Unusual game... I had just one choice throughout!
Edward D. Collins
although certainly not unheard of.
For the entire game, I only had to make one decision! (After an opening
roll of 2 6 by my opponent, how to play 5 1.)
I split my 24 point with the 1 and was promptly hit. I rolled a 2 6 and
failed to enter (an 8 to 1 shot), failed to enter AGAIN on the next few
moves and watched my opponent close his board and successfully bear of
seven men before I left the bar and promptly resigned a certain gammon.
After the game I realized that the above "record" can never be broken
since you're always going to have a least one decision to make, your
opening roll.
Here's the game in Jellyfish 3.0 format:
3 Point Match
Game 1
rocketman : 0 EdCollins : 0
1) 26: 24/22 22/16 51: 24/23 13/8
2) 64: 8/2* 6/2 26:
3) 64: 24/20 16/10 66:
4) 35: 10/5 8/5 25:
5) 34: 8/4* 4/1* 25:
6) 41: 6/5 5/1 61:
7) 56: 20/15 15/9 16:
8) 55: 13/8 13/8 8/3 8/3 41: 25/21
9) 11: 9/8 8/7 5/4* 5/4 13:
10) 46: 13/9 13/7 52: 25/20
11) 43: 9/5* 5/2 22:
12) 31: 13/10 10/9 33:
13) 54: 9/4 7/3 24:
14) 26: 7/1 6/4 13:
15) 65: 6/0 6/1 25: 25/20
16) 14: 4/3 4/0 41:
17) 14: 4/3 4/0 23:
18) 55: 3/0 3/0 3/0 3/0 61:
Wins 2 points
Best regards,
|'-''-'| ___ __ _
|______| (_ _/ _ _ _/ / ) / ) //' _
|====| /__(/((/(// (/ /(_/. (__()((//)_)
| |
| | ecol...@inficad.com
|____|
|====| Edward D. Collins proudly presents...
/======\ his very own homepage!
(________) http://www.inficad.com/~ecollins
Bob Ebbeler
it seem to follow this theme: an opening roll where you split your back
checkers, either running with one or having one each on the 24 and 23
points, opponent then rolls 5-5, points twice, you fan and never move again.
The fact that its not as unusual as you think doesn't make it any less
painful. Next time, ROLL BETTER !!!
NIHILIST
Edward D. Collins <ecol...@inficad.com> wrote in message
news:j4fZ3.1$mX5....@typhoon-la.pbi.net...
> |______| (_ _/ _ _ _/ / ) / ) file://' _
Gregg Cattanach
hehe, i HATE it when that happens! :-))
==========
Zox
Bob Stringer
>
> I recently played a game on FIBS which I thought was quite unusual,
> although certainly not unheard of.
>
> For the entire game, I only had to make one decision! (After an opening
> roll of 2 6 by my opponent, how to play 5 1.)
> [snipped]
I had a similar experience. In a 7 point match a while back, I was
gammoned, gammoned again, and then backgammoned. I had some choices to
make, but they were on the scale of bad versus grotesquely bad, and so
it wasn't difficult to decide what to do. And through most of the
match, though, I simply got pounded on. I also had a more recent 5
point match where the same thing happened -- I was gammoned and then
backgammoned in very fast order. At the end of that one, when I said
"thanks for the match," my opponent said "sorry".
The funny thing was that in each case, when I ran the match through
Snowie, the bot said that I had played like a "world class player." I
thought "That's great! Bring on Kit Woolsey!"
Bob Stringer
To reply please replace "REMOVE" with "bob" in my address
Daniel Hollis
Bob Stringer <REMO...@pacbell.net> wrote:
>
>The funny thing was that in each case, when I ran the match through
>Snowie, the bot said that I had played like a "world class player." I
>thought "That's great! Bring on Kit Woolsey!"
The classification of your play is cute and it feels good to have the bot
"think highly" of your playing skill. However, I just want to point out
that it's mostly useless.
The base factor, I believe, is a sum of the total errors in equity you
make throughout the game. Most obviously - as you noticed - if you only
get to split your back men on the first move and then get closed out, your
error is extremely small. My cat can be a world champion player if I make
the opening move for her.
