While eXtreme Gammon has some features that are an improvement over Snowie, has fared better in some comparisons and is less expensive, it is still new and may have yet unknown bugs in it.

When Windows XP came out, Microsoft made claims about how it was orders of magnitude more stable than previous versions of Windows but it soon became clear that it had stability problems (which Microsoft sorted out over time). My understanding is that the Window Vista as installed on computers would essentially be replaced with a stable version through "updates" once you booted up your computer.

Snowie, despite it age, may have bugs in it as well, and as best as I can tell there has not been any update to Snowie in over 2 years, including bug fixes. From all indications development has stalled. On the other hand XG's developer has been amazingly responsive to bug fixes and feature additions, releasing updates frequently, often to fix on one of two minor issues. XG's developer is also currently actively working on version 2.

Snowie supporters also seem to ignore the fact that the current version does not do rollouts for match play positions at score, which many would consider a major flaw.

"XG does better in some comparisons" implies that Snowie does better in other comparisons; where are they? Right now Michael Depreli's work seems to be the only comparison with any kind of systematic approach, and in that comparison XG certainly comes out on top. Are there other comparisons?

Is there any evidence to support that Snowie is better at backgames than XG?

So far the only reasons given for buying Snowie is its interface and backgame play. I have not used Snowie, but I really doubt that its features and interface are so much better than XG's that most people would consider it worth an extra $310. If it was proven the Snowie's backgame play is better than some people might consider that worth the extra money, but I think the vast majority of players would question whether better backgame play was worth the money.

MordellWeil said: "There is no longer any reason at all to buy Snowie."

Not a rational reason, no, but there is nonetheless a very powerful psychological motive for still using and recommending it: some of those people who spent $400 back in a lost age when it was the best, very naturally put on defensive blinkers when confronted by the distressing truth that there is now a far superior program at an eighth of the price. It is 'the pain of progress' at work, and causes them to say deeply bizarre things like "it may have unknown bugs..."

FWIW Snowie and JF are both dead to me. It's like asking "I have a 386, should I buy Windows 3.1 or Windows 95?"

And if anyone would like to back Snowie vs. XGR+ in a backgame of your choosing I am willing to back eXtreme.

Stick Rice Proto-Back game No. 11 making the 5 point vs. the 2 point (Phil Simborg Homepage). Stick Rice said, that he don't trust bots in backgames and that he is sure because of his deep understanding of backgammon, making the 2 point is right (20/15 Bar/2). Snowie evaluation disagreed. But in 2ply and 3ply rollout Robertie settings with live cube the old turd snowie agreed to Mr. Rice. I am cautious. If you are interested, i'll send you the rollout. The 3ply rollout lasted on a core 2 duo E5400 overclocked to 3,4 Ghz 24 hours. This is the limitation of snowie, his slowness. But else? By the way, Mr. Rice don't answered me on my email with rollouts. Perhaps, there was no money to earn. After all, the beginning and the end will head back into the void. Think about it.

Rollouts aside, does anyone think Bar/20 makes sense here? Bar/23 is mandatory in my mind.

In this case your example supports that XG would be a much better program for backgames than Snowie. Accepting that Stick's answer is correct--and I am not trying to argue that point-- XGR+ comes up with the right answer of bar/23 20/15 in a matter of seconds. So, in this case anyway, which program would be considered superior, the one that gets the right answer in a matter of a few seconds or one that requires a 24 hours?

If you are looking closer, you can also find arguments for making the 5 point.
But you are comparing apples with pears. A snowie 3ply evaluation is not XGroller+. XGroller+ is a sort of minirollout. XG performs here better because of multicore support. And if we refer to Frank Berger from BGBlitz, we have the AI and the framework of the AI. If you wrote an AI, you can for example use BGBlitz as platform to run. The platform flying is the main problem and manpowertime in programming. As snowies platform was written, multicore machines were beyond the horizon. But i think, the AI of snowie is as good as XG, perhaps even better in deep backgames. XGroller+ and other rollout modes are a question of platform. The snowie heart of engine isn't out of time, which in my opinion cranked up rollouts have proved. If you can lure someone in a bot proposition to make money, this will fine under the aspect to make a living with backgammon knowledge. but it isn't honest, because of the explanations above.
And don't forget, there is no reference bot. Even the concept of volatility isn't in my point of view really solved.

