what about bgblitz backgammon software?

[webrunner]

I always read about snowie and gnubg but what about BGBlitz?
It looks like a worthy alternative.
Does anyone have experience with it?

[stiefnu]

I find BGBlitz very useful, though cannot offer a personal comparison with GnuBG because I have been unable to download that to my Mac.  However, BGBlitz is certainly a very powerful program, with excellent graphics.  It has won the Computer Olympiad on three occasions, in 2002, 2003 & 2007.  Check out http://www.bgblitz.com/

I bought the Professional Version for $46, which includes a rollout facility and a tutor giving comments on cube handling and game play.  Further features (e.g. Match analysis) are under development.

stiefnu

[webrunner]

I know all this, i think it is a great programm too. But i was wondering why nobody aver talks about it here on fibsboard. It has great functionality for a reasonable price.

[PersianLord]

I find BGBlitz very useful, though cannot offer a personal comparison with GnuBG because I have been unable to download that to my Mac.  However, BGBlitz is certainly a very powerful program, with excellent graphics.  It has won the Computer Olympiad on three occasions, in 2002, 2003 & 2007.  Check out http://www.bgblitz.com/

I bought the Professional Version for $46, which includes a rollout facility and a tutor giving comments on cube handling and game play.  Further features (e.g. Match analysis) are under development.

stiefnu

hmmm.. I'm getting interested. As I had said, I have some problems with gnubg and was thinking of snowie. But if BGBlitz is equally strong, I'll definitely go for it. $380 is way more then $46. Isn't it?

[spielberg]

Gosh BGblitz looks nice. As a Windoze (stet) user I oftentimes use Gnubg but BgBlitz seems to have a friendlier interface and an easier to use tutor mode. At US$46 for the professional edition this is a bargain for those not wishing to pay a certain car salesman $10 for a ripped off Snowie.

[stiefnu]

I know all this, i think it is a great programm too. But i was wondering why nobody aver talks about it here on fibsboard. It has great functionality for a reasonable price.

Probably because few people will pay good money for something when GnuBG is freely available, Webby.  Also, as I mentioned, the program is limited, in that there is currently no match analysis available.  While it's tutor facility is excellent and its playing mode cannot be faulted, it will not yet tell me how lucky I was, or at what amazing level I just played against it ;-)

stiefnu

[FrankBerger]

Hi,

one user made me aware of this thread  (tnx BTW)

I know all this, i think it is a great programm too. But i was wondering why nobody aver talks about it here on fibsboard. It has great functionality for a reasonable price.

Yes, this is a little mystery to me  . There are some areas where BGBlitz excels, but it is often ignored or even
unknown to people where I think they ought to know. It sometimes seems to me that there is a unwritten law that there could be only 2 well known programs

Probably because few people will pay good money for something when GnuBG is freely available, Webby.  Also, as I mentioned, the program is limited, in that there is currently no match analysis available.

I don't think it is the money. The player costs just $23 and with the current exchange rate that is approx. the amount one pays for the drinks at an evening, and you get free updates for years. Even $46 is less thn some people play per point.
I guess the analysis is the main point and I'm working on it. The roadmap is:

Version 2.6 brush up the rollouts (end of spring)
Version 2.8 Match analysis (end of autumn)

anywhere inbetween userstatistics, a brushed up user interface and a new AI (well this is ready. approx 0.6-0.7 points in 100 cubeless games).

But somehow things are strange anyway. As an example BGBlitz2go. It is by far the best BG you can get for your mobile but it's impact is negligible (except by the people that have it already. One user had even to go to the meds. He was playing in every free minute and due to the permanent looking down he got problems with the cervical spine. So much to how addictive it is)

But the ignorance just motivates me to work harder on BGBlitz.

ciao
Frank

[PersianLord]

Hi,

one user made me aware of this thread  (tnx BTW)
Yes, this is a little mystery to me  . There are some areas where BGBlitz excels, but it is often ignored or even
unknown to people where I think they ought to know. It sometimes seems to me that there is a unwritten law that there could be only 2 well known programs


I don't think it is the money. The player costs just $23 and with the current exchange rate that is approx. the amount one pays for the drinks at an evening, and you get free updates for years. Even $46 is less thn some people play per point.
I guess the analysis is the main point and I'm working on it. The roadmap is:

Version 2.6 brush up the rollouts (end of spring)
Version 2.8 Match analysis (end of autumn)

anywhere inbetween userstatistics, a brushed up user interface and a new AI (well this is ready. approx 0.6-0.7 points in 100 cubeless games).

