Czuch Chuckers: Is there really only one best play for every given situation?
Yes, but. With perfect knowledge there is almost always going to be abest play, the only exceptions being when there's a tie. But we don't have perfect knowledge and nor do the computers .. yet.
Maybe what is a good play against another computer wont be a good play against an amateur?
Let's change it round a bit .. a good play against an amateur might not be a good play against a computer
Not only against an amateur but even against the different robots (though that level of subtlety in way beyond me). The "perfect" moves as reported by any of these computers is only perfect against itself. However that doesn't mean that they are wrong, it means that in certain situations they'll be slightly less than optimal. There's a huge amount of agreement between the top robots.
Against amateurs you'll find that there are definitely situations where you can make what would be a bad play against a computer or top player. For instance I'll leave a vulnerable blot in certain situations. The computer would jump on it with glee but it's safe when played against some opponents because they are too scared to do the hit. And that blot then gives me more options to hit them, or has them making poor choices in order to avoid the danger that it poses. Generally, with a player who puts too much emphasis on safety, you can get away with, and should try, things that would get you a slap on the wrist from the computer.
Putting it the way that you did, where the good play against the computer is bad against the amateur, is less frequent, I'd say. The good moves will always be good moves, even if not the best against that particular player in that game.
Interestingly Backgammon, written by Paul Magriel in the 70s, is considered to be the "Bible" of Backgammon. It's a book that many a champion has devoured and is highly recommended left, right and centre. This is despite that fact that it contains a lot of errors as proved by today's computer rollouts.
Hrqls: You had a slightly better home, he might have danced, you had good options to continue the attack or go defensive and make the barpoint anchor. The four backrunners of yours gave redsales the advantage but it wasn't conclusive by any means. Plenty of scope for you to turn the game. I wouldn't have doubled in that position - unless I had hopes of frightening you off. ;-).
Hrqls: 12/14 1/2 would have been very bad because there were two men facing you: a direct 6 and the highest indirect shot, 7. Better is to fight for the 5-point with 6/5* (20/19*) and bring down the support with the 2 - 13/11 (14/12). But best there is the double-tap/double-slot that you did.
Hrqls: That first example where you drop two blots onto your 2-point and ace point.. that's a big no-no. Blots in your home table should be working for a living - that means either preparing to make high points or hitting twice and going deep (eg 4-1: 6/5*/1*). There are times when you'd hit with two blots, such as your opponent opening with a (doubtful but not bad) 4-1: 24/20 24/21. If you then roll a 2-1 you'd hit them both. You gain a tempo plus the chance to make one or both of those valuable points on your next go. [LOL. You were busy posting while I was writing this. The redsales games is that exactly!] Gaining blocks on 2 and 1 is very inferior.
In your own game example I would have done that hit. I frequently do a double-tap with a single man from the 6 or 8 points when given the opportunity.
Hrqls: is it always best to capture 2 pieces of your opponent ?
Not always but much more often than not. The stronger your board the better because the chances of their missing a turn or of only getting one man back in increases dramatically. With 3 points blocked, the table is 75% open for a single man on the bar but with two there the chances of getting both back in is only 25%.
suppose your opponent has 2 singles in his home, is it wise to hit both, but leaving 2 singles of my own in his home ? (again, early in the game)
I'm not sure I understand you here. Blots in his home are to be snapped up in great gulps!
What's a blitz?
This is a devastating attack, often starting with a double against split backrunners, in which the table gets rapidly closed. In the worst case (for the victim) they get to make one move at the start of the game and then have to wait until their opponenent is bearing off before they get another go.
Going for a blitz usually entails risk, eg. lots of blots as builders are brought into the area to get the table closed, midpoint blot left behind, backrunners left stranded. If the victim recovers (and again doubles often feature quite strongly), the attacker can be at a disadvantage and have the tables turned on them.
Hrqls: You've got 10 half-crossovers or 5 full crossovers to get all your men home. Barb has only got 2. That gives her a lot of time in which to make up for the gap on 23 and 24. (A gap on 4 or 21 is usuually more disaster-prone)
Another way of gauging the race is with the pipcounts. 88 vs 73 has you trailing by two full rolls. Generally, in reasonably matched positions, a deficit of 8% is enough to offer the cube and you'd take when up to 12% behind, depending on how long there is to the end. You're trailing by way too much.
Couple that with the half-crossovers measure and it's a bad picture. Even a juicy 5-5 next go will have you not even matching Barb's current position and she'd have had another roll by then and still be on roll.
That "feeling lucky" needs to be working well, and not be daunted by this spelling out of the situation! May the best dice win.
pentejr: Hmmm, but why waste juicy doubles that you could use to make home points with (having brought that 5 down to the 8-=point).
