Ivanhoe v 50 tests

[img][You must be registered and logged in to see this link.][/img]
Sorry Angels77 but i am tired of his prima donna like behaviour and this is how much "slightly" stronger his follow updated one is
he refuses to take advice and tests himself now so..

1+1 games
No bases , same book , ponder OFF , 512 engine hash . 3.4 ghz quad , windows 64

"B" wrote:

[img][You must be registered and logged in to see this link.][/img]
Sorry Angels77 but i am tired of his prima donna like behaviour and this is how much "slightly" stronger his follow updated one is
he refuses to take advice and tests himself now so..

1+1 games
No bases , same book , ponder OFF , 512 engine hash . 3.4 ghz quad , windows 64

:dncekt:

[img][You must be registered and logged in to see this link.][/img]

30 games were enough for me to say ..

Cant see any increase in elo but some increase in speed , wether that is from Ahmeds tinkerings as he suggests ( and very likely ) or the new source code only compareing this to his T0,50 will show

test was NO books NO Bases NO ponder - No significant difference

So:
Changes happened in robbobases (code) only!
no elo change (in version 50).
thanks for testing.

[img][You must be registered and logged in to see this link.][/img]

No bases ,No large pages ,No books , ponder OFF


speaks for itself realy

Tinker belly now says.......

I would just like to say i think the Time Management of Ivanhoe 50hC is bad
I think TimeOrdinaryFactor = 40 is a mistake
Ivanhoe seems to make moves too quickly sometimes in the opening phase and missing a better move,and then later waste time to catch up.

The default TimeOrdinaryFactor = 75 should be better i think.

I got these results so far with 5 minute time controls with TimeOrdinaryFactor = 40



match 910, Blitz 5m 0

1 Houdini 1.03a x64 4_CPU +37 +4/=23/-1 55.36% 15.5/28
2 IvanHoe B50hCx64 -37 +1/=23/-4 44.64% 12.5/28

Tinker belly

Paer 2


Ivanhoe B52aF values

BishopPairMatVal = 45
PawnMatVal = 110 <---- 100 here is better
KnightMatVal = 345
BishopMatVal = 365
RookMatVal = 500
QueenMatVal = 1000
TimeOrdinaryFactor = 75 <--- default decembrists values,which is better
Knight Scale value = 300
WLBishop Scale = 320
WDBishop Scale = 310
BLBishop Scale = 310
BDBishop Scale = 320

Ivanhoe B50hB values

BishopPairMatVal = 48
PawnMatVal = 100
KnightMatVal = 332
BishopMatVal = 348
RookMatVal = 520
QueenMatVal = 950
TimeOrdinaryFactor = 75
Knight Scale value = 300
WLBishop Scale = 315
WDBishop Scale = 310
BLBishop Scale = 310
BDBishop Scale = 315

Ivanhoe B50hB seems to do better against Houdini at longer time controls than B50hC,but still has to be verified.As in my last test where i changed B50hC TimeOrdinaryFactor=75 it was only even with Houdini,but more games are needed to verify this for sure.

match 911, Blitz 5m 0

1 IvanHoe B50hCx64 test +3/=16/-3 50.00% 11.0/22 121.00
2 Houdini 1.03a x64 4_CPU +3/=16/-3 50.00% 11.0/22 121.00

[img][You must be registered and logged in to see this link.][/img]

well draw books are killing engine chess so draw engines .............

TinkrerBell posts.........

Q6600 @ 3.4 ghz
No Bases
Perfect 2010 up to move 7
hash 256 megs each


match 912, Blitz 5m 0

1 IvanHoe B50hBx64 +20 +23/=137/-13 52.89% 91.5/173
2 Houdini 1.03a x64 4_CPU -20 +13/=137/-23 47.11% 81.5/173


~~ Note

this compiles piece values were tuned to play Houdini so vrs other engines...........

this test was sent to us...........


intel Xeon E7420 @ 2.13 GHZ 8 GB Memory
Windows 7 professional 64BITS
GUI: Rybka 4
Book: HS-7moves
Hash: 256 MB
RB and TB: ON
Ponder: OFF
Large pages: ON for both Engines
Blitz: 1+1
CPU usage:both using 4CORES/4T,Task Manager always open to check 100% cpu usage




