Comparison of the two versions

Discussion about development of draughts in the time of computer and Internet.
Post Reply
Krzysztof Grzelak
Posts: 1307
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Comparison of the two versions

Post by Krzysztof Grzelak » Fri Sep 16, 2022 20:56

How best to compare the two versions of Flits. I mean the versions from Turbo Dambase (Flits 3.1) and versions 3.02.

I am interested in several answers:
1. which of these versions is newer
2. which of these versions is better
3. if the Turbo Dambase version has bugs that are in Flits 3.02.

Ed Gilbert
Posts: 854
Joined: Sat Apr 28, 2007 14:53
Real name: Ed Gilbert
Location: Morristown, NJ USA
Contact:

Re: Comparison of the two versions

Post by Ed Gilbert » Fri Sep 16, 2022 21:06

From my conversations with Klaas Bor, they are probably very similar. The version in TDam can use the kingsrow egdb, but that's not going to make much difference in tournament play. There's a search bug in Flits where it sometimes makes bad moves in simple endgames, and it does this in both versions. Klaas didn't change any logic in the search or eval.

Krzysztof Grzelak
Posts: 1307
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: Comparison of the two versions

Post by Krzysztof Grzelak » Fri Sep 16, 2022 21:37

Such a question occurred to me. Is the impact of errors on the database in Turbo Dambase in this case. Does the 7 and 8-piece database provide anything in this case?

Ed Gilbert
Posts: 854
Joined: Sat Apr 28, 2007 14:53
Real name: Ed Gilbert
Location: Morristown, NJ USA
Contact:

Re: Comparison of the two versions

Post by Ed Gilbert » Fri Sep 16, 2022 21:47

The 8-piece db provides better information in analysis mode. As far as the bug is concerned, it doesn't happen very often, and I don't know what triggers it. I don't even know if the bug occurs at all in analysis mode. The use of draughts programs in TDam is primarily for analysis.

Krzysztof Grzelak
Posts: 1307
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: Comparison of the two versions

Post by Krzysztof Grzelak » Fri Sep 16, 2022 21:57

Ed Gilbert wrote:
Fri Sep 16, 2022 21:47
The 8-piece db provides better information in analysis mode. As far as the bug is concerned, it doesn't happen very often, and I don't know what triggers it. I don't even know if the bug occurs at all in analysis mode. The use of draughts programs in TDam is primarily for analysis.
Well, yes, but people also play with the engine.

BertTuyt
Posts: 1573
Joined: Wed Sep 01, 2004 19:42

Re: Comparison of the two versions

Post by BertTuyt » Fri Sep 16, 2022 22:51

It is a pity that the bug still exists in both Flits versions.
Another thing, although Flits is relatively old, and basically not changed since many years, it has really potential for improvements.
If one (with the proper knowledge) would include search extensions like better pruning and LMR, I'm convinced Flits could reach the level of the top programs.

The same holds for Truus, and many others.

So i hope that Adri (the author of Flits) gives permission to do so.
Much work was involved in these programs, and with relatively small changes they could perform much better.
And would make the outcome of these matches less predictable and far more interesting, as in the current situation, in the end, the usual programs surface towards the top positions.

Assume Ed supports me here.

Bert

Sidiki
Posts: 313
Joined: Thu Jan 15, 2015 16:28
Real name: Coulibaly Sidiki

Re: Comparison of the two versions

Post by Sidiki » Fri Sep 16, 2022 23:28

Hi all

Sorry
Last edited by Sidiki on Sat Sep 17, 2022 09:12, edited 1 time in total.

Ed Gilbert
Posts: 854
Joined: Sat Apr 28, 2007 14:53
Real name: Ed Gilbert
Location: Morristown, NJ USA
Contact:

Re: Comparison of the two versions

Post by Ed Gilbert » Sat Sep 17, 2022 00:09

BertTuyt wrote:
Fri Sep 16, 2022 22:51
It is a pity that the bug still exists in both Flits versions.
Another thing, although Flits is relatively old, and basically not changed since many years, it has really potential for improvements.
If one (with the proper knowledge) would include search extensions like better pruning and LMR, I'm convinced Flits could reach the level of the top programs.

The same holds for Truus, and many others.

So i hope that Adri (the author of Flits) gives permission to do so.
Much work was involved in these programs, and with relatively small changes they could perform much better.
And would make the outcome of these matches less predictable and far more interesting, as in the current situation, in the end, the usual programs surface towards the top positions.