Have a look at the opening move rollouts and see how much splitting with the 1 is in favour. 5-1 as 24/18 is considered to be a blunder. (See Backgammon Links above)
The other aspect of splitting is that it gives two options on attacking the outer table or making a break for it.
pentejr: 13/8 adds a safe spare to 8. As grenv says that gives you the option of using a double, 1-1 being especially tasty. Even without a double you get to use one builder to make a home point and still have another for subsequent rolls. A piece on 8 is a banked asset in a way that slotting your opponent's barpoint isn't.
If that barpoint blot is hit then you've lost the entire roll unless you get one of the 14 retaliatory rolls. You mention thos 4/36 spoilers for that barpoint blot but there's also 5-1 which gets you double-tapped and in danger of a blitz. Then, not only do you waste the opening move but also your second roll too, unless you get a 1 or a double.
The split and bring-down-builders is the best move with many rolls because it advances your cause on both sides of the board. A setback in one of these areas isn't a total loss.
Splitting with a difference of 1 is relatively safer than other splits. A point made on the 2 is deep and commits two builders. Splitting with 2 is vulnerable to a 5-5 and a blitz. Splitting to the 4 and 5-points and getting pointed on is much more uncomfortable than the same on 2.
None of these things is conclusive but they all add their tuppence worth to the mix.
grenv: lol, Yes, impossible's a strong word. I suspect that it's close to the truth though. I'd have to see the code too. (btw. The quick check bit isn't quite right. There would have to be a dice roll at the end of each move in order to know whether the next move is possible or not.)
Czuch: "when white goes to that game both dice are rolled until there is no double"
This is correct.
"why is it okay for me (as white) to roll for black in the initial roll?"
This is not quite correct.
The server rolls the dice not the players, though a player is required to trigger the rolling. So white isn't rolling one for himself and one for black. White is asking the server to roll for both of them. There's no need for turn taking at this stage. In real life both players can roll simultaneously or in any order - it makes no difference as long as there are two dice values coming out of it.
"if auto pass isnt acceptable, then I should be promted (as white) to enter the game and one die should be rolld, [..elaborate sequence..]"
But even allowing for this conceptual sequence.. Because auto-pass isn't unavailable out of principle, the sequence doesn't follow from there being no auto-pass. It's not that auto-pass "unacceptable", it's that it's impossible (without reworking of the design).
Сделано для playBunny (16. Февраля 2006, 17:49:05)
Czuch Chuckers: That's not an auto-pass because nobody skips a turn. The dice rolling code is entered and it doesn't leave until it has a non-double - totally different from auto-pass. And that initial roll still only occurs when the player visits the page. An if they lose that dice roll and the other player has to start, there's been an unnecessary delay there too!
Субъект: Re: Split your back checkers (24/23) and slot your five-point.
grenv: I disagree with the 24/22.
I agree, you're right. I'm doing too many things at once here and that one was hasty. 13/11 is vital cover for the slot. I'd never do 24/22 and slot!
So the typo was in "Split your back checkers (24/23) and slot your five-point. It " would be the 'and' which should have been 'or'. but worse, they've completely missed out the "bring a two down from the mid-point (13/11" (copied from the 3-2 above on that page) which makes that section sensible.
As far as I'm concerned there's only one real reason why auto-pass can't be done and that's because the original design didn't allow for it. The method of dice-rolling is to roll them when the user accesses the page. This precludes auto-pass.
Sites that implement auto-pass roll the dice at the first opportunity. At GoldToken, for instance, if you make a move and then visit the game (non-cube) page immediately afterwards you will be able to see the dice that your opponent will be playing - whether they're online or not. By rolling the dice automatically the system is in a position to skip the move if necessary. At BrainKing, because the dice are rolled only when the player visits the page, there is no opportunity to skip the move. I reckon the changes required would be quite considerable, hence auto-pass is not an option. Any other reasoning for auto-pass being BAD is spurious.
Субъект: Re: Split your back checkers (24/23) and slot your five-point.
Czuch, Grenv: There's no doubt that slotting the five must use the one so you can assume that 24/23 is a typo and should be 24/22.
However, that opening rolls article is the easy-to-use one and is older than the Opening Move Rollouts (link at the foot of that page) which show 13/11 6/5 to be the best move for 2-1 except in the gammon-save situation.
So for a nice read use the opening rolls page and for accuracy you should go by the Rollouts page.
5-1: Generally, it is right to split the back checker and bring one down (24/23, 13/8). When behind in the match and a gammon win is a major plus, you might bring one down and slot your five-point (13/8, 6/5).
Ie. it's more risk than it's worth unless you need the gammon.
The 5-point is hugely important on its own, let alone the advantage of sending a blot back by 20 pips, so, if there's a chance to hit someone who slots their 5-point, GO FOR IT!
