| |
BGonline.org Forums
Memorizing backgammon positions
Posted By: Stick In Response To: Memorizing backgammon positions (Timothy Chow)
Date: Monday, 4 July 2016, at 2:28 a.m.
No, a few thousand dollars doesn't count as real money esp. when you're speaking of consuming that much time over a couple/few months. I texted or emailed all the people I mentioned originally (UBK, Chris Yep, Nack, Jake). One key sticking point was the motivation to do it. Nobody really has the desire to do it to prove it can be done. Yep gave a telling answer when I asked him what he thought his chances would be. 'It would be an increasing function based on the size of the bet'. I think all players involved feel strongly this way. Jake said explicitly 'I am sure of one thing, that it would take a lot of money to motivate me'. To me this suggests the task is quite doable but again, who would do it for a couple grand?
A few notes from the group involved.
- Since you're demanding a 100% correct return, we have to somehow assure answers given aren't counted wrong because of a potential misclick or mistype. I suggested hand writing the answers but am no more sure that that's foolproof to avoid this happening. Over the course of at least a few hours I could see how it might be a concern that someone just 'oops' and writes down/types a wrong play, even an illegal play. They shouldn't be punished for an illegal play or even a legal play that obviously wouldn't be made by anyone.
- Jake went on to express how important it is to be interested in the project at hand. I totally agree with this. Like I said earlier, my memory is generally pretty bad, but for the things I like I can remember whatever I like. He also warned me that with age it will get harder and harder to do things like this. I believe that but don't think I'm of that age yet. Hoping to stave off that moment until...shit, what was I saying?
And here are Nack's two emails in full: #1 Thanks for alerting me to an interesting thread. I only read a few of the posts, so please forgive me if I missed any key points.
It's unclear to me what "fair" terms are for a bet, but the memorization of a single best move for each position does not fully reward the knowledge of an expert backgammon player (not that Tim stated otherwise). If an expert bg player gets an early play wrong, it will typically be an error of .01 to .03, whereas if the non-bg memory expert gets a play wrong, it might be an octuple whopper (if legal at all). In short, it is in your interest to bargain for a penalty system based on error size.
I wonder, too, about the arbiter. As far as I know, there is not yet a complete set of (sufficient ply and length) third-roll position XG2 *rollouts where the first- and second-roll plays are best, let alone just (say) "well played." So, as an opening expert, you have (at least) two challenges that decrease your advantage or increase your disadvantage:
(1) What you once "knew" as the best play based on a previous-bot rollout might have been overturned, and/or (2) An XG2 rollout can generate a different answer than XGR++ or whatever (e.g., the second-roll example 52S-52, and there will be a larger percentage of conflicting third-roll positions).
These conundrums hardly matter in an error-size penalty system, but they will make a great difference when even the slightest error is penalized the same as a horrifically conceived move.
I agree with what seems to be the consensus view that cards with both a diagram and the move sequence should be supplied, and the test-taker can give the answer in whatever notation he chooses. Interestingly, though, if you want to tax the non-backgammon "memory expert," you would want to barter (though perhaps unfairly) for using traditional notation -- it's MUCH harder to memorize than Nactation! The non-bg memory expert would have to choose between (a) memorizing a string of numbers with slashes, spaces and sometimes parentheses and an asterisk, or (b) memorizing the exact board position (plus the answer).
In other words, Tim's perspective (or prospect in a bet) has been greatly strengthened due to a shorthand method that didn't exist a dozen years ago, replacing up to 28 characters (though fewer with destination-only trad) with a single character.
That brings up a final query that might help clarify what is being compared: Should the memory expert be required to know how to translate (say) the answer of "P" or "N" or "Y" (or "Bar/24 11/10 6/5*(2)" if memorized that way) to the properly corresponding move on a backgammon board, or even to know the rules of backgammon? Memorizing sequences cold doesn't earn them the right to claim they can now play the third roll moves better if they can't figure out how to put the moves on the board.
#2
Evidently, 10,000 positions has been mentioned as a number. Does that mean the lowest-error-size 10,000 or some other 10,000? Assuming the former, there are 6615 best-played, and the cutoff (combined error size for the first two moves) for the other 3385 is probably around -.005. But many of the answers to those 10,000 are patently obvious, which is why I wonder if the intended 10,000 is a mixture of best-played and less well-played positions selected for some degree of difficulty. I'm not expressing an opinion about what's fair, just pointing out that it matters.
In cramming for the (biannual) world Scrabble championship, I set about to learn all 50,000+(?) British words not in the American dictionary (and reflex-anagram all the long ones with mid/high-frequency letters). This was of course much harder than memorizing the 125,000 words in the American dictionary, so many of which are commonplace on this side or both sides of the pond. (Admittedly, the analogy is imperfect.)
Stick
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.