RomiChess96 - Olithinkwin32 : 31.5/36 27-0-9 (===1==1111=111

[Michael Sherwin]

RomiChess96 - Olithinkwin32 : 59.0/71 51-4-16 (===1==1111=111111=1111111=111111=11111111111=110110=11=0==110=1111111=1) 83% +275

A little speed bump in the middle of the road, but still a great result so far. And Romi's picking her speed back up again. She called on her cell and said that she's going to fly by the corner store to pick up a six pack and then head over to Homer's to see what mischief they could get into.

This was Romi's old record:

RomiChess96 - Olithinkwin32 : 71.5/100 61-18-21 (101=111111111=110101=111111==1=1=1010===1=11111010011101=1=011=1=0111111111=111010111=101110001====0) 72% +164

This is Romi's new record:

RomiChess96 - Olithinkwin32 : 79.5/100 68-9-23 (===1==1111=111111=1111111=111111=11111111111=110110=11=0==110=1111111=1111110=10==1111=11011011==01=) 80% +241

Romi's points by half:

old: 37.5, 34.0

new: 44.0, 35.5

elo gain: 241 - 164 = +77

What are the error bars on this result?

If the error bars indicate that this version is best then I will just wrap it up and send it out!

[BubbaTough]

What are the error bars on this result?

If the error bars indicate that this version is best then I will just wrap it up and send it out!

If you have the pgn saved:

http://wbec-ridderkerk.nl/html/download.htm

-Sam

[George Tsavdaris]

There are two camps of thought.

1) open source is wonderful and leads to quick improvement in chess engine strength, period.

2) open source means lots of very similar chess programs and who needs that. Testers (most anyway) do not want to spend time and effort testing an engine just to find out that it is just more of the same. They like to test new original engines even though they might not be all that strong. Too have every new engine to be as strong as Fruit or stronger just seems phony to them. Also there is the cloning from source issue and all the hassle and ugly that goes with it. Also people tend to respect authors more if they struggled to create something from scratch that has a bit of uniqueness about it.

People that want from authors to reinvent the wheel should respect computer Chess advancement more than the authors. And by reinventing the wheel a considerable delay in Computer Chess is introduced.

Variety of engines is not an issue i think. This fear, if exists, of if there are strong open source programs then all engines will play the same, is an illusion since if there are e.g 1000 methods and tricks of programming a Chess engine, then some will work in an engine and others will not work, so engines will have their own style for sure more or less.

Engine's strength increases so much that we have reached a point today that from the best 1-4 moves in a position, top and semi-top engines to be able to choose form them most of the times, in contrast to the past where engines played crap moves much more frequently. So this gives the deception of similar style.
Junior's style of speculative moves, loses badly today in contrast with the past where mistakes were not punished so violently, so engines are diverting from this.

For me personally, the release of Fruit 2.1 was not helpful. If I had to say one way or the other, I would say that it was hurtful and still is. I do not enjoy all this Fruit, Rybka, Strelka, Ippo disharmony. I also have more selfish reasons for not liking open source engines that are monsters compared to RomiChess. If I ever get Romi up to Fruit 2.1 strength I will no longer continue to release the sources.

So i guess that means a release of EXE+code very soon. So stop the match now at 88%.

[Michael Sherwin]

What are the error bars on this result?

If the error bars indicate that this version is best then I will just wrap it up and send it out!

If you have the pgn saved:

http://wbec-ridderkerk.nl/html/download.htm

-Sam

Code: Select all

1/6/2010 4:51:28 PM :

    Program                          Elo    +   -   Games   Score   Av.Op.  Draws

  1 RomiChess96                    : 2618   70  67   100    79.5 %   2382   23.0 %
  2 Olithinkwin32                  : 2382   67  70   100    20.5 %   2618   23.0 %

so

241 - 67 = 174

and

164 + 70 = 234

Therefore the lesser result could be better 60 elo stronger than the better result.

more test needed!

[George Tsavdaris]

RomiChess96 - Olithinkwin32 : 59.0/71 51-4-16 (===1==1111=111111=1111111=111111=11111111111=110110=11=0==110=1111111=1) 83% +275

A little speed bump in the middle of the road, but still a great result so far. And Romi's picking her speed back up again. She called on her cell and said that she's going to fly by the corner store to pick up a six pack and then head over to Homer's to see what mischief they could get into.

This was Romi's old record:

RomiChess96 - Olithinkwin32 : 71.5/100 61-18-21 (101=111111111=110101=111111==1=1=1010===1=11111010011101=1=011=1=0111111111=111010111=101110001====0) 72% +164

This is Romi's new record:

RomiChess96 - Olithinkwin32 : 79.5/100 68-9-23 (===1==1111=111111=1111111=111111=11111111111=110110=11=0==110=1111111=1111110=10==1111=11011011==01=) 80% +241

What are the error bars on this result?

If the error bars indicate that this version is best then I will just wrap it up and send it out!

With 99% confidence level:

Old-Romi vs Olithink: +160 ELO most probable and ELO is inside [81 , 258]
New-Romi vs Olithink: +235 ELO most probable and ELO is inside [156 , 345]
-------------------
-------------------
With 95% confidence level:

Old-Romi vs Olithink: +160 ELO most probable and ELO is inside [99 , 232]
New-Romi vs Olithink: +235 ELO most probable and ELO is inside [174 , 314]

*When i say ELO i mean ELO difference.

This improvement came only by applying Dann's formula(modified by you with the coefficients you gave) and this +10 thing?

[Michael Sherwin]

RomiChess96 - Olithinkwin32 : 59.0/71 51-4-16 (===1==1111=111111=1111111=111111=11111111111=110110=11=0==110=1111111=1) 83% +275

A little speed bump in the middle of the road, but still a great result so far. And Romi's picking her speed back up again. She called on her cell and said that she's going to fly by the corner store to pick up a six pack and then head over to Homer's to see what mischief they could get into.

This was Romi's old record:

RomiChess96 - Olithinkwin32 : 71.5/100 61-18-21 (101=111111111=110101=111111==1=1=1010===1=11111010011101=1=011=1=0111111111=111010111=101110001====0) 72% +164

This is Romi's new record:

RomiChess96 - Olithinkwin32 : 79.5/100 68-9-23 (===1==1111=111111=1111111=111111=11111111111=110110=11=0==110=1111111=1111110=10==1111=11011011==01=) 80% +241

What are the error bars on this result?

If the error bars indicate that this version is best then I will just wrap it up and send it out!

With 99% confidence level:

Old-Romi vs Olithink: +160 ELO most probable and ELO is inside [81 , 258]
New-Romi vs Olithink: +235 ELO most probable and ELO is inside [156 , 345]
-------------------
-------------------
With 95% confidence level:

Old-Romi vs Olithink: +160 ELO most probable and ELO is inside [99 , 232]
New-Romi vs Olithink: +235 ELO most probable and ELO is inside [174 , 314]

*When i say ELO i mean ELO difference.

This improvement came only by applying Dann's formula(modified by you with the coefficients you gave) and this +10 thing?

The only thing between old and new is the +10 thing.

[Rubinus]

Probably find better opponent ...
Regard Pavel Háse

[Graham Banks]

If the error bars indicate that this version is best then I will just wrap it up and send it out!

Hi Mike,

by the time I start Division 4 of my 18th Amateur Series, Romichess is likely to be ineligible (unless a new version is released) due to the 2 year rule.
No pressure!

Cheers,
Graham.