Czuch, Hrqls: "i noticed before that i have to make 1 move first before i am given the option to double"
"You" being the first player to move rather than the second player making your first move. Yes, that's correct.
"i am not sure if its a standard rule for backgammon, but it seems to be that way on here"
Yep, it's a standard Backgammon rule. The choice of doubling must be made before the dice are rolled (too much advantage otherwise) and for the game's opening move the dice have already been rolled to determine who starts.
Pythagoras: I'm not sure what you mean by "force it to play". Play it's own moves, play both players, one player when both are human? There's always the Hint button, after which you can make the best move suggested.
DragonKing: You can double from the Game menu or using <Ctrl-D>. I don't do either of those - I just click the doubling cube, lolol. You can also click it to accept a double when it's offered. Decline by using the icon on the toolbar.
If the game you're following shows no play then enter the dice and gnuBg will tell you that it's no play. ;-) If you don't know the dice - such as with the game records here :-, :-/, :-(( - then just make up a dice roll that you know won't work.
DragonKing: That dice chart's an easy one to answer. Click it!
What isn't so obvious is where to click. If it's white to move then click whichever pair of dice you want. If you're creating a new match or starting a new game then it matters which pair of dice you click. If you want to be the one starting tha game then click the lower and leftmost dice that gives you the number you want, ie. below the doubles. If GnuBg is to start then click the same combination but in the top left half, ie. above the doubles. for instance, clicking 2-4 will mean you start while clicking 4-2 will have GnuBg starting.
If by "playing through a game record" you mean manually entering the moves ... the human vs human should do the job just fine when you create a new match. If it's not working I don't know why. What does it do/not do?
Hrqls, Sue, BBW: It's nice to have these links available but, please, don't clutter the top of the board with them.
It's better to do what I did with the Japanese Chess board and appropriate one of the messages as a resources area. I've done just that with one of my old messages .. http://brainking.com/en/Board?bc=26&ngi=342190 .. so you could add that link to the board header. I've only just created it so it only has what we've just been talking about. If anyone has links that they'd like me to add to it then send me a PM or post them here.
Сделано для playBunny (13. Февраля 2006, 18:06:55)
I don't know of anywhere that you can download GnuBg games which contain comments. It's certainly possible to make them, I just don't think anyone's done so.
Create a Backgammon favourites folder and save these. Some of the quzzes will be more relevant to you in due course.
DragonKing: GnuBg is what I use. It's the strongest in the world per $, lol. It also beats almost everything that you'd pay $ for.
Set evaluation and analysis to 2-ply (ie. Supremo) and it'll give you a good fight.
Small tip. Whenever you change any settings in the program, they are only temporary changes unless you use the Settings/Save Settings menu option. It's a pain in the bum when you forget, especially if you've been creating a beautiful 3D board setup and exit without saving. :-/
Сделано для playBunny (12. Февраля 2006, 04:10:01)
DragonKing: Having played them both a bit, the Neeley one has a Hypergammon feel. Fast action (once you're on the board), simple strategy, limited blocking and lots of blot hitting. The Terry's one has more the feel of Crazy Narde, what with the list of "dice" rolls that you get to play with, there being so many pieces on the board and the amount of blocking that can be done.
I'm trying the one from Terry's Egyptian pages. It's hugely different to the P.S. Neeley's one. "Rules vary WIDELY" is true - same board but a completely different game!
Сделано для playBunny (11. Февраля 2006, 05:08:37)
Jim Dandy, Vikings: Not being redundant, Jim, because only you can give your opinion. Just that it was in the wrong place.
I and Walter Montego both agree with you both that non-cube matches should have gammons. Well, N-wins matches shouldn't but N-points matches should. So the votes have gone up.
grenv: Yes, Although it was incongruous with the others, I reckoned you were referring to that player's second move (ie. third overall) just for that instance.
I agree that if your opponent has just made the 4-point then there's the danger of becoming closed in, but rather than put a blot on the 5-point with 24/20, it's better to grab the point entirely 24/20(2) and be secure, especially in Gammon-Save.
grenv: Yes, blot-splatting gets the high priority.
5-5: Always 8/3(2), 6/1(2) if you can hit a blot. Blitz time! :-)) else 13/3 because what else can you do?
4-4: Likewise 8/4*(2) 6/2(2) or 8/4(2) 6/2*(2)
3-3: And 8/5*(2) 6/3(2) or 8/5(2) 6/3*(2)
2-2: If you've done the 4-2 and made the four point then 6-4 is inadvisable. That man is a builder for the 5-point. In fact almost all advancement of builders past empty points in the home table is to be done only if forced.
