gnubg analysis statistics

[badgolferman]

I played a 7-point match against GNUBG where I got crushed 7-0. I
didn't feel like I played badly, just that the bot was incredibly
lucky. After performing a whole match analysis I was presented with
the statistics which confirmed my thoughts. But that raised another
question... looking at the overall statistics, what stats would you
consider important? Most of these stats with parantheses I don't
understand so if you want to explain any of them I'd appreciate it.
Sorry for the misalignments.

-----------

gnubg badgolferman

Chequer Play Statistics:

Total moves 87 85

Unforced moves 79 60

Unmarked moves 87 81

Moves marked doubtful 0 4

Moves marked bad 0 0

Moves marked very bad 0 0

Error total EMG (MWC) -0.000 ( -0.000%)
-0.391 ( -4.268%)
Error rate mEMG (MWC) -0.0 ( -0.000%) -6.5
( -0.071%)
Chequerplay rating Supernatural Expert

Rolls marked very lucky 3 1

Rolls marked lucky 3 3

Rolls unmarked 81 74

Rolls marked unlucky 0 6

Rolls marked very unlucky 0 1

Luck total EMG (MWC) +2.373 (+30.491%)
-1.060 ( -9.986%)
Luck rate mEMG (MWC) +27.3 ( +0.350%) -12.5
( -0.117%)
Luck rating None None


Cube Statistics:

Total cube decisions 38 44

Close or actual cube decisions 7 6

Doubles 2 1

Takes 1 1

Passes 0 1

Missed doubles below CP (EMG (MWC)) 0 0

Missed doubles above CP (EMG (MWC)) 0 0

Wrong doubles below DP (EMG (MWC)) 0 0

Wrong doubles above TG (EMG (MWC)) 0 0

Wrong takes (EMG (MWC)) 0 0

Wrong passes (EMG (MWC)) 0 0

Error total EMG (MWC) -0.000 ( -0.000%)
-0.000 ( -0.000%)
Error rate mEMG (MWC) -0.0 ( -0.000%) -0.0
( -0.000%)
Cube decision rating Supernatural Supernatural


Overall Statistics:

Error total EMG (MWC) -0.000 ( -0.000%)
-0.391 ( -4.268%)
Error rate mEMG (MWC) -0.0 ( -0.000%) -5.9
( -0.065%)
Snowie error rate -0.0 ( +0.000%) -2.3
( +0.000%)
Overall rating Supernatural Expert

Actual result +50.00% -50.00%

Luck adjusted result +9.52% -9.52%

Luck based FIBS rating diff. +126.60

Error based abs. FIBS rating 2050.0 1968.9

Chequerplay errors rating loss 0.0 81.1

Cube errors rating loss 0.0 0.0

[Nasti Chestikov]

On Tuesday, 2 March 2021 at 17:13:51 UTC, badgolferman wrote:
> I played a 7-point match against GNUBG where I got crushed 7-0. I
> didn't feel like I played badly, just that the bot was incredibly
> lucky. After performing a whole match analysis I was presented with
> the statistics which confirmed my thoughts. But that raised another
> question... looking at the overall statistics, what stats would you
> consider important? Most of these stats with parantheses I don't
> understand so if you want to explain any of them I'd appreciate it.
> Sorry for the misalignments.
>

First of all, don't take what the bot tells you as gospel. In the case of GNUDung, the software has a severe case of feeling sorry for itself......if it fails to land an 11 in 36 shot it will mark the roll as very unlucky.

See my previous posts on this.

Awful piece of software, it's free for a reason but it will throw you off at a tangent if you believe that you've been lucky when you haven't.

[ah....Clem]

On 3/2/2021 12:13 PM, badgolferman wrote:
> I played a 7-point match against GNUBG where I got crushed 7-0. I
> didn't feel like I played badly, just that the bot was incredibly
> lucky. After performing a whole match analysis I was presented with
> the statistics which confirmed my thoughts. But that raised another
> question... looking at the overall statistics, what stats would you
> consider important? Most of these stats with parantheses I don't
> understand so if you want to explain any of them I'd appreciate it.
> Sorry for the misalignments.

Here's what I'd look at:

> gnubg badgolferman

>
>
> Moves marked bad 0 0
>
> Moves marked very bad 0 0
>

> Chequerplay rating Supernatural Expert

> Cube decision rating Supernatural Supernatural
>

> Overall rating Supernatural Expert

This will give you an idea of your overall level of play, but it won't
really help you improve.

If you're looking to improve your game, find the play that resulted in
the biggest loss of EMG and try to figure out why the bot play is so
much better than yours. Maybe the next one or two big errors. Don't
focus so much on the small errors - you can only internalize so much at
a time.