But somehow things are strange anyway. As an example BGBlitz2go. It is by far the best BG you can get for your mobile but it's impact is negligible (except by the people that have it already. One user had even to go to the meds. He was playing in every free minute and due to the permanent looking down he got problems with the cervical spine. So much to how addictive it is)

But the ignorance just motivates me to work harder on BGBlitz.

ciao
Frank

Hi Frank.

What a surprising story. I was reading this page http://www.bgblitz.com/olympiad_2007.html and was seeing your pics and when I returned to forum, I saw your post! 

If BGBlitz beat GNUBG's grandmaster, there would be NO reason to not buy it. I will definitly but the full version. $46 is nothing, really nothing for such a strong software. I had read that GNUBG had beaten Snowie using it's expert level by a score of 56-44. Now I wonder how badly Snowie would be defeated by BGBlitz? I'm really fond of these software vs software battles and if you can provide us with more details, I'd be delighted.

Thank you again Frank and keep the good work up man. With a little more strong advertising campaign, BGBlitz will be the best-seller BG software. A nice proposal would be affilliate programs and paying webmasters (like me  , kidding) who advertise for BGBlitz.


Regards

PS:

As you may know, many anti-virus and anti-trojan softwares have problems with java and often prevent it from running or downloading. Is offering BGBlitz in a non-java format among your future plans? That would be great.

[stiefnu]

Version 2.8 Match analysis (end of autumn)

That's great news Frank!  As to the price, this is certainly not expensive ~ I was just suggesting a possible reason, in answer to Webrunner's question.

One thing which would be good to be able to do would be to run saved FIBS matches through BGBlitz.  There is an Import FIBS/IBPM Match (text) option but it does not seem to recognise JavaFIBS .match files.  Is this something that maybe will come along with 2.8?

stiefnu

[FrankBerger]

One thing which would be good to be able to do would be to run saved FIBS matches through BGBlitz.  There is an Import FIBS/IBPM Match (text) option but it does not seem to recognise JavaFIBS .match files.  Is this something that maybe will come along with 2.8?

Just looked at JavaFIBS. You can do this already:
If you have selected that matches are saved (Preferences/Log games), they will be saved
in the internal format. When you select "Tools/Matchconverter", select the match you want and press "convert" you find the match in Jellyfish-format beyond JavaFIBS in the folder matches/Jellyfish.

Does this solve your demand?
ciao
Frank

[webrunner]

Thanx for joining this topic Frank 

[spielberg]

WARNING : off topic

A quick point for all JavaFibs users: the new update, currently being tested, offers the ability to save matches in Jellyfish format automaticaly. It's also much more user friendly for checking repbot's data. There are a few remaining bugs which are being solved as I type. I must stress it's not me doing the update - inim, Tom and Cthulu deserve all the credit.

What was already the best front end for fibsters is about to get a whole lot better.

[FrankBerger]

If BGBlitz beat GNUBG's grandmaster, there would be NO reason to not buy it. I will definitly but the full version. $46 is nothing, really nothing for such a strong software. I had read that GNUBG had beaten Snowie using it's expert level by a score of 56-44. Now I wonder how badly Snowie would be defeated by BGBlitz? I'm really fond of these software vs software battles and if you can provide us with more details, I'd be delighted.

