lichess.org

Beating highest level stockfish handicapped by a piece

@nayf
"I look at it as an amazing opportunity to sit down and play the greatest player in history, who could wipe the floor with Fischer, Kasparov and Carlsen playing together."

Except the fact that even Stockfish level 8 is limited on this site and Carlsen would almost certainly have not much problem crushing the version you play against, this would be true If you somehow managed to play stockfish used for actual analysis of games which would be a bit of a problem as you would be "spying" on its ideas.

Holy cow, I ve just f.cked up a captcha, someone kill me please!

I understand better your perspective now. you really find its play style brilliant. and not only that it is the strongest player, but also the greatest.

I` hope actually that you were the brilliant one when you managed to keep or convert your material advantage.

Because i am very skeptical of those qualificative applying to that player. perhaps the developers that chose its strict behavior constraints artfully or successfully, could be considered instead, not the mechanical risk averse and not very surprising engine.

i rarely find its moves, whether i could see it or not before, surprising. or temporarily baseless, or expressing threats with unnerving ambivalence, to keep some unknown into what are its actual intentions in the particular moves chosen. or maybe i want to keep some illusions so i always find its moves non-surprising...after the fact (easy posture to adopt).

so maybe exhaustive search will be glorious. i am not sure either.

@gayweather

The version I play is on droidfish. I think its elo is around 3200, more than any human has ever approached.

@Irishman964

How do you know?

Winning a blitz game a piece up against SF is very difficult. I failed miserably.

(SF in the "Fritz"-GUI unleashed.)

@Sarg0n

A minor piece? Yes, I think it must be close to impossible, although I wonder if some GMs could do it.

Have you tried with a rook? Kingside rook should be easy for you.

so SF-Fritz, SF-lichess, SF-lichess-eval, SF-droid-fish are not comparable? not having any consistent way of defining strength?

So the gui is affecting the strength of the engine hooked to it?
are there no tables of equivalence based on the engine exposed parameters (to the gui)?
are those gui allowing precise algorithm parameter settings?

i only know SF-lichess opponent, and the eval engine with
memory, core number, and number of lines shown, and the depth being adjustable to some extent.
I get the meaning of all these but for the number of lines shown: is that just a change of display, or is the eval engine actually changing how it drops line candidates or keep them for some deeper search as well? *

ELO does not mean much to me, as i never played in any competition, and never will.

Chess search parameters might be more palatable for me. so if the guis are using UCI communication protocol to change the search parameters, i would start looking more into those alternatives.

there are more dimensions than just rating or a 1 parameter strength level, i believe, or would like to figure out.
Different training benefits could be obtained from changing those 2 or 3 dimensional parameters values.

I thank the op for his magnanimous attitude in allowing me this slight drift, taking advantage of the posters experiences with various GUIs interacting with same engine.

one more: why is only the search depth being often exposed in lichess engine evaluations? paraphrasing *

if you have ideas related to my questions, feel free to PM me.

No, I just tried a minor piece which proved to be impossible.

„Die Uhr ist auch eine Figur.“ (= the clock is also a piece which rhymes in German)

I literally posted this about a week ago. "Handicaping the engine"....

en.chessbase.com/post/komodo-9-odds-matches-against-gms
so GM's year 2015 struggled with two pawn and offers a table for estimating impact of odds (based on kaufmans researcs I assume)

Odds
Description
Appr. Elo value
Queen odds Remove Qd1
2000
Rook odds Remove Ra1
1400
Knight odds Remove Nb1
1100
Two-pawn odds Remove two pawns
400-700
Exchange odds Remove Ra1 and Nb8
500
Pawn and two moves Remove f7, White plays two moves to start
600
Pawn and move Remove f7
450

So Stockfish is about 3300 and knight shoudl be about 1100 == 2200 Elo shoudl have chance. But yet another thing is that computers gain less on time than humans. branching factor makes sure that engine gets most of things right in just few seconds. Leela/A0 type engine gains more on increased thingking than traditional

Reconnecting