Hi and sorry for the late reply. In Greece where I'm from, we actually play 3 different games collectively known as Tavli. One of them, Portes, is what you call backgammon but without the doubling cube. Comparing the skill in these games is a favorite topic amongst Tavli players. Backgammon players of course think that Portes have the most skill, especially when played with the cube. To me, this isn't obvious at all. Sure, the cube adds a few decisions in the games that continue with a Take, but it removes a lot more decisions in the games that end with a Pass. The situation seems even worse in match play because cube decisions tend to be easier for the trailer who is more likely to be the weaker player.
When I discovered that GNU had the capability of rollouts between unequal players - to me, this capability is enough to prefer GNU over XG if I had to choose - I realized that I could test my hypothesis. Having a background in math, I also realized that I could define skill in match play using the ELO system. I had read lots of old forum posts on bkgm.com that claimed the ELO formula doesn't agree with real-world data. Thus, I could kill 2 birds with 1 stone: compare skill and arrive at data-driven formulas for each variant. I finished my research in 2024 and uploaded the results on viXra in order to share them and get feedback. In the first version I modelled certain cases mathematically which turned out to have been a bad idea and which lead to lots of revisions as I ended up rolling everything out. I have withdrawn all past versions to avoid any confusion. The current version can be found at https://vixra.org/abs/2506.0144 or at https://doi.org/10.5281/zenodo.16375841 where you can also find .xml files containing the METs I rolled out in GNU format. Happy to answer any questions. Tilemachos Zoidis PS: Not sure how to reply so that the response is properly threaded.
