Championship 22 Breakthrough
14 replies. Last post: 2015-08-25Reply to this topic Return to forum
wanderer_bot at 2015-04-22
If I may paraphrase Tasmanian Devil: If you are signed up for, or are interested in participating in the next championship round, you are hereby notified that there is currently one single move left in Championship 21.
edbonnet at 2015-05-28
Is it okay for you to play a different continuation in the games that went back a few moves?
Of course, I mean a really different continuation (not a simple transposition).
Personally, I think it’s highly disrespectful towards the opponent and should not stay unpunished.
wanderer_bot at 2015-05-28
If you are saying that Wanderer did this, I apologize. He plays automatically and the random nature of his algorithm causes him to play different moves from the same position. Certainly no disrespect was intended nor was he trying to undo past mistakes.
However, if you are talking about everybody else, I agree with you!
edbonnet at 2015-05-28
No, I was speaking in general. It was a collective “you” ;-)
edbonnet at 2015-08-15
Ok, it seems that we’ll have to congratulate the alliance MojoRising-Stop_Sign for winning the 22nd championship.
I find somewhat strange that someone of the strength of MojoRising played 49.d3-c4 in his game against Stop_Sign
(and went on to lose what was a totally winning position)
when there is a straightforward win with 49.f5xg6 50.f7xg6 51.e4-f5 (it’s so winning that White don’t even have to exchange first on c6).
At that time, MojoRising had already lost too many games to be in contention for the title, so probably lost interest in his remaining games;
with the small side effect that it will change the winner of the championship.
Anyway, the Breakthrough championship is too heterogeneous to really make sense.
Stop_Sign at 2015-08-16
michelwav ★ at 2015-08-16
Ed, I don’t understand what you mean in your last sentence.
I think this championship is very hard : the 7 presently best ranked active players are playing in it and every body played good games in it.
But every one among us (except wanderer) is able to make big mistakes (not only Mojo R).
Maybe Stop S. and you will get the 2 first ranks : it would not be a surprise.
Sorry if I misunderstood your post.
wanderer_bot at 2015-08-18
... “except wanderer”. Don’t be so sure. From this position Wanderer played d8-c7.#1689308 ahhmet vs. wanderer_bot
8 7 6 5 4 3 2 1 a b c d e f g hWanderer blunder
michelwav ★ at 2015-08-18
I see ; so I should remove “except wanderer” from my last post.
Ray Garrison ★ at 2015-08-24
why would a computer blunder by simply hanging a piece? Is this a human data entry error? (my guess is the computer suggested trade first followed by d8-c7. but the human operator was analyzing the board after the trade, then forgot abut the trade first, oops!)
wanderer_bot at 2015-08-24
It was not a human error. The bot both figures out the move and plays it on the board. There is no human intervention.
Indeed, the blunder is a bit surprising and caused me considerable grief. The short explanation is: even though Wanderer is an MCTS based program, it still uses an evaluation function. (The algorithm it uses we refer to as MCTS with EPT -- “Early Playout Termination”.) In particular, the evaluation function contains weighted values corresponding to the piece counts (of course) and to strategic values that encourage things like not moving d and e pieces off of the first row to early, not moving too deep into enemy territory too early, etc. This game shows that the weight on the strategic values was too high. In particular, though it’s hard to track down precisely, I think it thought the white piece on c5 was a liability to White while Black had a very solid defensive position so it compensated for the piece lost.
Needless to say I dialed back the strategic weight. It won’t make that move again. :) But dialing it back too far and then all it cares about is piece count and basically ignores strategy entirely. The evaluation function is extremely delicate. (I’m wondering if that’s why Wanderer seems to be performing a bit worse recently....)
edbonnet at 2015-08-24
To me, the real mistake is 26.g7-f6 instead of 26.d8-c7.
Even if wanderer does not let c5 hang, it has to trade on b4 and then White is completely winning.
The pawn d1 will advance to a4, and, depending on how Black defends on the queenside, something along those lines:
b4-b5, a4-a5-a6 or a4-a5, c4-c5, a5-b6, b4-a5 followed by a sac on dark squares will eventually be crushing.
It may seem far-fetched but, playing White with two extra tempi (one natural, the other due to the trade in b4)
AND an unbalanced position (kingside/queenside) is really easy.
michelwav ★ at 2015-08-25
Congratulations to edbonnet for his victory in the breakthrough championship.
The game between edbonnet and Stop_Sign was particularly interesting.
We can also congratulate Stop_Sign and wanderer_bot for their good tournament,
and Mojo_Rising and David_Scott for their victory against wanderer_bot.
According to me, it was probably the most difficult breakthrough championship I ever played.
edbonnet at 2015-08-25
Wait for Luffy to step in for next championship ;)
I agree it was (still is?) tough. I was worse/losing against at least Stop_Sign, wanderer, David Scott...
In late response to your post of 08-17, Michel, I also agree that the games in the championship are interesting and difficult
but I’m not convinced the present system qualifying the first two (instead of just the winner, or the winners in case of tie) in both championships 2 is desirable.
I think it’s a bit brutal for the (second) qualifiers. Indeed, there is quite a gap between the average rating of championship 2 (~1750) and of championship 1 (~2050).
Then again, it’s just my point of view and if the qualifiers find the challenge interesting, then it’s probably better that way.