5
Now in MakeChessMove() at Time0000000342512066, nChsProgNr = 0, ChsStrct->PlyPtr = 5
Now in CheckForEndOfGame() at Time0000000342512082
EcoStr = g8f6 Nf6 => 1.d4 e6 2.Nf3 d5 3.g3 Nf6
SideToMove = White.
Now in ProbeGuiBook() at Time0000000342512097
Now in MakeChessMove() at Time0000000342512097, nChsProgNr = 0, ChsStrct->PlyPtr = 6
Now in MakeChessMove() at Time0000000342512097, nChsProgNr = 1, ChsStrct->PlyPtr = 6
Now in CheckForEndOfGame() at Time0000000342512097
Current position (Ply 22) : rnb2rk1/p1q2pbp/1p3np1/3p4/3P4/5NPP/PPPQPPB1/R4RK1 w - - 1 12 = rnb2rk1/p1q2pbp/1p3np1/3p4/3P4/5NPP/PPPQPPB1/R4RK1 w - - 1 12
System time lag per move = 1034 milliseconds
Eng01 (Seer 1.1 64-bit) thought for 3556 milliseconds, time left = 166444 milliseconds
Almost in ParcelMove(from Seer 1.1 64-bit) at Time0000000342538212
e2-e3
Now in MakeChessMove() at Time0000000342538212, nChsProgNr = 0, ChsStrct->PlyPtr = 22
Now in MakeChessMove() at Time0000000342538228, nChsProgNr = 1, ChsStrct->PlyPtr = 22
MatAvail = , MatDiff = .
Now in DrawGraphics() at Time0000000342538243