Blitz 1m+1s


1 Ivanhoe B 52aFx64 +9 +104/=254/-92 51.33% 231.0/450
2 IvanHoe B50hCx64 -9 +92/=254/-104 48.67% 219.0/450

*Update

1 Ivanhoe B 52aFx64 +51 +25/=67/-10 57.35% 58.5/102
2 IvanHoe B50hCx64 -51 +10/=67/-25 42.65% 43.5/102

[img][You must be registered and logged in to see this link.][/img]

No Books, No ,Bases No Large Pages, No intention of wasteing more time testing it

It is faster but NO stronger and i doubt another 100 games would show it to be more than equal

About the speed is may be link with Magic Bitboard use now ?
magic_mult.c than doesn't exit for IvanHoe-Beta-999957a-Beta for example.

2 minute blitz

i7-920 @3.67 ghz, 6 gb ram, same books engine hash 512, large pages ON , ponder OFF



1 IvanHoe B50kAx64p +13 +40/=179/-31 51.80% 129.5/250
2 Houdini 1.03a x64 POPCNT 4_CP -13 +31/=179/-40 48.20% 120.5/250

Blitz 6+3 / 64-bits/ POPCNT / i5-750 / 512 Hash / Large Pages OFF / Ponder OFF

1 IvanHoe B50kAx64p +4/=14/-2 55.00% 11.0/20
2 Houdini 1.03a x64 POPCNT 4_CP +2/=14/-4 45.00% 9.0/20

It's a really short test (only 20 games) but in a "long" time control (6+3). That makes my conclusions less acurate but made a great match to watch and analyse!
Ivanhoe has won in my computer, and that's very good because it never happened in my last tests (Houdini was very strong here, I don't know why...).:

Aitor

PS: If someone is interested in the games, tell me where and how to do it.

EDIT: Corrected a tYpo (Ponder OFF)!

Aiorla wrote:

Blitz 6+3 / 64-bits/ POPCNT / i5-750 / 512 Hash / Large Pages OFF / Ponder ON

1 IvanHoe B50kAx64p +4/=14/-2 55.00% 11.0/20
2 Houdini 1.03a x64 POPCNT 4_CP +2/=14/-4 45.00% 9.0/20

It's a really short test (only 20 games) but in a "long" time control (6+3). That makes my conclusions less acurate but made a great match to watch and analyse!
Ivanhoe has won in my computer, and that's very good because it never happened in my last tests (Houdini was very strong here, I don't know why...).:

Aitor

PS: If someone is interested in the games, tell me where and how to do it.

"B" suggests Ponder OFF is better as each engine benifits from full cpu useage, and Base OFF is best .but a great result and,,

re games you can upload to download site and post link in data base forum or use attatchment option in your post, games would allways be appreciated espesially by book makers

Angels77 wrote:

Aiorla wrote:

Blitz 6+3 / 64-bits/ POPCNT / i5-750 / 512 Hash / Large Pages OFF / Ponder ON

1 IvanHoe B50kAx64p +4/=14/-2 55.00% 11.0/20
2 Houdini 1.03a x64 POPCNT 4_CP +2/=14/-4 45.00% 9.0/20

It's a really short test (only 20 games) but in a "long" time control (6+3). That makes my conclusions less acurate but made a great match to watch and analyse!
Ivanhoe has won in my computer, and that's very good because it never happened in my last tests (Houdini was very strong here, I don't know why...).:

Aitor

PS: If someone is interested in the games, tell me where and how to do it.

"B" suggests Ponder OFF is better as each engine benifits from full cpu useage, and Base OFF is best .but a great result and,,

re games you can upload to download site and post link in data base forum or use attatchment option in your post, games would allways be appreciated espesially by book makers

Oh it's a typo, I did test in Ponder OFF :)

Aitor

PS: I will post the games soon in this post!

EDIT - The games! [You must be registered and logged in to see this link.]

[img][You must be registered and logged in to see this link.][/img]



sample "B"

RATING LIST (1), Blitz 1m+1s


1 Houdini 1.03a x64 POPCNT 4_CP +35/=138/-27 52.00% 104.0/200
2 IvanHoe B50kAx64p +27/=138/-35 48.00% 96.0/200

Using Sedat TOP-100 positions / NO Large Pages / No Ponder / YES Popcnt. / i5-750 NO Overclocking / 64-bits

Well, Houdini has won the match. This may prove that Ivanhoe is better in long games but Houdini is the first in short ones.