Another approach is to methodically go through the entire match looking
at each of your plays - most of them will (or should) be correct but it
helps to have that reinforced so you're more confident about your
correct plays.

Don't worry about the luck figure. There's nothing you can do about it
anyway.

[badgolferman]

Thanks for the good advice. Start with the big errors.

Looking at the luck rating helps me soothe my pride when getting
crushed...

[Timothy Chow]

On 3/2/2021 10:02 PM, badgolferman wrote:

> ah....Clem wrote:
>> Don't worry about the luck figure. There's nothing you can do about
>> it anyway.
>
>
> Thanks for the good advice. Start with the big errors.
>
> Looking at the luck rating helps me soothe my pride when getting
> crushed...

I agree with ah_Clem. I never look at any statistic involving luck.
The less one cares about luck, the better.

---
Tim Chow

[Axel Reichert]

"badgolferman" <REMOVETHISb...@gmail.com> writes:

> Start with the big errors.

Definitely. This is the obvious place to start on improving your
game. There is a nice article about this by Kit Woolsey:

https://bkgm.com/articles/GOL/Sep03/burger.htm

I have done an experiment on this. Let's say you play with the tutor on,
warning level set to "bad" (equity loss > 0.08). If you then each time
the tutor warns you about a bad play take an alternative play that seems
reasonable to you and loses less than 0.08 of equity, you will typically
end up with a rating of "advanced" or even "expert". Since GNU
Backgammon's standards for rating are very high, this is quite an
achievement (if you can do this without tutor consistently).

You should definitely also read "All about GNU Backgammon" by Albert
Silver. Make sure you work through it completely and understand it. If
you don't get some of the theory, search on Backgammon Galore or ask
here for pointers.

> Looking at the luck rating

Don't. There is nothing to gain here, except dice paranoia. But you
should read an article by Douglas Zare:

https://bkgm.com/articles/Zare/AMeasureOfLuck.html

> helps me soothe my pride when getting crushed...

One thing you can be proud of is when you have reached a level of play
so high that your opponents NEED more luck than you to beat you: An
average game lasts, say, 25 moves for each player (more statistics again
on Backgammon Galore). If on each of your moves you throw away 0.02 of
equity (while your opponent plays perfectly), in total you will have
squandered 0.5 of equity. If net luck (you will understand after reading
Zare's article) cancels out in that game, your opponent is at the point
where he doubles, you have a borderline take and play on.

So without having more luck than you your opponents will not reach the
cash point and be able force a win. Please note that I am heavily
simplifying things here, but the general idea is sound. And I think it
is nice to able to say:

I cannot be beaten without you having more luck than me.

Best regards

Axel

[ah....Clem]

Well, I always look - XG displays it so prominently that it's impossible
to not look.

But I don't think about it very much, other than to note that anytime I
beat a bot my luck has been quite good.

[Philippe Michel]

On 2021-03-02, ah....Clem <ah_...@ymail.com> wrote:

> Another approach is to methodically go through the entire match looking
> at each of your plays - most of them will (or should) be correct but it
> helps to have that reinforced so you're more confident about your
> correct plays.

Another interest of this kind of thorough review (if done soon enough
that you remember what you were thinking while playing) is to find
plays that were right but where you either:
- thought they were obvious when there were one or more close alternatives
- agonized on when the right play was in fact clearly best

[Timothy Chow]

On 3/3/2021 3:01 PM, ah....Clem wrote:
> Well, I always look - XG displays it so prominently that it's impossible
> to not look.

It's not impossible. I use XG all the time, and I literally
couldn't even tell you where XG displays the luck.

I will admit that I sometimes notice that certain rolls, or
games, are marked lucky or unlucky.

---
Tim Chow

[Stick Rice]

I have no idea where it is either. Couldn't care less.

Stick

[Nasti Chestikov]

On Saturday, 6 March 2021 at 04:09:04 UTC, Stick Rice wrote:
> I have no idea where it is either. Couldn't care less.
>
> Stick

I'm not a qualified meteorologist but I can tell you when it's raining.

Likewise, if I have the bot with 3 counters on the bar and only the 5 point open and it rolls 5-5 and comes back in and then hits my blot on the 15 point, I'm calling bullshit on the dice.

I don't need GNUDung or XG to tell me that its roll was "very lucky".

[Axel Reichert]

Nasti Chestikov <nasti.c...@gmail.com> writes:

> if I have the bot with 3 counters on the bar and only the 5 point open
> and it rolls 5-5 and comes back in and then hits my blot on the 15
> point, I'm calling bullshit on the dice.

Some observations/correlations from my experience:

1. The more often someone calls out a favourite dice roll ("Now 55!",
"Please a 62!"), the poorer the level of play.

2. The lower the error rate of a player, the less dice paranoia.