Is there a point to playing against Twixtbot? TWIXT PP
21 replies. Last post: 2019-11-26Reply to this topic Return to forum
Hairspring at 2019-11-17
This is a real question I am pondering. On the one hand I see it as a waste of my time. I only have enough time to take a look at a few possible moves in any one of my games. In fact I had so much actual life going on that I timed out of a lot of my games last championship (sorry about that guys). I just don’t ponder all that long before I plop down a peg (they are still pegs to me, I have my original 3M board still) because I have things to do like work and make dinner, pay attention to the news and raise my kids. Twixtbot does nothing but play twixt and can run through multiple iterations of outcomes for every move I might make against it. It is a cool program. It is a well crafted machine. It is like trying to hit against a pitching machine that throws 400 mile per hour fast balls. Might as well close your eyes and swing, maybe you will hit something. (sorry about the metaphor from a peculiar, American sport.)
I Guess the other side of the argument is that Twixtbot can raise our level of play if we human players study what it does and learn to integrate lessons from it into our games. Can we do that? Like I said, my time is limited and maybe I just can’t commit that much effort to a board game. Is anyone looking closely enough at the games Twixtbot plays to garner actual practical knowledge for how to be a better TwixT player?
P_Guz at 2019-11-17
Actually, yesterday I spent some time on playing with Twixtbot (https://littlegolem.net/jsp/game/game.jsp?gid=2133166). I decided not to bother to much about the result but I tried to bring some quality. And I enjoyed a bit this game. I am living under assumption that till 11th move the winner was not determined. After all, what I am missing when playing with bots, is some small feedback given to me about my biggest mistake or suggestion about better moves. Learning from games with human players is much simpler, cause we are `on the same wavelength`.
TwixtBot at 2019-11-17
Sounds like it is time for TwixtBot to retire. After this tournament is over then.
MisterCat at 2019-11-18
emergency post from cat: NO, it is not. Hairspring started this thread to invite discussion, and that’s a good thing; rash actions are a bad thing. Sorry to be brief, but please bear with me – my day off is Tuesday, and I will elaborate then.
Over? No. No point in playing? No. Learning? Problematic, and needs discussion. What is not needed is anything rash.
isometry at 2019-11-18
While super bots can be frustrating to play against, they are great tools for teaching us about openings. Hey TwixtBot, would you consider sharing which first moves you think should be swapped and which should not? (In theory, some could be neutral – leading to a draw. I think I brought that possibility up many years ago, to which some protested.)
P_Guz at 2019-11-18
leela_bot revealed his opening in hex (thread: https://littlegolem.net/jsp/forum/topic2.jsp?forum=50&topic=739)
isometry at 2019-11-18
Hairspring at 2019-11-18
I Really did open the question as a discussion not as a condemnation of Twixtbot playing here on LG. I was feeling philosophical on Sunday morning while taking a quick look at the openings in the current 1.1 tournament and saw that some had forfeited their games against Twixtbot. It may be that I can be convinced to play the games through with Twixtbot. I just have to figure out how to learn something from it that is worthwhile. Maybe it will require applying it to Alan Hensel’s opening move data to change how to view and play the opening game. Maybe there it is just useful if I play enough games against Twixtbot to get some lesson through my thick skull. I don’t know, I really was curious what others thought when I posted, and I still am curious.
Galdian at 2019-11-18
In my opinion, if bot is an undisputable champion, it should be retired from the championship and other competitions, while people, who are interested in playing with it, still might do it in non-tournament games.
Being an LG champion is something to strive to, I don’t see any reason to take those chances from humans.
Florian Jamain at 2019-11-18
But is he really unbeatable now?
Maybe gonna play him one more time if I got the motivation.
Hairspring at 2019-11-18
I do not know about unbeatable but it is certainly hard to beat. I would be equally hard to beat if I could map out outcomes for move combinations at the same rate that the bot does. At least I would like to think so. If I had my way I would learn to play TwixT in a manner that was much more akin to art. By that I mean that I simply wouldn’t have to try to run down all of the alleyways of potential moves but, rather, would be capable of looking at the board and judging where the next peg should go based on accumulated knowledge of the game and my own intuition and creativity. The human mind that has gained true expertise at something can put things together that are astoundingly astute and can do so on a subconscious level. I almost always at first look at my position in a game and plunk down a peg where I think it should go. Just throw it out there as a first glance move. I am always tempted to just hit the send move button but usually I go and look at it on one of the analysis tools to see if it really is a decent move or not. Ultimately I guess my question boils down to this: Can playing against Twixtbot help me to get closer to a place where I am a true expert at the game? Can it help me be able to look at a position on a TwixT board and know where to put my next peg without using any deep tree of move analysis at all? Or is it just going to frustrate the crap out of me? Like I said, I was feeling philosophical when I posed the question.