Aiorla wrote:

RATING LIST (1), Blitz 1m+1s


1 Houdini 1.03a x64 POPCNT 4_CP +35/=138/-27 52.00% 104.0/200
2 IvanHoe B50kAx64p +27/=138/-35 48.00% 96.0/200

Using Sedat TOP-100 positions / NO Large Pages / No Ponder / YES Popcnt. / i5-750 NO Overclocking / 64-bits

Well, Houdini has won the match. This may prove that Ivanhoe is better in long games but Houdini is the first in short ones.

AnythingSedate has touched becomes highly suspect in my opinion , no offence to you intended

tests can vary and ive not a clue about this book if it is a book or a suite that tests engines knowledge in certain postions , id prefer you re ran test with normal book or no book as 90 & of tests weve seen have this compile beating Houdini at 1+1 and longer TC , but it could just be that your test was longer and more vigerouse .......... only repetition can realy answer the question , but then the net 3 identical tedts could well produce three different results

Great to have you back testing

"B" wrote:

[img][You must be registered and logged in to see this link.][/img]



sample "B"

[img][You must be registered and logged in to see this link.][/img]

sample "B" +


some changes in paramiters

gave slight increas in elo ( for this test at least ) but more draws

* many games wer over 100 moves long , one was 255 moves

me4u wrote:

Aiorla wrote:

RATING LIST (1), Blitz 1m+1s


1 Houdini 1.03a x64 POPCNT 4_CP +35/=138/-27 52.00% 104.0/200
2 IvanHoe B50kAx64p +27/=138/-35 48.00% 96.0/200

Using Sedat TOP-100 positions / NO Large Pages / No Ponder / YES Popcnt. / i5-750 NO Overclocking / 64-bits

Well, Houdini has won the match. This may prove that Ivanhoe is better in long games but Houdini is the first in short ones.

AnythingSedate has touched becomes highly suspect in my opinion , no offence to you intended

tests can vary and ive not a clue about this book if it is a book or a suite that tests engines knowledge in certain postions , id prefer you re ran test with normal book or no book as 90 & of tests weve seen have this compile beating Houdini at 1+1 and longer TC , but it could just be that your test was longer and more vigerouse .......... only repetition can realy answer the question , but then the net 3 identical tedts could well produce three different results

Great to have you back testing

The suite is not the problem, it's only a recopilation of highly played positions in the openings and I made them repeat colors, it's a good work by Sedat.
As you said every test can be different, and in my computer, Houdini has always had a really high level (maybe because Robert and I have the same processor (fact) and it's really optimized ).

Those "big" test are made to make in a future a Rating List, and I will post the games in a database after more matches :)!

Aitor

"B" wrote:

"B" wrote:

[img][You must be registered and logged in to see this link.][/img]



sample "B"

[img][You must be registered and logged in to see this link.][/img]

sample "B" +


some changes in paramiters

gave slight increas in elo ( for this test at least ) but more draws

* many games wer over 100 moves long , one was 255 moves

Sample B plan B from "B"

LOVE IT

RATING LIST (2), Blitz 1m+1s 0


1 Houdini 1.03a x64 POPCNT 4_CP +50/=127/-23 56.75% 113.5/200
2 IvanHoe T0.5.3 +23/=127/-50 43.25% 86.5/200

Not a big improvement, Ahmed .

Next match T0.5.3 vs B50kA

Aiorla wrote:

RATING LIST (2), Blitz 1m+1s 0


1 Houdini 1.03a x64 POPCNT 4_CP +50/=127/-23 56.75% 113.5/200
2 IvanHoe T0.5.3 +23/=127/-50 43.25% 86.5/200

Not a big improvement, Ahmed .

Next match T0.5.3 vs B50kA

you mean ...its rubbish



message to hamhead,,,,,,,,,,,,,quality b4 quantity ...............

Bettina (Sobel ?) wrote:

message to hamhead,,,,,,,,,,,,,quality b4 quantity ...............