Γειά σας, Τηλέμαχος. Ευχαριστούμε για αυτό τον πίνακα.

The only other asymmetric tables we have are the Trice-Jacobs ones, which
are very old and pre-date the bots. I haven't yet compared it with the
Trice 50 table, which will be interesting.

I wonder why you've used a rollout without variance reduction. That seems
very inefficient. A quick test I did recently had a 0-ply rollout ready
statistical significance about 8 times faster with VR than without. You
might want to consider that if you try a rollout with a greater skill
difference, such as 2-ply against 0-ply.

Best regards,
Ian Shaw


On Sat, 22 Mar 2025, 23:31 Τηλέμαχος Ζωίδης, <[email protected]> wrote:

> Hi,
>
> I have been interested in rollouts between unequal players, specifically
> between 0ply and 1ply. With the aim of obtaining accurate results at score,
> I rolled out a 13-point MET with GNUbg, which I'm sending to you in case
> you want to include it in the next version.
>
> It turns out that the rollouts don't agree with the current theoretical
> model, the Elo system. This has been noted by many people over the years,
> including Joseph Heled. For more information:
> https://vixra.org/abs/2409.0106
>
> Keep up the good work,
> Tilemachos Zoidis

Reply via email to