And something for you to understand:
3ply 648 trials truncated after reaching database with live cube 3ply is a very deep and sound setting to get clear. But it is time consuming. Even in a 2ply snowie came up with the right answer. And if we set away the live cube it gets even faster. Looking for the right rollout setting is a world of it's own. Don't compare apples with pears. Perhaps we should ask the opinion of the experts, what AI is better. But that will be then only the opinion of a master. Look at the postings of Frank Berger in the fibs forum about the principle problems. And my point of view is perhaps a little bit philosophical, not money making.

The core question is whether there is any reason to purchase Snowie? The only reasons given thus far have been it's interface and ability to play backgames. The interface question is more of a personal preference matter, if someone thinks the interface is worth the additional price then that is there choice. But the real question is does Snowie do better with backgames than XG. I suppose one could argue that if Snowie 2 ply, 648 game rollout gets the right answer while XG requires a 3 ply, 1298 game rollout to get the correct answer than Snowie's overall backgame AI is better. But, from a practical standpoint if I can run the 3ply 1298 XG rollouts in a quarter of the time as I can the 2 ply Snowie rollouts then there would still be no practical need for Snowie.

I realize that making the above claims are somewhat over-simplistic because it would require testing a large number of backgame positions with "known" solutions to conclude that either rollout setting actually gives reliable results.

I am not sure comparing XGR+ to a 648 game 2 ply rollout is really all that apples to pears because one could argue that both are mini rollouts

Now we are coming nearer to the heart of darkness. Thinking, that one AI is better then the other, is, even if you are a B-52, an educated guess. Deep backgames are the most complicated things in backgammon. We are talking about decision making. And for that, Herbert Simon wrotes: "behavior was shaped by a scissors whose two blades were the structure of task environments and the computional capabilities of the actor." It is a sort of bounded rationality. Choice, as the evaluation of options and the selection of actions depends on the structure of information (task environment). Thus, framing the information, influences the information. Judgement, defined as the processing of information and estimation of probability, can have it's biases, feeded by emotion, education, culture, experience, and world view. For example, how much experience do we really have in playing deep backgames? Have the bots not told, that this is a goal-line stand and has to be avoided? I think, it is impossible, to said Snowie or is XG is for sure better to play backgames. Then you have to know the right answer for yourself. And the decision, what to buy is a question of your personal goals and your measures to reach them. If you have a limited budget, XG might be fit. If you want to research, perhaps both bots. And you have to look for BGBlitz.

Snowie: slow apple
XG: fast apple
quasi-metaphysical discussions of backgames: pear

quod erat demonstrandum

I hoping one of the bots implements graphics co-processer to double + the speed.

It is not metaphysical (perhaps the question if we have by now already a pipcelot machine). But it is not honest to say this or that bot is better, to get a prop. You have to know, what you want. And Snowie can be still a choice. I am working with gnu, snowie and xg. Neil Kazaros favors gnu and xg, because he works with the developer of xg, Bill Robertie favors Snowie, because he is familiar with this bot and he sells it. So that is bounded reality. But if you make a book out of the problems of the week, you have a very fine book about backgammon and a fragmented theory of back games based on snowie research. Everything depends on your goals. For serious goals, only jellyfish is outdated. But to have a bot to play with for fun and training, you can still trust him. If you are honest, you will see, that snowies problem is his platform, not the AI. If you want a speed machine, you have to take xg with an overclocked core i7 gulftown (for 4.000.000 positions per second). So goal orientated decision making under uncertainty.

The whole point is the OP is obviously a fairly new player that is trying to determine the bot they should be using. The answer they got basically breaks down to Jellyfish is an okay playing opponent but if you want to have a bot that will really improve you then shelling out $360 for Snowie is necessary and the "array of wonder features" make Snowie "well worth the price." I think it is very poor advice to tell someone that is looking for a bot these days that they should spend that much for a bot when there are very good options that are either free or $50. Perhaps if Snowie is better for deep backgames (and that as we have determined is questionable) than someone doing really complex analysis might have a need for it. But for 95%+ of bot users there is no need to waste over $300 on Snowie and to not give them the options is, in my mind, damaging both the reputation of the person giving the advice and to backgammon in generally.

Dear BigWill, if you have read all of my last post, i have everything said. And the "homeplayer" can even take jellyfish or gnu for free. But if he will improve, and if he is not a genius like Bill or Stick or Phil, then he has to buy also several books, to understand, what the machine tells him. It isn't just buy XG for 50 bucks and you will get a B-52. Have you worked yourself through the 77 problems of the week? By far i would know no backgammon book, which contains more material in so detailed analysis about back games (snowie research). And a back game is a goal-line stand, but if you want to play competitive backgammon, you have to know how to defend a back game and how to make decisions in a proto-back game. You can easily dump 350 bucks in one session or one tournament, so why whining about 350 bucks for a software? What are you burning every month with your car? The homeplayer should take gnu and try to be happy with the crappy interface. Goal-orientated decision-making, i can only repeat.

