RE: Réf. : [Bug-gnubg] Understanding the stats

[Ian Shaw]

From:  Albert Silver
Sent: 22 September 2006 15:32
 

I think Albert has a point. There are only four factors that prevent a game ending in a draw: Luck on my rolls, Luck on opponent's rolls, errors in my play, errors in opponent's play.

 

The way I read the results below, Opponent won 3 points, but if he hadn't been so lucky he'd "only" have won 1.958 points. However, he is still coming out ahead, which means that he must have more net skill. However, the error analysis says Opponent was a worse player.

 

Why does this contradiction exist? It might be because luck and errors are normalized to a 1-cube. For example errors on 2 cube are rated the same as errors on an 8-cube, even though they cost you less equity.

 

I'm not sure how this works its way through to the overall stats, where the result does depend on the cube level (obviously).

 

-- Ian

I don't think you are understanding the problem. If he was worse, but luckier, why is he the one with an advantage over me? Why should he have a luck adjusted advantage of +0.279 points per game over me when he plays *worse*?

Unless GNU is trying to convince me that it is better to play worse than my opponents, that way I'll have a bigger edge over them than if I were better.

Albert

On 9/22/06, address@hidden <address@hidden > wrote:

>I'm trying to understand the money game stats, but clearly am having
>no luck, so I'm requesting help.

Funny way to state your problem ... no luck ... :))

>I played a session, albeit poorly, which I lost 10 points to 13 over
>7 games. According to the analysis, I played better overall, better
>in checkers, and better with the cube. Despite this, and despite
>losing, I am told I was the luckier of the two. Highly disputable IMHO.
>
>                                      Opponent             Me
>Overall Statistics:
>Error rate (total)                    -3.523 ( -4.811)     -2.269
( -3.939)
>Error rate (per decision)              - 25.7 ( -0.035)      -14.2
( -0.025)
>Equiv. Snowie error rate               - 11.0                 -7.1
>Overall rating                        Casual player        Intermediate
>Actual result                          +3.000               -3.000
>Luck adjusted result                   +1.958               - 1.958
>Advantage (actual) in ppg             +0.4286              - 0.4286
>95% confidence interval (ppg)          3.1388               3.1388
>Advantage (luck adjusted) in ppg      +0.2797              - 0.2797
>95% confidence interval (ppg)          0.9709               0.9709

No, your opp was the luckier.
The actual result for him is +3.000 while the luck adjusted result for him
is
+1.958 : +1.958 < +3.000, this means that he had more luck than you.

Other way to see it, his ACTUAL advantage in ppg is +0.4286 (that is equal
to +3.000points / 7games), but his luck adjusted advantage in ppg is
+0.2797 :
+0.2797 < +0.4286, this means that he has been luckier than you.

But notice that the fact you played better has nothing to do with it.
You can play better (have a lower error rate) and be the luckier of the
two.
Luck does not depend on how you play, it depends only on rolls.