Hi,
about the so-called "basic" levels, the manual is somewhat vague, even a bit obscure. I quote:
'
Basic: The computer takes a certain average time per move ̶ shortest on level 1, longest on level 10.'
Very quick tests show level 1 is obviously much faster than level 10, therefore the manual statement is true, yet not much informative.
Nothing is shared about the thinking time behavior in-between the extreme levels 1 and 10; nothing about the average time per move you should expect.
From this thread:
Zitat von
Wolfgang2
Udo, diese "Basic" bzw. eingedeutscht "Normal"-Spielstufen sind in der Suchtiefe limitiert, aber auch etwas anders abgestimmt als die Zeitstufen. Diese sind im Prinzip geeignet für Lernende die über die Spaß- bzw. ("FUN") hinausgewachsen sind.
Zitat von
Bryan Whitby
ChessChampion 2.0 Levels
Basic Level 1-10 (1 sec - variable minutes)
Zitat von
Wolfgang2
Die "Normal"-Stufen sind zeitlich variabel, das stimmt. Da diese Rechentiefe orientiert arbeiten. In komplizierten Fällen, bei welchen nach 30 Sekunden die Suchtiefe noch bei 2 Halbzügen ist, dann einfache Fehler passieren, kann man mit "Normal 10" besser bedient sein. Allerdings denke ich, dass per Saldo 30s/Zug (fix) etwas stärker spielt, da im Endspiel dort tiefer gerechnet wird.
In order to shed a ray of light, is used a stopwatch to assess the moves thinking time across two games from the Spacious Mind series: #1 F. DI CASTELLVI vs N. VINYOLES (41 half-moves), and #5 Capt. SMITH vs F.-A. DANICAN PHILIDOR (58 half-moves).
The first one is an open game, clear and rather tactical, and the second is a closed one, complex and rather positional. First tests revealed much faster moves in game #1 than game #2, thus providing an interesting range of measures. The average times result from the total time used over the whole game, both White and Black moves, divided by the number of played half-moves.
From the second game I also selected three positions where most levels spent a very large thinking time with regards to the usual ones, and I used the Save/Load feature to repeat these complex, time consuming positions for a dedicated, single move time measure.
Here is the data.
As you can read, the distribution of the thinking times is not simply rising from low to high levels; there is a breakthrough change at level 6. The ranking of levels do not result from an average time granted; the very large time occuring for complex positions using upper levels (variable minutes, as stated by Bryan) is typical of the search depth approach mentioned by Wolfgang.
Considering the thinking times that are very close per couple of levels 2-3, 4-5, 7-8 and 9-10, plus levels 1 and 6 that look apart, there must be something else than search depth creating the scaling of levels. I think this must involve software skills throttling, and the basic levels ranking would result from combining these with the search depth. Level 6 could be full strength evaluation combined with minimalist search depth, whilst level 5 would reach deeper depth but using limited evaluation.
From this point on, I understand why did the user manual not explain the basic levels in further details: that's too complex. They definitely are "levels" with the meaning of variable strength; the thinking time being no designed target - just an outcome.
As a conclusion, to keep something useful out of the data, I made this educated guess about the
thinking times.
I bet the actual average times would fit within these, as the games I used did not involve long endgame phases, potentially reducing the overall average (per limited search depth effect).
MfG,
Tibono