Assume Ed supports me here.

Bert
Indeed, as we discussed earlier, many of the programs play very well until they get into late midgame positions with around 18 pieces remaining, and then they sometimes make a losing move. From the symptoms it appears that they do not search deep enough in this stage of the game. It's likely that adding a search reduction technique like LMR would allow these programs to search substantially deeper, and perhaps avoid many or even most of these losses. If Adri was willing to let us, I could work with you to make some improvements to Flits, including fixing the endgame bug.

-- Ed

Krzysztof Grzelak
Posts: 1307
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: Comparison of the two versions

Post by Krzysztof Grzelak » Sat Sep 17, 2022 07:13

Ed Gilbert wrote:
Sat Sep 17, 2022 00:09
Indeed, as we discussed earlier, many of the programs play very well until they get into late midgame positions with around 18 pieces remaining, and then they sometimes make a losing move. From the symptoms it appears that they do not search deep enough in this stage of the game. It's likely that adding a search reduction technique like LMR would allow these programs to search substantially deeper, and perhaps avoid many or even most of these losses. If Adri was willing to let us, I could work with you to make some improvements to Flits, including fixing the endgame bug.

-- Ed
I will ask if the author of the program will allow it.

Krzysztof Grzelak
Posts: 1307
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: Comparison of the two versions

Post by Krzysztof Grzelak » Sat Sep 17, 2022 07:14

Sidiki wrote:
Fri Sep 16, 2022 23:28
Hi all

Sorry for be late, a friend Luzimar, fixed this endgame bug of flits, and it's more stable and more strong.
Sorry again to remember at the end of the tournament.

This's the link that I done for this fixed with endgame database.
You can test the difference with some endgame positions.


https://pixeldrain.com/u/AKAy1QTU
You write that you fixed all the mistakes in how you did it.

Sidiki
Posts: 313
Joined: Thu Jan 15, 2015 16:28
Real name: Coulibaly Sidiki

Re: Comparison of the two versions

Post by Sidiki » Sat Sep 17, 2022 09:11

Krzysztof Grzelak wrote:
Sat Sep 17, 2022 07:14
Sidiki wrote:
Fri Sep 16, 2022 23:28
Hi all

Sorry for be late, a friend Luzimar, fixed this endgame bug of flits, and it's more stable and more strong.
Sorry again to remember at the end of the tournament.

This's the link that I done for this fixed with endgame database.
You can test the difference with some endgame positions.


https://pixeldrain.com/u/AKAy1QTU
You write that you fixed all the mistakes in how you did it.
Hi Krzysztof,

Please read carefully messages before writing an answer.

Friendly, Sidiki

pontel
Posts: 41
Joined: Tue Jan 26, 2021 21:48
Real name: João Anselmo Pontel

Re: Comparison of the two versions

Post by pontel » Sat Sep 17, 2022 18:53

Krzysztof Grzelak wrote:
Sat Sep 17, 2022 07:14
Sidiki wrote:
Fri Sep 16, 2022 23:28
Hi all

Sorry for be late, a friend Luzimar, fixed this endgame bug of flits, and it's more stable and more strong.
Sorry again to remember at the end of the tournament.

This's the link that I done for this fixed with endgame database.
You can test the difference with some endgame positions.


https://pixeldrain.com/u/AKAy1QTU
You write that you fixed all the mistakes in how you did it.
Hello Grzelak!

It seems that our friend Luzimar noticed that the milliseconds for the answers were too short, and this was embarrassing some searches in the finals base, there was only a small time delay for the answers.

I was able to test the positions I have where Flits made mistakes, if you have others, I suggest you keep testing. He follows:

[Event "Ending Error - Flits"]
[Date "2022.09.17"]
[White ""]
[Black ""]
[Result "1-1"]
[FEN "B:W20,33,37,42:B11,15,27"]
*
After the capture 15:44 delivered the piece by 37-31 and ended up losing the game that was in a tie position.


[Event "Ending Error - Flits"]
[Date "2022.09.17"]
[White ""]
[Black ""]
[Result "1-1"]
[FEN "B:W28,30,31,44:B11,18,25"]
*
After the capture 25:34 delivered the piece by 44-39 and ended up losing the game that was in a tie position.

Post Reply