Well the Computer-Oympiad is a fun thing and exiting, but it doesn't resolve the question who plays better. I personally believe that Gnu plays a little better (my bet is somewhere between 5-15 rating points), really no idea related to Snowie just that the difference will be even smaller so for all practical purpose it doesn't matter (Does it make a difference whether the car that knocks you down drives with 80, 81 or 82 mph?) There are only two benchmarks I know about that I feel bring some more light in this, that's the bencmark of Michael Depreli and the "Big Bot Shotout" by transeuropaexpress.
Oh... and BGBlitz isn't on hold on that. 2.6 will bring a new AI that wins 503:497 on 1000 cubeless games with 1-ply (deviation 0,15 deviation with 95%, The benchmark was 4 million games) that will close the gap further.

And, pretty important, strenght and weaknesses of the bots are different, so it pays to crosscheck their suggestions anyway.

As you may know, many anti-virus and anti-trojan softwares have problems with java and often prevent it from running or downloading. Is offering BGBlitz in a non-java format among your future plans? That would be great.

I'm really not aware of any problems with Anti-Virus software... but that only means that no one told it to me. I was aware that the bad guys uses e.g. a certain installer and have therefore choosen another one and an older version of BGBlitz had some permission issues with Vista but that's all I'm aware. Delivering as a native programm is not easy, there is only one way I know about (Excelsior Jet) and that software in the version I need is beyond my budget. 

ciao
Frank

[PersianLord]

Well the Computer-Oympiad is a fun thing and exiting, but it doesn't resolve the question who plays better. I personally believe that Gnu plays a little better (my bet is somewhere between 5-15 rating points), really no idea related to Snowie just that the difference will be even smaller so for all practical purpose it doesn't matter (Does it make a difference whether the car that knocks you down drives with 80, 81 or 82 mph?) There are only two benchmarks I know about that I feel bring some more light in this, that's the bencmark of Michael Depreli and the "Big Bot Shotout" by transeuropaexpress.
Oh... and BGBlitz isn't on hold on that. 2.6 will bring a new AI that wins 503:497 on 1000 cubeless games with 1-ply (deviation 0,15 deviation with 95%, The benchmark was 4 million games) that will close the gap further.

And, pretty important, strenght and weaknesses of the bots are different, so it pays to crosscheck their suggestions anyway.
I'm really not aware of any problems with Anti-Virus software... but that only means that no one told it to me. I was aware that the bad guys uses e.g. a certain installer and have therefore choosen another one and an older version of BGBlitz had some permission issues with Vista but that's all I'm aware. Delivering as a native programm is not easy, there is only one way I know about (Excelsior Jet) and that software in the version I need is beyond my budget. 

ciao
Frank

Thank you for the kind response.

But I didn't realize this :

Oh... and BGBlitz isn't on hold on that. 2.6 will bring a new AI that wins 503:497 on 1000 cubeless games with 1-ply (deviation 0,15 deviation with 95%, The benchmark was 4 million games) that will close the gap further.

Against whom BGBlitz wins? Would you elaborate plz?

Thanks

[stiefnu]

Just looked at JavaFIBS. You can do this already:..........
Does this solve your demand?

Yes Frank, indeed it does!  Thank you!  

stiefnu

[FrankBerger]

Oh... and BGBlitz isn't on hold on that. 2.6 will bring a new AI that wins 503:497 on 1000 cubeless games with 1-ply (deviation 0,15 deviation with 95%, The benchmark was 4 million games) that will close the gap further.

Against whom BGBlitz wins? Would you elaborate plz?

with pleasure.

When I develop a new AI I test it against the current AI. To get statistical relevant data you have to play a very large number of games. A test e.g. with Tony Lezards Dueller would take much to long. To give you a feeling for that: when you play 1.000.000 cubeless games the deviation that is statitically not significant is 0,003 i.e. when the result is 501.499:498.501 you can't conclude that the new AI is better (with more than 95% confidence), it could be a random deviation. Therefore 1-ply (rest-of-the-world-counting, for GnuBG ist's 0-ply) and self play is the only way to get results in a reasonable amount of time. If I would use cubes in the benchmark, the standard deviation increases a lot and I would need much more games (and it is also a albeit small possibility that the standard deviation in cubeful game isn't defined at all).