The second point that I don't like is that often the same error can be made
more than once on consecutive turns. For instance, suppose I hit a shot in a
back game, but I broke my 6 point in the process. Instead of slotting the 6
point, I bring my back men around. Suppose my play is worse than
slotting by x points per game on average, each time.
If I make the bad play and get away with it, I haven't lost any
equity. By making the bad play several times, I'm risking
the same bit of equity each time. If I make the bad play 5 times,
then I got away with it 4 times (for no loss), and I am still risking
my opponent coming in (for average x loss). I will not on average
lose 5x points more than if I made the correct play, yet this is what
the error rate uses for measuring.
The net result is that one conceptual error can be counted several
times. Even though the total error is compared to the number of
unforced moves (I think), it's still the case that if you have to
make more moves, your error rate will still go up.
Dan
Bob Stringer
>
> Bob Stringer <REMO...@pacbell.net> wrote:
> >
> >The funny thing was that in each case, when I ran the match through
> >Snowie, the bot said that I had played like a "world class player." I
> >thought "That's great! Bring on Kit Woolsey!"
>
> I think the way snowie classifies players is "funny" all the time!
> The classification of your play is cute and it feels good to have
> the bot "think highly" of your playing skill. However, I just want
> to point out that it's mostly useless.
>
I wouldn't go so far as to say "mostly useless," but I agree with much
of what you say.
In my experience, there are two principal areas where the "level" rating
isn't meaningful -- both similar to what you've described. First, where
the position has certain characteristics that persist over several
moves. From the standpoint of the Snowie evaluations, I find that this
isn't an "issue" so much with respect to checker play, as it is with
respect to my failure to double when Snowie thinks that I should. As
you mention, in such a case I'm really making only a "single" conceptual
error, but the computer naturally sees several errors, and reports them
all. [Of course, that's still useful in going over the analysis,
because you still get to see Snowie's analysis of each position. And
the evaluations will be somewhat different -- for example, in the first
position, if I had doubled, Snowie would have taken, whereas in the
second position, it would dropped. The errors, therefore, aren't
necessarily on the same level. The important thing, in any case, isn't
whether at the end of the analysis Snowie says I was an "expert" or a
"putz," but that my errors are identified, so I can look them over.]
The second general case is the one that I mentioned in my first post --
where I lose because I don't have many choices, or at least the choices
are obvious and all bad. In such cases, my error rate of course will be
low, and Snowie will "think highly" of my play. But in such a case, I
already know that my "level" of play is unimportant. I analyze such
games simply to see what errors there were -- both in my game and in my
opponent's.
Over the long run, the error rates do seem to reflect reality, at least
to the extent that they reflect my *relative* progress. With the
exception of a couple of months in the early '80's, I started playing
backgammon just a little over a year ago. I purchased Snowie right away
and used it to analyze my matches. In the beginning, my error rates
were much higher, and my level of play, more often than not, was
"intermediate." The error rates are lower now, and the more common
evaluation is "advanced," suggesting, perhaps, that I have managed to
rise out of the bog to a limited extent.
Chuck Bower
Daniel Hollis <hol...@math.umn.edu> wrote:
>In article <3835F54D...@pacbell.net>,
>Bob Stringer <REMO...@pacbell.net> wrote:
>>
>>The funny thing was that in each case, when I ran the match through
>>Snowie, the bot said that I had played like a "world class player." I
>>thought "That's great! Bring on Kit Woolsey!"
>
> I think the way snowie classifies players is "funny" all the time!
>The classification of your play is cute and it feels good to have the bot
>"think highly" of your playing skill. However, I just want to point out
>that it's mostly useless.
>
> The base factor, I believe, is a sum of the total errors in equity you
>make throughout the game. Most obviously - as you noticed - if you only
>get to split your back men on the first move and then get closed out, your
>error is extremely small. My cat can be a world champion player if I make
>the opening move for her.
>
> The second point that I don't like is that often the same error can be made
>more than once on consecutive turns. For instance, suppose I hit a shot in a
>back game, but I broke my 6 point in the process. Instead of slotting the 6
>point, I bring my back men around. Suppose my play is worse than
>slotting by x points per game on average, each time.
