In a different threadm Chow gave a link to an old thread and while
skimming through posts in it, I saw this:

https://groups.google.com/d/msg/rec.games.backgammon/-Y9AWXOFwrc/bxGv4TIUuNgJ

On September 3, 2009 at 8:55:02 AM UTC-6, tc...@lsa.umich.edu wrote:

> I didn't see an answer to my question about using dice generated
> by some method other than the Mersenne Twister.  Does your offer
> stand for other dice generation methods too?

It got me curious also. I had never played "money sessions" against
GNUBG or/nor using other dice rollers.

GNUBG has 4 other dice rollers besides Mersenne Twister. I thought
"perfect"! I decided to play a session of 100 games, using each of
those rollers for 25 games and so I did (setting the seed to the
current date "20170201" for each roller).

First I wanted to play against GNUBG 4-ply but it's a little too
slow and I get distracted by trying to do other tasks while waiting
for it. So, I played against the Grandmaster level, which I find a
little faster than XG Roller++, which in turn seem a little faster
than GNUBG 4-ply.

Each 25-game section took about an hour and here are the results:

Games 1-25 using ANSI roller, MK 44, GB 19
Games 25-50 using Blum Blum Shaub roller, MK 52, GB 31
Games 50-75 using ISAAC roller, MK 51, GB 24
Games 75-100 using MD5 roller, MK 31, GB 29

My results against the first 3 rollers were similar but I had quite
a difficulty against MD5... Probably just a bad streak??

Combined results: MK 178, GB 103 (MK = +75)

The session was video captured similar to a session against XG Roller++