Is gnubg luck related to skill?
bobsw...@nospam.comcast.net
own luck ratings. I just played twenty five games with the bot and it
ended the game with a higher luck rating than me 16 times, I was
higher on 4 and we had the same on 5. I know this is a tiny sample
but I seem to consistantly get out-lucked by the bot and only did the
25 game test because of what seemed like a higher than average luck
factor for gnubg over the last 4-5 months that I've been playing
gnubg.
One other thing. In the last game of the 25 game test we ended up
with the same 'None' uck rating, even though the bot had been at 'Go
to Las Vegas Immediately' just two rolls previous, and from the
numbers below it appears gnubg had a clear luck advantage. Can
someone please tell me how these luck ratings are calculated?
gnubg user
Luck rate (total) +1.748 (+17.564%) -0.754(-16.877%)
Luck rate (pr. move) .019 (+0.189%) -0.008 (-0.170%)
Luck rating None None
Thanks,
Bob Sweeney
Gregg Cattanach
works:
For each dice roll, the bot calculates the equity of the resulting position
for each dice roll if it is moved 'correctly'. These 36 equities (actually
21 unique positions but you count the non-doubles twice) are averaged. Then
the equity for the roll you actually got is compared to the average. If the
roll is above the average you will get a positive luck factor for that roll,
if it is below average you will get a negative luck factor. The actual way
you make your move doesn't matter at all, it does all this based on what it
thinks is the best play for the 21 rolls. This is added up for every throw
for the whole game/match.
Also, GnuBG keeps two separate numbers, the luck factor for your rolls and
for your opponent's rolls. Snowie just merges these two numbers into a
single number. As far as the 'word' ratings for luck in GnuBG, I have no
idea.
In backgammon by making good moves you often give yourself many more rolls
that will improve your positions. However, this skill doesn't come into
play either with this luck factor calculation, because if most of your rolls
improve your position, the average roll will have a high equity to start
with, and only the rolls above that average will come out as 'lucky'. Most
of the time, there are more than 1/2 of the rolls above or below the
average, not exactly 1/2. Some large jokers or anti-jokers skew the number
of actual rolls that come out above or below the average. If you have
Snowie, you can see this real easily in the 'Dice' panel.
Hope this helps.
Gregg C.
<bobsw...@nospam.comcast.net> wrote in message
news:cqgaqug9r7uliach2...@4ax.com...
Douglas Zare
bobsw...@nospam.comcast.net wrote:
> I ask because the bot is consistantly luckier than I am based on its
> own luck ratings. I just played twenty five games with the bot and it
> ended the game with a higher luck rating than me 16 times, I was
> higher on 4 and we had the same on 5. I know this is a tiny sample
> but I seem to consistantly get out-lucked by the bot and only did the
> 25 game test because of what seemed like a higher than average luck
> factor for gnubg over the last 4-5 months that I've been playing
> gnubg.
Suppose to start off each match, you make a blunder worth 20% mwc, but
then you and your opponent play perfectly. Then you will win 30% of the
time, and every time you win, your net luck will be +70%. Every time you
lose, your net luck will be -30%. If you win 3 games in 10, then your
total luck would be 210%-210%=0%.
The total luck should average to 0 in the long run. However, the stronger
player will be luckier most of the time. In a match between players who do
not play at a novice level, the luckier player wins, so it is not useful
to check who is luckier if you already know who won the match. What is
useful is to see how much luck was used.
Douglas Zare
Jørn Thyssen
> I ask because the bot is consistantly luckier than I am based on its
> own luck ratings. I just played twenty five games with the bot and it
> ended the game with a higher luck rating than me 16 times, I was
> higher on 4 and we had the same on 5. I know this is a tiny sample
> but I seem to consistantly get out-lucked by the bot and only did the
> 25 game test because of what seemed like a higher than average luck
> factor for gnubg over the last 4-5 months that I've been playing
> gnubg.
>
> One other thing. In the last game of the 25 game test we ended up
> with the same 'None' uck rating, even though the bot had been at 'Go
> to Las Vegas Immediately' just two rolls previous, and from the
> numbers below it appears gnubg had a clear luck advantage. Can
> someone please tell me how these luck ratings are calculated?
Gregg has explained how the luck rate is calculated in another posting,
so I'll just explain the luck rating. I mostly introduced it for fun :-)
r = luck rate pr. move in in terms of equity
That is, 0.019 and -0.008 per move for gnubg and you, respectively.
-oo < r < -0.10: Haaa-haaa (imagine that it's shouted by
Nelson from the Simpsons)
-0.10 < r < -0.06: Go to bed
-0.06 < r < -0.02: Better luck next time
-0.02 < r < +0.02: None
+0.02 < r < +0.06: Good dice, man!
+0.06 < r < +0.10: Go to Las Vegas immediately
+0.10 < r < +oo : Cheater :-)
I must admit that these numbers are close to arbitrary. If someone can
suggest better numbers please let me know.
Jørn
Murat Kalinyaprak
Douglas Zare wrote 3DA5C3EB...@math.columbia.edu
> The total luck should average to 0 in the long run. However, the
> stronger player will be luckier most of the time.
Douglas buddy,
Are you smoking something there or is it just a language issue...?
I thought I just read several articles explaining in detail that how the
players actually move their pieces doesn't effect the "luck factor"...
You must know something different...??
> In a match between players who do not play at a novice level, the
> luckier player wins,
And which one of the novices wins in a match between novices...?
And, of course, why...?? :))
> so it is not useful to check who is luckier if you already know who
> won the match.
Does this mean that it is useful to check who is luckier if you don't
know who won the match...? For what purposes...?? Further, what
would be the best ways of accomplishing this...? With a crystal ball
or Tarot cards...??
> What is useful is to see how much luck was used.
Now, this sounds like it could be a really interesting subject if you
would develop on it a little more... :)) I will even pay you for it...
MK
-----= Posted via Newsfeed.Com, Uncensored Usenet News =-----
http://www.newsfeed.com - The #1 Newsgroup Service in the World!
-----== 100,000 Groups! - 19 Servers! - Unlimited Download! =-----
Murat Kalinyaprak
Jørn wrote 3DA5D49C...@NOSPAMchem.sdu.dk
> -oo < r < -0.10: Haaa-haaa (imagine that it's shouted by
> Nelson from the Simpsons)
> -0.10 < r < -0.06: Go to bed
> -0.06 < r < -0.02: Better luck next time
> -0.02 < r < +0.02: None
> +0.02 < r < +0.06: Good dice, man!
> +0.06 < r < +0.10: Go to Las Vegas immediately
> +0.10 < r < +oo : Cheater :-)
> I must admit that these numbers are close to arbitrary. If
> someone can suggest better numbers please let me know.
I don't know if they would be closer to arbitrary than your
numbers but I would suggest 0.033 and 0.066 instead of
your 0.02 and 0.06...
I don't know who Nelson is but I heard about "Bart Simpson";
can you make the "Haaa-haaa" sound like him instead...??
If you can't be a man to know the difference between "close
to arbitrary" and "arbitrary", change that "Good dice, man!"
to "Good dice, fag!", errr, I mean "Good dice, dude!"...
And take out that silly smiley after "Cheater"... You should
know better... :)))))))