Whatever the case, I hope you play the bot again Florian Jamain, I would love to watch. I might learn something from that, maybe.
gzero_bot at 2019-11-19
One option for the (neural network) bots is to run with a very constrained amount of search. In other words it will rely on the intuition of the network rather than try every possibility up to depth n, giving it an impossible tactical advantage to overcome. At least for gzero in hex and connect6, I guess it would be still be pretty strong, but certainly not super human strength.
I am going to run gzero (in waiting room) on hex19 going forward with limited search time (around 10 seconds) to see it how it fairs. I welcome anyone to put this to the test!
MisterCat at 2019-11-19
I am going to make lots of remarks in the threads below.
David J Bush at 2019-11-20
Instead of dumbing down the bot, how about implementing row handicapping?
Of course this is a lot of work for Richard as well as Bony J. It is available on ig Game Center.
The smallest handicap is to eliminate the swap rule. Beyond that, one dimension of the board is reduced.
The weaker player moves first and has less distance to cross. Of course there is never any swap in a handicap game.
Such games should probably not be rated.
passenger at 2019-11-22
I’m a little bit dissapointed that Twixtbot “stole” the last championship from me, but it’s not hill I’m ready to die on. I would leave the decision to better players like Florian, Maciej or David since they are most qualified to say if humans are still able to compete with the bot. If not, I know that I’m not:) With such a big level gap, it’s not fun for me anymore.
Or maybe I’m just still bitter about the championship:)
David J Bush at 2019-11-23
I believe the bot has made its point. It is far stronger than I am. I would love to see a match against Maciej, Florian, or Gyorgy, but this could always happen as a challenge game outside of the championship. So I personally would prefer no more bot in rated tournaments, at least not until and unless someone asks for its return.
I still harbor the conceit that I might learn from the bot’s lessons and improve eventually, but I’m definitely not there yet.
MisterCat at 2019-11-23
Hey, David, I made some comments about this in my ‘Twixtbot (a)’ thread. I’m not TELLING you to post there – just mentioning that I made remarks in this regard. Who am I? Well, just an experienced Twixt player, NOT at the level of you or Passenger, who CARES very much about this game. Also, my remarks are pretty much IN AGREEMENT with what you said here. Also, Bony Jordan HAS read my comments, and replied to me.
Also, once I am retired, my plan is to get in my car in New Jersey, drive down to pick you up in Virginia (??), drive us BOTH down to North Carolina, meet up with Hensel (my buddy), and play live Twixt? Yes, I will lose miserably to both of you, but I will have LOTS OF FUN!
David J Bush at 2019-11-24
Heh how about you pick us both up and fly us to Hamburg?
Galdian at 2019-11-26
What is the relation of Hamburg and TwixT PP?
David J Bush at 2019-11-26
There was some mention, in a thread years ago, of a tournament there. But I was just preying on Mr. Cat’s generosity. Hey hey, why not fly me, Alan, AND his wife to Hawaii? And a bag of chips please.
MisterCat at 2019-11-26
You may recall some years back when I orchestrated a ‘live Twixt’ event, successfully, in New York City. Four of us played games and had some worthwhile discussion regarding the games here on Little Golem. Over the years, there have been a few other attempts at live events; I believe there may have been one or two, in Europe.
Not everyone has the money to attend the Mind Sports Olympiad. I certainly don’t, and even if I did, I have no passport or any inclination to get one.
But playing games live is an enjoyable social event; for example, I play Chess now with some players at my local Panera Bread on Saturdays. Also, when you play live, cheating is next to impossible. (‘bathroom breaks’ notwithstanding, as seen in Chess tournaments)
My comment here was because I know that David, Alan, and myself are all on the East Coast of the U.S., albeit not all that proximate. If the time could be found, and the inclination were there, I’d love to do another live games event. I would certainly help organize and arrange things, as possible. My reference to retirement is simply because these things take time.
And if I sounded overly generous in my proposal, David, perhaps its because I was a bit drunk at the time.