I am reading your post, and getting more and more confused as I do. I was talking about a new player, I am not sure were "homeplayer" came from. And I am certainly not opposed to spending money on backgammon. I have several books and drive 150 miles round trip every week just to get to a club to play live! I never said not to buy books, or that somehow XG alone would give you all the answers. However, I think many (if not most of us) no matter how serious we might be about the game have a limited amount of funds to spend on it. $300 may not be a lot for many of us in the whole scheme of things, but most of us would be pretty upset to find out that we spent $300 needlessly. To tell someone that Snowie is "well worth the money" without even telling them about free and much less expensive options is the issue that I have. And when one of those options is faster, arguably stronger, and is the program of choice of many of the best players in the world then failing to tell someone to consider it does a major disservice to them.

Yes, if one is doing significant backgammon research than running XG on a multicore, overclocked machine is the way to get a speed machine. But, even if you are not investing in a super faster machine XG is going to allow you to do faster, deeper, analysis on any given machine than Snowie. Why be limited to 648 game 2-ply rollouts when you have a tool that can give you much longer and deeper rollouts in less time? The only reason would be if the longer, deeper rollouts are somehow less reliable, but no one in making that case. And the Snowie supporters still want to ignore the lack ATS rollouts!

I am also not sure quite sure what you point about the 77 problems posted here is. It seems that you are arguing that among these 77 problems is a collection of backgame problems that taken together form an excellent resource for playing backgames, and that because these are based on "snowie research" it argues that we should continue to use Snowie. I have not been through all 77 problems, but I will grant for the sake of argument that among them is a collection of backgame problems that form the best available resource for backgames. But just because Snowie was used for this research is no reason to continue to use Snowie. Unless we can show that Snowie is better for backgames then XG (and you seem to say that we can't be sure of this) then the same body of work could have been developed using XG had it been around for the whole time.
Otherwise you argument is pretty much the same as saying that because quill pens were used to write much of the greatest literature in the world anyone wanting to write great literature should still be using a quill pen.

Saying the problem with Snowie is the platform not the AI is pretty much the same as saying the problem with GNU is the interface not the AI! Personally I think the whole crappy interface argument against GNU is blown out of proportion. If one skims the online GNU manual for a few minutes it is really quite easy to pick up the interface.

So the core question still comes down to why is Snowie worth $310 more than XG, I still have not seen a good answer to that one.

Finally, to set the record straight, Phil Simborg readily admitd that he is not in the same class as Stick or Bill. Phil is a strong player, but not a world class player. His strength lies more in his passion for teaching and promoting the game.

I wouldn't download Snowie if it was free.

Just one opinion.

+1 to your excellent posts, BigWill, but one correction of fact: I've just checked, and the price of snowie seems to have come down to $85, so players like the OP will only waste $36 buying this antiquated program, instead of the mustn't-mention-it-because-we-don't make-money-from-it XG. Also, your post gives the impression the speed difference is marginal, or that one needs a really high-end machine to do it justice. I have a stone-age Core Duo, but XG still gives me a 1296-game variance reduction rollout, in the background while I'm playing, of every error and blunder I make. This isn't just faster than snowie / gnubg, but on a different planet. I also find it far more attractive, intuitive and versatile to use, and my interest in the game has quadrupled since buying it.

I just checked the Snowie site and the $85 is for the student version, which is a limited version (for example it does not do rollouts). The full version is still listed at $365. The prices from Gammon Press are $15 higher for each version because they are the cost of the CD version whether than the download version.

I did not mean to imply that the speed differences between Snowie and XG were minimal, although not having Snowie I cannot do an exact comparison. My point was people using Snowie are limiting themselves to short 324 or 648 game 2 ply rollouts and limited evaluations because to do otherwise is so slow. Using XG one can get results using much longer rollouts and deeper settings in significantly less time than the limited Snowie rollouts. That is what I intended to say.

How often do you guys beat these programs? Just curious because gnu is currently destroying me over quite a few series.

well aside from very specific backgame leaks gnu is far beyond any human player in a long run
the point is not to beat it but to improve your game, although I wouldn't lie, it feels good every time I beat it in 17-point match, which doesn't happen too often