5 ms·

Adaptive Rock Paper Scissors

10 points by dangoldin 18y ago

trickjarrett 18y ago
I went 20-5-10. Interesting algorithms, I could see it's "strategy." 3 of my losses came from me picking scissors three times in a row. And then I correctly deduced it would guess that I would switch to rock and thus play paper, so I clicked scissors a fourth time. Will be curious to see how the statistics play out. Would also be better to randomize the order of Rock - Paper - Scissors links as it's given to the user.
- dangoldin 18y ago
  For me I was getting a lot of wins in a row and as soon as I got a loss I would switch strategies. That ended up working pretty well.
- tectonic 18y ago
  Yes, randomizing the order of the links for the user each time would let the user press the first link every time and have an unbeatable strategy. :)
kqr2 18y ago
They should try their AI on one of the rock-paper-scissor (rps) variants. For example, this guy has developed rps-25 versus the traditional rps-3: http://www.umop.com/rps25.htm http://www.umop.com/rps25.htm
tl 18y ago
"Against me, you've won 28 games, lost 14 games, and tied 18 games." Is there an api that we can use to create our own rps decider to play against your site or other people?
- dangoldin 18y ago
 This isn't my site. I just stumbled into it.
- kqr2 18y ago
 You should be able to make requests of the form: http://iterationprojects.com/rps/?mode=p&id=<integer player id>&rps=<r,p, or s>&result=<t,w,l>&r=<random # between 0 and 100000> where: r,p,s represent your rock, paper or scissors choice. t,w,l represents whether tie, win, or loss against previous prediction The response will be like: Prediction: p I have played a total of 25751 games. I've won 40.895% of those -- 33% would be chance . (7537 ties, 10531 wins, 7683 losses.) strat1 (0, 0, 0, 0, 0, 0) using prior where the computers next move is charAt(12), in this case 'p':paper
herbyderby 18y ago
This one has been around for a long time: http://chappie.stanford.edu/cgi-bin/roshambot http://chappie.stanford.edu/cgi-bin/roshambot
tectonic 18y ago
This is one of my projects. I made it a while ago. I'm glad you guys are enjoying it! It implements an approximation of this strategy: http://www.ofb.net/~egnor/iocaine.html http://www.ofb.net/~egnor/iocaine.html
lionhearted 18y ago
My results: Against me, you've won 11 games, lost 0 games, and tied 9 games. Maybe it´d make sense to add a slight random element to throw off intelligent human players? 10% chance of picking one completely randomly every time? It might lead to people who could break the system picking up patterns wrong.
lazyant 18y ago
Nice. I chose as random as possible (for a human, we're bad at that) and I tied: "Against me, you've won 23 games, lost 23 games, and tied 21 games".

10 points by dangoldin 18y ago

I went 20-5-10. Interesting algorithms, I could see it's "strategy." 3 of my losses came from me picking scissors three times in a row. And then I correctly deduced it would guess that I would switch to rock and thus play paper, so I clicked scissors a fourth time. Will be curious to see how the statistics play out. Would also be better to randomize the order of Rock - Paper - Scissors links as it's given to the user.

dangoldin 18y ago
For me I was getting a lot of wins in a row and as soon as I got a loss I would switch strategies. That ended up working pretty well.
tectonic 18y ago
Yes, randomizing the order of the links for the user each time would let the user press the first link every time and have an unbeatable strategy. :)

kqr2 18y ago

They should try their AI on one of the rock-paper-scissor (rps) variants. For example, this guy has developed rps-25 versus the traditional rps-3: http://www.umop.com/rps25.htm http://www.umop.com/rps25.htm

tl 18y ago

"Against me, you've won 28 games, lost 14 games, and tied 18 games." Is there an api that we can use to create our own rps decider to play against your site or other people?

dangoldin 18y ago
This isn't my site. I just stumbled into it.
kqr2 18y ago
You should be able to make requests of the form: http://iterationprojects.com/rps/?mode=p&id=<integer player id>&rps=<r,p, or s>&result=<t,w,l>&r=<random # between 0 and 100000> where: r,p,s represent your rock, paper or scissors choice. t,w,l represents whether tie, win, or loss against previous prediction The response will be like: Prediction: p I have played a total of 25751 games. I've won 40.895% of those -- 33% would be chance . (7537 ties, 10531 wins, 7683 losses.) strat1 (0, 0, 0, 0, 0, 0) using prior where the computers next move is charAt(12), in this case 'p':paper

herbyderby 18y ago

This one has been around for a long time: http://chappie.stanford.edu/cgi-bin/roshambot http://chappie.stanford.edu/cgi-bin/roshambot

tectonic 18y ago

This is one of my projects. I made it a while ago. I'm glad you guys are enjoying it! It implements an approximation of this strategy: http://www.ofb.net/~egnor/iocaine.html http://www.ofb.net/~egnor/iocaine.html

lionhearted 18y ago

My results: Against me, you've won 11 games, lost 0 games, and tied 9 games. Maybe it´d make sense to add a slight random element to throw off intelligent human players? 10% chance of picking one completely randomly every time? It might lead to people who could break the system picking up patterns wrong.

lazyant 18y ago

Nice. I chose as random as possible (for a human, we're bad at that) and I tied: "Against me, you've won 23 games, lost 23 games, and tied 21 games".