>
> If I make the bad play and get away with it, I haven't lost any
>equity. By making the bad play several times, I'm risking
>the same bit of equity each time. If I make the bad play 5 times,
>then I got away with it 4 times (for no loss), and I am still risking
>my opponent coming in (for average x loss). I will not on average
>lose 5x points more than if I made the correct play, yet this is what
>the error rate uses for measuring.
>
> The net result is that one conceptual error can be counted several
>times. Even though the total error is compared to the number of
>unforced moves (I think), it's still the case that if you have to
>make more moves, your error rate will still go up.
True that if you make only a few plays in a match you are more
likely to get a high evaluation from Snowie ("Extraterrestrial", "World
Class", etc.) But what would you like it to do? If you log your matches
then in the long run these high evaluations won't mean much because plays
are weighted by number of chances (unforced moves), not number of matches.
As far as being repeatedly chastised for the same error, that seems
to make sense to me. Each turn is a new challenge. Ignorance is no
excuse (to quote the law). The fact that the dice didn't punish you is
simply luck, not skill. Snowie is evaluating your skill only.
Imagine you're playing golf, you have two paths, one of which is
risky (e.g. driving over a lake) and the other is safer, but causes you
to play farther and having less chance for a birdie. You aim over the
lake, hit your shot and... plop. That's two strokes (one for the swing
and one for the penalty of going in the water). You re-tee aim over the
water, and... same result. That's two more strokes on the scorecard.
You made the same 'mistake' (if that's what you want to call it) and got
punished TWICE. Of course at backgammon you don't get such stern immediate
feedback. But you did make TWO mistakes. And that is what Snowie is
evaluating you on--each chance (roll).
Chuck
bo...@bigbang.astro.indiana.edu
c_ray on FIBS
Bob Ebbeler
PLAYED LIKE A JAGOFF ?
NIHILIST
Daniel Hollis <hol...@math.umn.edu> wrote in message
news:816g2k$678$1...@news1.tc.umn.edu...
> Dan
>
>
Bob Stringer
>
> You paid a couple hundred bucks for Snowie, do you expect it to say YOU
> PLAYED LIKE A JAGOFF ?
>
It's almost a running joke around my house. My wife doesn't hear any
sounds from me during a match, and afterwards she'll ask "so, does
Snowie think we're a lucky idiot today?" When she hears wailing and
keening, she knows that I'm playing one classy game, but have to deal
with those #$%^! skewed dice.
One of the top chess-playing programs, Fritz, has an audible commentary
feature, which you can able or disable as you please. Since the program
is made in Germany and sold internationally, you have a choice of
languages. The English guy sounds prissy to me, and his comments are
insipid for the most part. But the German guy is something else.
That's the voice I always use, even though I don't understand a word of
the language. [Well, all right. I'm familiar with Sweinhundt (sp? --
it's not in my spell check) and a couple other exceptions.] Sometimes
he sniggers so hard at my moves he actually snorts. Great stuff.
When a bot [instead of some clown on FIBS] does it, it's funny. Imagine
Snowie yelling "idiot!" "schmuck!" as you play.
Vince Mounts
Chuck said, the World Class games get mixed in with the games where you had
more choices to make an average.
What I think is more interesting to explore is the luck factor. I wonder
what it is really measuring because it doesn't seem to be what humans would
call luck. Here snowie is measuring how good your dice are. How does it do
this? Well for each time someone rolls it looks at all the possible dice and
determines the best move for each assigning it an equity. It then averages
the equities (remembering to count rolls like 6-1 1-6 twice). If your actual
roll is better than this average then you get a plus in your luck column and
the opponent a minus. This has some interesting consequences I think.
