Centaur vs Unassisted Engine

[MonteCarlo]

Draw agreed

Indeed. a5! is a nice resource. Otherwise white keeps at least some token advantage (although still rather drawn).

Nc3 instead of Re1 on move 12 was the engine's anticipated reply, much preferred to Re1.

It also thought Bf4 was a modestly better try than Bg5.

I'll play out the first several moves after 1.e4 in these conditions and let you know what it chooses.

Cheers!

[mbabigian]

Draw agreed

I'm glad everyone finally admitted I was right!

[jefk]

mbbg,

some comments

1) this was just a test game, initiated by a post by ' carldaman'
to see how strong Nnue engines can play nowadays at slow time
control and without book, in other words just an experiment,
whereby i acted as 'typical' correspondence chess player.

2) lessons learned: later in the main thread, about ICCF correspondence
chess draw 'problem' (apparently not yet here but in sight i.e. it
arrive probably soon; it was initiated by a post by LK and rightly so).
So as result of this game against MC, although it was only one game
and we don't know which engines he used (neural network at least i suppose)
i can conclude it's indeed very hard to win for White (and even more
so for Black). Yet with strong play White may sometimes get a slight endgame
advantage, of e.g one pawn more (how to compensate such a
first-move advantage in possible new rules, we don't know yet
although there are some ideas.

3) maybe there's even a better chance to get a better endgame
for White when starting with 1.e4, yet against the Berlin, even with
Re1 instead of d4, the winning chances for W remain very slim
(maybe 'MonteCarlo' now has some more info about this !?!?!).
Anyway, opening theory may well be changed or at least updated as result
of the Nnue evals and analysis, which only is an interesting development
as far as i'm concerned (with the implications for human otb chess).

4) despite the often occurring (and sometimes maybe a bit annoying)
SF 0.00 score in endgames (before the egtb's), it plays strong in the endgame;
yet personally i still think there is scope for improvement in the transition
between middle game and endgame; see eg. also the (controversial) CB posting
by ICCF GM Stephen Ham as discussed under 'general topics',

General conclusion: i don't think (ICCF) correspondence chess will be 'finished',
any time soon, but hopefully it will be changed/improved a little .
Some people prefer blitz/bullet, some others long(er) games. And if we
change some rules in some appropriate and minor/modest style, then it
won't be only battle of the comps as we now foresee is coming soon. Ideally
it most likely still would require some human understanding/experience
with most likely also indeed some computer(chess) skills...So be it.

[carldaman]

Thanks for running the test, jefk and MC! The 'null' result can't be too encouraging for corr players.
Maybe they can get back to the honor system and not use engines. I know, it's virtually impossible to enforce.

[jefk]

well it was only one game, and i played in a careful style.

When a player would try hard to win, eg. by deviating from
the engine-recommended book lines, he/she would actually
(probably) increasing losing chances; nevertheless i think this
this still will happen a lot in correspondence chess coming years.

Going back to forbidding computer use in ICCF is most likely
not going to happen (as you said, impossible to check).