Hrqls: If your opponent has split his back men and there's a direct shot at your 8-blot then you're better off making only the 5-point. Use the other two 1s to split your own back men. 6/5(2) 24/22
Make your opponent's 5-point and your 8-point (as you would when moving second (unless there's a blot on your own 5-point in which case you hit it 13/5*(2) with glee)).
Сделано для playBunny (10. Февраля 2006, 12:32:55)
Hrqls, KotDB : Well, the dice to go first could be an independant roll. I know that one option in playing (though not on any online site that I know of) is to roll again if you don't want to use the who-goes-first roll. That gives a considerable advantage to the starter.
I imagine that doubles are considered to give too much advantage to the first player on top of being able to go first.
Double 6, 5 and 1 - as you say.
Double 4: Make your opponent's 5-point and your 9-point as you would when moving second (unless there's a blot on your own 5-point in which case you hit it 13/5*(2) with glee).
Double 3: Attack as you suggest, though Gammon-Save would be 13/10(2) 24/21(2)
Double 2: Again attack, 13/11(2) 6/4(2).
grenv: When I returned to Backgammon a couple of years ago I wanted to concentrate on chequer play. The cube was a complication that I didn't need - in fact, couldn't handle. I was playing a very aggressive robot and, almost invariably, accepting a cube meant that I'd lose even much of the 25-30% that were "supposed" to go my way. I just wasn't good enough at that point. Eventually (after about a year, I'd guess) I got to a very good rating and decided to add the cube. Naturally my rating plummetted as I was offering bad cubes and taking huge drops. So then came the second learning curve and ratings climb. In fact I've only recently got to the top of the list against that particular bot. (It's a GnuBg a couple of versions old and with a voracious appetite for blots and a love of blitzes and primes.)
Now I can't speak for Walter, and he's a much better player than I was when I first came back to the game, but I think doing it in stages is a good idea. Polish your chequer play, which means gammons, and then learn the cube. BrainKing's missing that middle stage, ie. multi-game matches with gammons. Perhaps N-wins matches could be just a collection of single games and N-point matches could include gammons.
alanback: I'd love to do it and I'm sure it makes sense. But it also makes for a million cups of coffee. Well, maybe not but a 0-ply is instantaneous while 2-ply can take 3 seconds or so per dice roll and 3-ply takes about 7 seconds (both timings longer for a double and perhaps highly spread positions, etc). The rollout at 0-ply took about an hour and a half so I shudder to think what a 2-ply would require, let alone a 3-ply. There is the option of doing just the first N moves at a higher play and then reducing but I haven't tried that. I don't know that it would make that much difference on an opening move rollout. I'll have to try it ...
The writers of Gnu say that 0-ply is adequate as the errors would tend to balance out for the two players and the dice are random beyond the first two rolls anyway and blah, blah, but only the future will tell us how correct they are.
I thought I had heard that computer analysis had ressurected the 2-point opener. Aye, I'd gained that impression as well and it's been my preferred move for some time now. There's always the possibility of operator error, lol.
Paul Magriel's updating his 70's classic to correct the "mistakes" of that era and also include more about the cube. He's doing it conjunction with some other guy (whose name I can't remember) and they're doing full rollouts on Snowie of all the openings (and maybe all replies?). Hopefully they'll be able to go well beyond 1296 which has far too high an error range for the rolls with subtle differences. I guess it'll be taken as the definitive data for the next wee while..
Hrqls: So in the double match point (DMP) situation the best moves are the drop-and-split and the run-for-home. The difference between them (0.008) is only just bigger than the possible error, so they make be closer than reported or even reversed.
In the gammon-save situation the run-for-home is more clearly judged to be the better move. In both cases GnuBg judges making the 2-point to be a fairly Doubtful move.
Hrqls: Lol. I forgot the column headings. I'll put them in...
Okay, the Move number and Move columns are obvious. The percentage columns are the expected Wins, Wins by gammon (included in the Wins) and Wins by backgammon (including in Wins and W.G) plus the corresponding Lose percentages.
The Equity, is how GnuBg evaluates the potential earning power of the move. 0.000 would mean that there is as much lost as won. 0.018 would indicate a small gain. The -0.202 in the gammon-save situation means that the roll and move is not good in match terms. Perhaps Alan or Pythagoras can explain that one as I'm still uncertain about what equity means in absolute terms.
The important value from the point of view of judging the relative merits of the moves is the last figure, in brackets. This is the difference in equity between that move and the best move. With the levels that I use anything worse than -0.010 is Doubtful and anything from -0.050 is a Bad move. (Thus move 5 in the gammon-save situation is an error because 24/20 24/18 leaves the player open to several double-taps, which is hardly a gammon saving kind of thing to do!)
(убрать) Если Вы хотите узнать больше о некоторых играх, Вы можете проверить секцию Ссылки и увидеть какие-нибудь интересные ссылки там. (pauloaguia) (Показывать все подсказки)