First realize that given perfectly even ditribution of dice (same number of
doubles, number of shots hit per percent chance ,exactly the same number of
times rolling each combo etc.) that unless you play exactly what snowie
thinks is "right" you will always be luckier than the bot. Another way to
say this is that bot would have to be luckier than you would have to be to
show the same luck rate. Why? Well, because every time you make an "error"
you are giving away equity to the bot. This makes it seem more likely that
whatever miracle the bot rolled was more likely. Say in one game you need a
1-6 to enter and hit a blot on the bar pt to save the game. You do so. The
next game the same thing happens for the bot. The bot is considered less
lucky even though it is the same exact thing that won the game. This isn't
so bad. It makes sense that the less skilled player needs more luck to win
the match. But it is interesting that the "mistakes" you made, that makes
you look luckier, had nothing to do with the lucky shot that won or lost the
game. So the main point is that due to equity you give the bot you will
always appear luckier in the long run. This even applies to such
inconsequential moves as how to build your board in a holding game. All
those tiny disagreements make the bot look less lucky. I experienced this
recently. I was playing a series of 11 pointers. One match I wiped snowie
off the board with great dice in 4 games. My luck rate for the matchwas
big... almost 3.0 . The very next match I got beat in 4 games. The match had
a very similar number of rolls, and the same number of 2 cube gammons and
very similiar flow with the loser dancing repeatedly etc. The bots luck
factor was 0.025. This seemed absurd to me. Reviewing both matches I still
didn't see how my streak was that much better than the bots.(note: this is
what prompted me to realize the bot would always be less lucky)
Also... When people discuss luck on this group you often hear "There are
three things that can happen. 1)tHe better player gets the better luck and
wins. 2)The lesser player gets the better dice and wins. or 3) The lesser
player gets the better dice and the better player wins through his skill."
This also implies that with exactly equal opponents the best dice always
win. I may be wrong but with the way snowie measure "luck" there is another
possible outcome. "The lesser player gets the worse luck and wins
anyway."(or with equal players the less lucky player wins). Maybe there is
some mathematical proof that this can't happen but my intuition says it can
with the way luck is being measured. Let me construct an example.
Lets assume for a minute both players make the all the moves in agreement
with snowie.
One player starts the game with a series of jokers and the other player
rolls poorly. Maybe the first guy gets 3-1 followed by 6-6 ,(double/take)
and then 4-4 or something. While the other player can't even seem to get a
good anchor due to uncooperative dice. The unlucky guy eventually gets the
anchor but this doesnt affect his poor "luck" all that much. Now during the
holding part of the game the leader rolls very slightly better than average
to bring it home while the other guy rolls less than average while building
his board. But a fly shot chance at the last guy on the midpoint finally
comes about and the unlucky fellow gets lucky and hits it. At this point
surely the first player still has the luck in his favor overall. Now assume
the opposite from before. The new leader rolls slightly worse than average
to bring the position home. The new trailer isn't hitting the fly shots but
this expected and not unlucky or below average. The expected outcome is that
the new leader will win which he does but always rolling a little worse than
he could have. At the end of the game the least lucky player has won. With
the above assumption that no checker errors were made the least lucky person
won the game. This can be made more pronounced if the the least lucky person
can cube the other out after the hit.
Is this really what we mean when we talk about luck? What is snowie really
measuring? Is there a better way to measure luck? perhaps measuring only key
plays against their likelyhood? I don't know.... what do you all think?
Carl Garringer
Vince Mounts <vmo...@mindspring.com> wrote in message
news:81fgvp$p2$1...@nntp9.atl.mindspring.net...
I disagree with several of your statements above. First of all the luck is
determined just by the rolls and not how they are played. In your 1-6
example above the luck should be same the same for both, but this would be
determined by the average expected equity for the roll. However, when a poor
player does give away equity, that will probably result in more good rolls
for Snowie (or the better player). This in turn should raise the average
equity which requires a better roll to gain any luck. Any gained luck is
likely to be a smaller amount and any luck loss is likely to be larger
amount.
Daniel Hollis
remembers the average equity you will come out with after your roll, and
compares it to it's estimate for the best play on that roll.
I was talking about the Snowie error rate measurement. I happened to be
gone for a bit for Thanksgiving and didn't get a chance to respond to
Bob or Chuck, unfortunately.
I have reconsidered on the point that, over the long run, these error
rates will show one's improvement. I wouldn't Snowie's error rate
figure from a small number of games; any given game might have the kind
of compounded error that I was talking about earlier. However, it's not
going to happen all that much. It's only that when it does happen, the
effect is too large for me. Since it's not going to happen that often,
I would trust Snowie's error rate for a bunch of games (my idea of
large and trust here are general and subjective, but I don't think
anyone is using these numbers for predicting anything anyway).
