Twixtbot vs. The World part II TWIXT PP
13 replies. Last post: 2020-01-14Reply to this topic Return to forum
MisterCat at 2020-01-07
The first thread is getting SO LONG that you have to yank that scroll bar down a whole lot, which is inconvenient.
Speaking of yanking the scroll bar down a whole lot, the WORLD vs. TWIXTBOT analysis at Alan’s commentator is getting HUGE. You really have to keep scrolling and scrolling to get to the latest comments. With that in mind, I asked Alan Hensel for a fix – and he delivered!
You can add a tab to your Chrome Bookmarks bar on top to automatically scroll to the bottom of the open Commentator page. Here are Alan’s instructions, verbatim:
TwixtBot at 2020-01-07
What if you made a new commentary game for each move? That would help with the scrolling issue.Alan Hensel at 2020-01-07
Hm, just before I read this, I searched for “6.r20” in the Commentator thread to see what we had previously said about it. That would not work so well over separate pages.David J Bush at 2020-01-10
What happens if we LOSE this game? What happens if we WIN? A draw??
If a team of top players, working together in consultation, STILL loses to Twixtbot, then besides being DEPRESSED, I believe that folks are going to feel this way: there is no point in ever playing Twixtbot; it is too damn strong. Twixtbot should stay completely out of all tournaments and rated play. I will go further, in MY viewpoint, that Twixtbot can’t even be used as a tool for learning an analysis; no human being can preview 50,000 games per moves and tabulate the results, so it will not be possible for human play to model itself after Twixtbot ‘type’ moves. This will point to BonyJordan basically giving up and retiring Twixtbot permanently, unless he just wants to keep his server attached for the odd practice game (or new player) that might occur. He set himself a task years ago – create software capable of defeating all the best players at Twixt; after many years, the goal has been reached. Bye! Kind of a shame, though, and I batted around a few suggestions in another Forum thread, which I’ll add to here:
a) bring back Twixtbot jr., who may suck, but is still a useful practice tool, IMO. Of course, I don’t know the particulars of BonyJordan’s hardware, and whether it is practical to keep both programs active on his computer.
b) expand the Twixtbot to include levels of play; you can weaken it by allowing the trials to vary from 50,000 down to, if need be, zero. Could TB still play an intelligent game with NO trials, based on the existing neural net, or won’t that work at all? It doesn’t HAVE to be zero; it could be, for example, 10,000, or 1000, etc. I also proposed randomly programming in ‘mistakes’; cases where it plays 2nd-best or 3rd-best moves; the opponent won’t know when this happens, so can prove to be a useful tool for training. Also, David proposed ‘handicap’ games, where the number of rows and columns changes to favor the human player, though personally I’m not a fan of changing the game this way.
c) add an analysis tool, where any specific position can be programmed into TB, and it will respond with a ‘best move’ and give sample lines.
d) create a portable Twixtbot that can be downloaded, plays off-line, or can be played on a cell phone (and, as I said, sell this).
These suggestions are to preserve the value of Twixtbot if we LOSE, and nobody ever wants to play it (and there already ARE a number of folks here who are of that mindset). Look, if the team loses and is simply determined to repeat the experiment and try again – well, I’m all for it, but perhaps next time with a different moderator, and maybe different rules for selecting the team move in place.
On the other hand, if we WIN, then HOORAY My above suggestions remain for the creator, but won’t be as necessary – as Twixtbot will have plenty of use here for all.
And if a draw? Well, I think I’ll leave this possibility alone, for now.
(meow)MisterCat at 2020-01-11
that’s strange – my paragraph in the above post after the word ‘HOORAY’ got CLIPPED! Now I have to try to remember what I said. What’s up with this, Richard??! Besides the sense of accomplishment and satisfaction that will be felt all around, it means something else: Twixtbot CAN BE DEFEATED by human play; play carefully, accurately, and with precision and you will WIN. I think people will be MORE inspired to play games against the Bot on their own, and TB will get plenty of use. I still think that it should stay out of tournaments, however.
I think that’s basically what I said above; maybe I said something else, but I don’t remember. It is annoying that my post got chopped – perhaps happened because I was using different font setting and colors.
meowypercube at 2020-01-11
Yeah, that’s more or less what you posted there. It got lost due to some formatting issue. Copied, formatting removed:
> Besides the sense of accomplishment and satisfaction which will be felt all around, I believe that it will change everyone’s attitude about the Bot: it’s play is EXTREMELY STRONG, but CAN BE DEFEATED by humans, playing the most accurate moves with care. In this case, I would expect that more people will be inspired to try on their own (and maybe succeed), and play against the Bot will increase with enthusiasm (though, still, probably it should stay out of the tournaments). In fact, we could see a big jump in the overall quality of human play in general.
> My above suggestions remain for the creator, but won’t be as necessary – as Twixtbot will have plenty of use here for all.
> And if a draw? Well, I think I’ll leave this possibility alone, for now.
> (meow)Alan Hensel at 2020-01-11
286 comments into the TwixtBot vs. The World discussion, and we’re still discussing the 4th move. Three and a half days – about half a week – remain on the clock.
Mr. Cat, I’d like to suggest a “Pop-Tarts are Ravioli, Change my Mind” strategy. Every time TwixtBot makes a move, suggest the next move and start a 48-hour clock. If someone changes your mind, post what you’ve changed your mind to. When the 48 hours are up, play the last move you changed your mind to.
That still loses 12 hours per move. What do we have, 20 vacation days?MisterCat at 2020-01-11
Alan, I’m the laundromat and will be home in an hour or so. You can tell the gang at Commentator, because I can’t log in on my cell phone. Not easily.
Something above about my taking steps to speed up Team analysis and resolution of moves. I’ll read again, later, and see what I can do.
Ypercube, thanks for retrieving my post. You’re a wiz with this stuff. How about adding some comments and critique, while you’re at it? Also, rescue our Twixt game with a brilliancy!David J Bush at 2020-01-13
In a few more moves, the position should clarify, and then we can move quickly enough to accumulate time.David J Bush at 2020-01-14
I invite everyone to the new analysis page.
less time loading & scrollingDavid J Bush at 2020-01-14
Sorry please don’t use this. Use the link in the “part 3” thread.
Reply to this topicContact 2014 ©LittleGolem --23--