Hey guys just wondering your opinion on this. I am just getting into backgammon again after a couple year hiatus to focus on poker. Now that I'm getting serious again I'm wondering if I'm getting to the point that I need snowie and to get more involved in posting in forums and get involved to really step my game up. After a warm-up week my results from last week seem to have me at about a 9.3 SER.(Though I would like to have about 10 more games in my sample size to make me feel a little more comfortable with that number but its probably close from what I can tell. Right now I'm analysing with GNU on 3ply and my very bad moves showing up at .03. Is snowie worth my money?

Stick with GNU or try Extremegammon. Both are a bit stronger than Snowie at this point and a lot cheaper.

Extremegammon is the strongest bot right now (so the studies say anyway), and it only cost $50. However Extremegammon 2 is probably close to it's release, so you might want to wait a couple of months.

I don't remember exactly what terminology GNU uses for errors/blunders/bad moves/very bad moves etc., but 0.03 is what I would call a small mistake (but still a mistake worth caring about).

I would recommend 2-ply for analysis (use the predefined world class or supremo setting) instead of 3-ply, which actually is a bit weaker and slower than 2-ply.

Thanks, I didn't realize 3ply was worse, I just thought it was better. Hopefully I haven't skewed my results with that. And I run .01 doubtful .02 bad and .03 very bad for my mistakes.

Any other training tips? I would like to be down to 6 ER by Christmas. I have started reading more, playing more, and doing 10 training games with GNU and analyzing the play.

Grab a copy of the Backgammon Encyclopedia. Its a book of cube reference positions. The book is very clear, and systematic in its approach. It really helped me drop my ER when I was floating around between 8-10. For checker play you should definitely read Modern Backgammon by Robertie. Very strong book. Also, at this point when your ER is in that range I think that you should be playing fewer games before going through and analyzing. That way you can focus more on fewer errors, without being overwhelmed. Try 5 games at a time (if just cash games).

Yeah I have three new books on the way. Lots of the matches I play are on Daily Gammon so I can sit and think about my positions before I make my move but I will be scaling back here to absorb some of this information.

On a side note I kind of understand Snowie ER's and somewhat the ranges that one needs to be in to be world class or the best in the world, but what is the difference between one point in the SER. Is an 8 that much better than a 9 or does it get more difficult the better one gets. As in the difference between and 8 and 9 is less than between a 5 and 6?

3ply is hardly much worse, so it shouldn't matter much, but it still is slower, so no need to use it.

You can set your boundaries for errors anyway you like of course, but you are gonna make a lot of "very bad" moves with those settings. Something like doubtful: 0.02, bad: 0.05, very bad: 0.10 would sound more reasonable to me.

Maybe however in my mind any error is an error and my goal is to only make top plays. Maybe I'm being to harsh on myself but that is what I want to do though I know it is pretty much impossible to play perfect every move.

You want to keep in mind that your error rate will fluctuate wildly over short sessions. If your true error rate is, say, about 7, you'll have sessions where it goes down to 4 or so, and others where it's over 10.

I think the best approach for learning at your stage is to play 10-game money sessions (not matches) and average the results over 10 sessions. Then play another 10 sessions and average those results, and so on. Over time your average for each 10-session group should show a slow decline.

Why money games instead of matches? It makes learning the game clearer and easier. Playing matches creates all sorts of situations where the score affects the cube and checker play, and you're liable to get confused and learn things that aren't really true. Stick to money games for the first few months, then when you have a solid foundation, start to play matches. (This is why almost all the problems in my weekly problem quiz are for money games. I could add tricky match score situations and make them all harder, but they're hard enough as is.)

Thanks Mr. Robertie I'm going to give it my best and hopefully I'll get good enough to show up to some big time tournaments one day.

Forget Snowie. Yes it plays at a World Class level (tests show 3-ply plays around 1.2 ER) buy XG which is stronger and faster and has more features.
Some plays Snowie will flag as an error which in fact are not and some where it will agree with you which in fact are errors.
Providing you have an up to date PC, you'll be able to get a fast and accurate analysis initially using XGR++ setting or even a rollout.
$50 is a steal if you're serious about improving your BG.
IB

See I have been out of the loop. Is it worth it to wait for the supposed EBG2 or just get the EBG original?

It depends. You should be able to upgrade to XG2 for less than full price (but I don't believe the exact upgrade price has been disclosed yet). I don't know the timeplan for the final release of XG2, but the betaversion was released three weeks ago.

You should also be reading the bgonline forums. Among other reasons, the two primary XG developers(Xavier, NeilKaz) discuss XG issues there.

That is exactly what Stick says. He even said, until your PR isn't about 5, you shouldn't play matches. You need to develop a firm grip for ceckerplay and cube decisions under an even gammon price, before you go to adjust for match play.

Given that XG has been so well received and is now the de facto standard for top backgammons players, I suspect it will be split in to 2 products.
XG2 Pro and XG2 Lite (or whatever).

I wouldn't be surprised to see XG2 costing more and is probably justified given the major improvements in strength, speed and features plus there's no real competition.
Admittedly gnubg is free but progress hasn't been so great of late and the engine hasn't been significantly improved for quite some time.
BGblitz costs only slightly less than XG and overall isn't as good.
The only minus for XG is that it's a Windows only product.

IB

I have played about 120 money games with cube vs Snowie on Expert and that's about all my bg experience. My ER over those games is 9.5. Is playing this way and analyzing on 2ply going to make me learn bad habits, or is this bot strong enough so that i will improve for a long time coming?

Would you recommend setting it to higher than expert? i have noticed its mistakes on expert are mostly small and have the same tendecies (like hitting too loose in my home board early game).
Though, I am afraid that if i practice only vs a really good opp, i will not learn to play vs bad plays (like cubing way too soon/late) and when to beaver

I know virtually nothing about Snowie, but I have seen many on this site recommend setting your bot to play at a very high level. Otherwise, your bad play will be reinforced by the bot's poor replies. Don't worry about playing against inferior players. Your good play (against their misplays) will only make you win more.

FYI: I have set GNUBG to "supremo", and XG to "XG Roller+."

Just set Snowie to its highest level (3-ply) and proceed from there. The way you beat bad plays (and bad players) is to learn to play well. You want an opponent playing at as high a level as possible. Otherwise, you'll learn some mistakes along with some good ideas.

ok, let me put it this way and see if there is an error in my reasoning:

when snowie doubles me, i know it was not a double for him last turn.
so, if nothing much has changed on the board, i know it must almost always be a take now. Or at least by how much has changed, i learn a lot about if i should take or not.

however, if a bad player doubles, i even have to think about beavering, which i will never do playing vs snowie.


and similarly, if snowie beavers me, i KNOW it must have been a terrible double. whereas, if a bad player beavers me, i can still be clueless about if it was a good double.


ofcourse you will always beat a bad player in the long run, that is not the concern.

Beavers are very rare even when you play bad players. You'll pretty much never see Snowie offer a beaverable double on 2-ply, so this isn't a reason to avoid playing on 3-ply.

1 ER = 33 Elo points (approximately)

The difference between 8 and 9 is the same as the difference between 5 and 6.
8 is not much better than 9.



But yes, the lower your ER, the harder is to improve by one point, if that was your question.

Is there a thread around the place that gives a good starters approach to the different backgammon bots and training programs,. I downloaded one of them before and couldnt make heads of tales of it...

Thanks and hope this isnt a derail

Download GnuBG from HERE, and HERE you have "All about GnuBG" (It is 5 years old tutorial, so some things changed in the user interface).

Good luck!