Dan
toishan
Edward D. Collins <ecol...@inficad.com> wrote in message
news:j4fZ3.1$mX5....@typhoon-la.pbi.net...
> I recently played a game on FIBS which I thought was quite unusual,
> although certainly not unheard of.
>
> For the entire game, I only had to make one decision! (After an opening
> roll of 2 6 by my opponent, how to play 5 1.)
>
Gregg Cattanach
Just FYI,
The luck factor calculated by Snowie works this way. It calculates the
resulting equity of the 36 rolls, assuming each is moved 'correctly'. It
then subtracts the equity resulting from the roll that was rolled from this
average. A negative results is an 'unlucky' roll, a positive number is a
'lucky' roll. This is accumulated for each game, and the result is shown
and reported. All good luck for one person is counted as bad luck for the
other, so it is a zero sum result.
--
Gregg Cattanach
Zox at GamesGrid, VOG, Zone
http://gateway.to/backgammon
ICQ #2266410
gcattana...@prodigy.net
"Daniel Hollis" <hol...@math.umn.edu> wrote in message
news:8248cd$ctl$1...@news1.tc.umn.edu...
Hobbyish
however:
1. They are cube-independent. A lucky roll with the cube on 8 is not
given greater weight than an equally lucky roll with the cube on 1.
2. They are cubeless. To me, the "luckiest" roll in most cases is the
roll my opponent gets that lets me offer an efficient double. If my
opponent's equity on a given roll goes from, say, -.55 to -.75, he
will be assigned negative luck, when in fact it really did me no good
at all - I went from an optimally efficient double to a pass.
3. They are gammon-dependent. If you play shorter matches you will
have a greater proportion of games where gammons do not count for one
side or the other. In the limiting case, suppose you played only
one-point matches. All the luck factors would be smaller since
gammons don't count.
4. This is perhaps more a case of asking how many angels can dance on
the head of a pin - but - consider the following:
Trailing 160-148 in a non-contact position you roll 6-6. Your
pre-move equity was (in the position I chose to set up) about -.20,
after the roll is is about +.32. The roll was lucky by .52.
You have a checker on the 5 and one on the 2, your opponent has one on
the ace. Your pre-roll equity is .056. Your luck factor on the next
roll will be either +.944 or -1.056.
Somehow, I personally feel that if I roll 6-6 in a race, that ought to
be the "luckiest" roll of the race, not the final roll on which I do
or don't get off to win. If I roll 6-6 and reach the 5-2 position,
well, I would never have even gotten to that if I hadn't rolled the
6-6.
Don't take any of this to mean that what Gregg said is inaccurate, he
was quite correct in his comments.
On Wed, 1 Dec 1999 20:58:59 -0500, "Gregg Cattanach"
<gcattana...@prodigy.net> wrote:
>
>Just FYI,
>
>The luck factor calculated by Snowie works this way. It calculates the
>resulting equity of the 36 rolls, assuming each is moved 'correctly'. It
>then subtracts the equity resulting from the roll that was rolled from this
>average. A negative results is an 'unlucky' roll, a positive number is a
>'lucky' roll. This is accumulated for each game, and the result is shown
>and reported. All good luck for one person is counted as bad luck for the
>other, so it is a zero sum result.
>
>--
>
>Gregg Cattanach
>Zox at GamesGrid, VOG, Zone
>http://gateway.to/backgammon
>ICQ #2266410
>gcattana...@prodigy.net
>
>
>"Daniel Hollis" <hol...@math.umn.edu> wrote in message
Vince Mounts
some of hobbyish's post on the effect of the cube.
My gut feeling about it is that it is just another way of measuring equity
in a position rather than "luck",, though it does correspond to luck in a
loose way also. I had thought of some examples but I'm not really into
dredging them out of memory again and maybe my "gut" is just shy of some
analysis.
V
Walt Swan
I still believe it slows the analysis down greatly.
"Vince Mounts" <vmo...@mindspring.com> wrote in message
news:8270m5$u6q$1...@nntp3.atl.mindspring.net...