[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Stats labels

From: Robert-Jan Veldhuizen
Subject: Re: [Bug-gnubg] Stats labels
Date: Fri, 29 Feb 2008 12:28:42 +0100

Hi all, here's some of my suggestions:

I agree some changes in the statistics would be very helpful.

Personallly I'd say the most important figures are: total EMG lost (error), EMG/mv lost (error) and the corresponding "Snowie error rate";
for matches, total %MWC lost.

Then it would be nice if GNUBG gave the sum of the total EMG/MWC lost for both players, for comparison with the luck-adjusted result.

Further, for money: total luck abs. points; for matches: total luck MWC.

In my opinion, "luck per move" hardly makes any sense, since it approaches to zero as the number of moves goes to infinity. It's therefore not really comparable to error per move anyhow. Especially the labels that GNUBG now gives, based on a "luck per move" figure, give results that just don't give people a good understanding of what luck really means in backgammon, IMO. Labels based on total luck would make a lot more sense, especially on longer matches. The only thing that counts is your total luck minus opponent's total luck; a higher number of moves (read: a loner match/session) should not lead to being called less (un)lucky or something...

It would also be nice if something could be done about GNUBG's way to harsh cube decision labels. Since they only get divided by the number of "close" or actual decisions, the error per move here tends to be very high when compared to the checker play error per move. Unfortunately GNUBG doesn't allow for this when giving its label (I believe the labels are still based on GNUBG's old system where all cube decisions were being counted). A new range for the labels would be more realistic.


On Fri, Feb 29, 2008 at 11:43 AM, Massimiliano Maini <address@hidden> wrote:

Hi all,

in an old post to the mailing list (thread "Luck rates ?") we discussed about changing
the somehow confusing labels shown by gnubg in the stats (Error and Luck stats).

You can fetch the entire thread, but I just recap the issue here below (cut&paste of
previous posts):

Here are examples of money play and match play from the current implementation:

                                      yanwolfy             ianshaw

Error rate (total)                    -7.290 ( -8.329)     -3.441 (-3.549)    
Error rate (per decision)              -34.5 ( -0.039)      -15.9 (-0.016)    
Equiv. Snowie error rate               -19.4                 -9.2

Luck rate (total)                     +2.929 ( +2.118)     -2.330 (-4.734)    
Luck rate (per move)                   +15.7 ( +0.011)      -12.3 (-0.025)  

                                      Bennet               Shaw

Error rate (total)                    -2.808 (-23.283%)    -3.397 (-22.420%)  
Error rate (per decision)              -19.8 ( -0.164%)     -22.6 (-0.149%)  
Equiv. Snowie error rate                -9.1                -11.0

Luck rate (total)                     -5.198 (-49.754%)    -1.288 (-8.433%)  
Luck rate (per move)                   -33.3 ( -0.319%)      -8.4 (-0.055%)  

The current system is confusing because:
1) the bracketed part of the title does not refer to the bracketed part of the data
2) it is not clear what the first and second (bracketed) values mean
3) "Error rate (total)" and "Luck rate (total)" are not rates at all (as
Christian said). They are simply "Total Errors" and "Total Luck".
4) the units are not specified, which would be good practice. Totals are
in points for money play or MWC for matches; rates should be in
millipoints per decision for errors or millipoints per move for luck.

At th end of the thread, the last proposals were:

                                      yanwolfy             ianshaw

Overall Statistics:
Total Errors (points)                  -7.290               -3.441

Total Errors as EMG (points)           -8.329               -3.549

Error rate (mp/decision)               -34.5                -15.9    
Error rate as EMG (points/decision)    -0.039               -0.016

Equiv. Snowie error rate               -19.4                -9.2

Luck Statistics:
Total Luck (points)                    +2.929               -2.330
Total Luck as EMG (points)             +2.118               -4.734
Luck rate (mp/move)                    +15.7                -12.3
Luck rate as EMG (points/move)         +0.011               -0.025

The display occupies four extra lines of text, but they would fit within
the existing area of the gui. It has the advantage of clearly
identifying both measurements.

The units labels are a bit ugly, so maybe we can dispense with them
after all. (Can we do a mouse-over tooltip on this section?)

It appears that the EMG luck rate and EMG error rate have not been
multiplied by 1000 to convert from points to millipoints per move. There
ought to be consistency here.


                                      yanwolfy             ianshaw
Overall Statistics:
Total Errors in points (EMG)           -7.290 (-8.329)      -3.441 (-3.549)
Error rate in mp/decision (EMG)        -34.5  (-39.0)       -15.9  (-16.0)  
Equiv. Snowie error rate in mp/move    -19.4                -9.2

Luck Statistics:
Total Luck in points (EMG)             +2.929 (+2.118)      -2.330 (-4.734)
Luck rate in mp/move (EMG)             +15.7  (+11.0)       -12.3  (-25.0)

The only downside of this setup is, to me, that the most meaningful error
values (the total error EMG and error rate EMG) are in brackets.

My last comment was about the fact that the meaningful quantities to look at
at the end of a match are

        - Normialized error (total or rate)
        - Unnormalized luck (total or rate)

Hence a proposal of:

(1)                                               yanwolfy             ianshaw
Overall Statistics:
Total EMG Errors (Errors) in points               -8.329 (-7.290)      -3.549 (-3.441)
EMG Error (Error) rate in mpoints/unf.decision    -39.0  (-34.5)       -16.0 (-15.9)
EMG error rate in mpoints/decision (Snowie)       -19.4                -9.2

Luck Statistics:
Total Luck (EMG Luck) in points                   +2.929 (+2.118)      -2.330 (-4.734)
Luck (EMG Luck) rate in mpoints/decision          +15.7  (+11.0)       -12.3  (-25.0)

- most meaningful figures are shown without brackets (in brackets we have the
less interesting ones)
- units are clearly stated

- EMG values with brackets for Luck and without brackets for errors. This may look
confusing, but in the GUI version Error and Luck are shown in separate panels ...

If the last point is considered really annoying, I wuold live happily with:

(2)                                                yanwolfy             ianshaw
Overall Statistics:
Total EMG Errors (Errors) in points                -8.329 (-7.290)      -3.549 (-3.441)
EMG Error (Error) rate in mpoints/unf.decision     -39.0  (-34.5)       -16.0 (-15.9)
EMG error rate in mpoints/decision (Snowie)        -19.4                -9.2

Luck Statistics:
Total EMG Luck (Luck) in points                    +2.118 (+2.929)      -4.734 (-2.330)
EMG Luck (Luck) rate in mpoints/decision           +11.0  (+15.7)       -25.0  (-12.3)

I'm for #1, Jim is (probably) for #2. If anybody has a preference ... let's agree
on the best one and change the current confusing display. I think it's a must if
we're going for a new "stable" release.


Bug-gnubg mailing list

Robert-Jan Veldhuizen
reply via email to

[Prev in Thread] Current Thread [Next in Thread]