So the new AI coming with 2.6 wins against the current AI roughly 503:497 (that is more descriptive than an equity of 0.0067) in 1000 games. The equity value was calculated by 4.000.000 games so the statistical significance is very high and the 95% confidence interval is 0.0015 so the "true" value is with 95% probability between 0,0052 and 0,0082.

That sounds not too impressive, but the level is already very high so any improvement gets more difficult. But be sure, I'm working on it

[PersianLord]

with pleasure.

When I develop a new AI I test it against the current AI. To get statistical relevant data you have to play a very large number of games. A test e.g. with Tony Lezards Dueller would take much to long. To give you a feeling for that: when you play 1.000.000 cubeless games the deviation that is statitically not significant is 0,003 i.e. when the result is 501.499:498.501 you can't conclude that the new AI is better (with more than 95% confidence), it could be a random deviation. Therefore 1-ply (rest-of-the-world-counting, for GnuBG ist's 0-ply) and self play is the only way to get results in a reasonable amount of time. If I would use cubes in the benchmark, the standard deviation increases a lot and I would need much more games (and it is also a albeit small possibility that the standard deviation in cubeful game isn't defined at all).

So the new AI coming with 2.6 wins against the current AI roughly 503:497 (that is more descriptive than an equity of 0.0067) in 1000 games. The equity value was calculated by 4.000.000 games so the statistical significance is very high and the 95% confidence interval is 0.0015 so the "true" value is with 95% probability between 0,0052 and 0,0082.

That sounds not too impressive, but the level is already very high so any improvement gets more difficult. But be sure, I'm working on it

Many thansk for the elaborated reply.

I'm going to buy BGBlitz in the coming days.

Regards

[webrunner]

So guess this topic has opened some eyes

[PersianLord]

So guess this topic has opened some eyes

Yes, and thanks for that. 

I have been in search of a solution to i) having a  more bug-free, or at least less trouble-making sofware then gnubg and ii) not paying $380 for snowie.

Now BGBlitz is the solution and now I wonder why didn't I find it sooner.

Regards

[Hardy_whv]

I have been in search of a solution to i) having a  more bug-free, or at least less trouble-making sofware then gnubg ...

Just to clarify that: If your computer crashes while using GNUBG, I believe, it's a general problem of your computer and not of GNUBG. I have one computer running for more than 3 months now without shutting it down once, two instances of GNUBG running on it since that time without any problems. GNUBG is running on 5 computers here at my home (Windows 2000, Windows XP and Linux [Ubuntu]) and none is having any problems because of GNUBG.

Hardy 

[playBunny]

<Something about "good, but intuitive" players" and "better, but analytical" ones....>
Posted by <a player> at Wed Mar 12 11:28:29 2008

@<a player>: ~~Using the math usually separats the "good, but intuitive" players from the "better, but analytical" ones.

It's more like "good intuitive players" and "better intuitive players who are analytical as well". You can't remove intuition from the picture. Attempts to program backgammon bots analytically have failed to produce anything near the strength of play of the intuitive bots (such as GnuBg and BGBlitz).**

**My new policy of mentioning BGBlitz instead of Snowie. Yay, let's hear it for BGBlitz! ;O)

Posted by playBunny at Wed Mar 12 11:51:50 2008

The above was copied from the DailyGammon forum.

Hey there Frank, you should see if you can get more people to do the same as me.

Might I suggest that you create an account at DailyGammon. Here are a couple of posts about the perception of BGBlitz out in Backgammon Land which may be interesting to ponder on.
http://www.dailygammon.com/bg/forum2/main/read/23213#90

[FrankBerger]

The above was copied from the DailyGammon forum.

Hey there Frank, you should see if you can get more people to do the same as me.

Might I suggest that you create an account at DailyGammon. Here are a couple of posts about the perception of BGBlitz out in Backgammon Land which may be interesting to ponder on.
http://www.dailygammon.com/bg/forum2/main/read/23213#90

Thanks for that hint. I post there in a minute