On Saturday, May 6, 2023 at 5:13:46 PM UTC+1, Robert Zimmerman wrote:
On Saturday, May 6, 2023 at 7:54:48 AM UTC-4, [email protected] wrote:
I like the following concept (see end of post) for playing the bots, but it's
convoluted and unlikely to catch on.
For me the most likely thing is that it interests nobody who reads
this thread, but I do have a slight chance of gaining the interest of
some others with Bob and Tim being reasonable prospects.
So why am I posting then? Well, I'm looking for a job and I feel I
should be studying C++ over the weekend to help with this, and
bg is a good procrastination strategy.
So after that tedious preamble, I will now describe my (bad) idea.
The idea is that, when playing bots, you can hedge your PR losses by assigning percentages to each play. For example, with an opening 41,
I can say 13/9 6/5 (40%) 24/23 13/9 (40%) 13/8 (20%). Then the error
for the play is marked at 40% of the error for 13/9 6/5, 40% of the error for 24/23 13/9 and 20% of the error for 13/8. (Of course correct moves have zero error).
The same concept applies to cube actions.
For deciding which move to play, the bot picks randomly among
those with the highest percentages. So, in this case, the bot selects randomly (and uniformly) between 24/23 13/9 and 13/9 6/5 and doesn't
play 13/8.
With this game, a player can mitigate the frustration of obtaining bad luck
with the PR by making wrong 50/50 guesses.
With cube actions, the same concept applies and a player can go 50/50 between doubling and holding with the computer deciding randomly.
BTW, even if this post is about as much use as decaffeinated coffee,
the word "mitigate" is a really excellent word (in my opinion).
Paul
Interesting idea, but wouldn't you be susceptible to making performance errors in your percentage assignments? I suppose you could not let the bot judge you on that.
I'm not a fan of PR at the single match level, although I do look at it to gauge my progress as a player over time. I particularly don't like the idea of tournaments where there is an award, or something like that, related to PR. It sort of takes the
fun out of the game. It's akin to having the players do strength, endurance and agility tests as part of Wimbledon.
You might think of PR performance as a measure of endurance in a match, rather than sheer backgammon brawn. The closer you get to the end of the match, the greater the effect on PR of any given error.
Just let the frigging game happen and let people make their own suppositions about who actually had the better backgammon brain on that day. There's more fun in that, IMO.
If, however, they introduce arm wrestling to BG tournaments, I'm all for it. Bob
"Interesting idea"
As an Aspergers person, I'm often vulnerable to characterizations such as "unable to understand other people".
But didn't I call the situation correctly here -- identifying Bob and Tim as being likely to be interested?
As I've said before in another thread, it's no wonder people call me "Mr. Acurate".
"wouldn't you be susceptible to making performance errors in your percentage assignments?"
In the usual situation where there's only one optimal move, all percentage assignments would contain
performance errors. The point isn't to lessen those errors but to make them less volatile.
if my scheme allowed players to simply hide their errors, it wouldn't be good at all.
I'll give another example. Your initial roll is a 65. I roll a 55 and you dance.
Now, unless you've learned opening theory, the doubling decision is tricky -- I understand that holding
is correct but the double is close. So, if I don't know what to do, I can minimise my max_error by
saying Double (50%) Hold (50%). The bot then picks randomly for me.
Assigning this 50/50 wasn't really an "error" -- it just reflected my psychological state.
"The closer you get to the end of the match, the greater the effect on PR of any given error."
I don't believe this is true. You're probably assuming that PR is directly derivable from losses in MWC,
but I think that the great bg player and writer, Nommag Sengiu, deliberately defined the PR to create
some type of normalisation, or adjustment, so that equivalent errors don't lose more PR simply for
coming later and having more effect on the match.
Paul
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)