EinStein in Computer Olympiad

Hello,

this year the Computer Olympiad will take

place in Tilburg (NL), in the middle of November.

Website Tilburg 2011

Amongst several other games also "EinStein wurfelt nicht"

will be played, for the first time. All bot programmers

are invited to join the event.

Ingo.

I'm thinking about it...

I guess submitting a a bot with a simple heuristic could be a good idea since you just might win :p

I cannot find game logs on the site sadly :/ would be interesting to see.

What do you been by game logs?

I mean logs of what moves were made. Like on little golem where you can see the moves from old games.

The event is in NOVEMBER 2011, so it is still to come.

Perhaps, some logs will be available after that event.

Yeah I meant from previous events.

@Thnikkaman: It is the first time that Ewn is played

at the Computer Olympiad.

I have written a report on the origins of EinStein würfelt nicht, see

here.

Ingo.

I haven't joined this event officially but I want to join it in the next days...for the very first time on a 'computer olympiad' with my 'naive' little child...I'm the total beginner in programming, let's see what happens ;)

Which bots or members of LG will also be there?

I will be there with my Chess program 'The Baron' and Havannah program 'Gambler', although the latter needs a lot of work to become competitive. Perhaps, if time permits, I'll have a go at EWN as well.

I would have rather liked to see a real life EWN tournament with human players (propably also because I was not able to take part in the one years ago in Jena).

Are there any plans for such an event, Ingo?

Richard Pijl wrote:

> ... Perhaps, if time permits, I'll have a go at EWN as well.

That would be very nice. I have some hope that also Johan de Koning

might enter the competition. We will see. By the way,

my idea is NOT to make the competition a race in hardware.

We should not have remote machines, and also the computers in

Tilburg should be "normal" ones.

Ingo.

McGregor wrote:

> I would have rather liked to see a real life EWN tournament with human

> players (propably also because I was not able to take part in the one

> years ago in Jena).

> Are there any plans for such an event, Ingo?

Not directly. But the next "Long night of Sciences in Jena" is approaching.

It will be on Friday, November 25. It will be possible to play EinStein in

that event. An when several strong players will come to Jena, we might

even have a "pro tournament".

Ingo.

Here is a link to the games station in the "Long Night of Sciences"

in Jena. The text is in German only, unfortunately. However, also

international visitors are welcome. Please, contact me, when you

need help.

Optimierung - Geldwelt - Spiele

Ingo.

Unfortunately ICGA managed to schedule the Olympiad and WCCC in the same week as the long night of Sciences. As my priority 1 is still chess is means that I will not be coming to Jena this time either :-( And with only 2 months to get Gambler competitive in Havannah (again?) the chances that I will produce an EWN version of Gambler is not big either. But I will be watching the games in Tilburg.

> Unfortunately ICGA managed to schedule the Olympiad and WCCC in the

> same week as the long night of Sciences.

Yes, that also came as a shock for me. I will be in Tilburg until Thursday

(November 24), and the station of my group for the long night will be

smaller than in previous events. Especially, Havannah will be played

only on the back burner: perhaps one exhibition game between a human and

a computer - and a table where visitors can learn the rules.

> ... the chances that I will produce an EWN version of Gambler is not

> big either.

No problem at all. By the way: your name "Gambler" fits better for an EWN

bot than for a Havannah bot.

My information and the link above was mainly meant for human Ewn players

who would like to participate in a real world Ewn tournament.

Ingo.

Registration form for the Computer Olympiad is online now

(especially also for Ewn, Amazons, Havannah).

See at the

ICGA event site.

Early registration ends on October 12, 2011.

Double fees after that.

Ingo.

I registered for Chess, EWN and Havannah. Let's hope I have enough time to actually build the programs :-)

I'm almost surely going with EWN, Havannah, and Amazons. (I guess I better register.) It should be fun. What is there to do in and around Tilburg?

I do not know Tilburg very well but here's a site describing the things to do: http://www.vvvtilburg.nl/pagina/37/places-to-go.aspx

There is an excellent train connection to the nearby city of Breda: http://www.vvvbreda.nl/en/history/161

You can find a strange situation in the villages of Baarle-Nassau/Baarle-Hertog: http://english.baarledigitaal.org/historystart.php which is reachable by bus from Tilburg. Other cities like Rotterdam, Eindhoven, 's Hertogenbosch are a bit farther away but can be reached very well by train. A bit farther away (45 mins by car, train will take a bit longer) there's Antwerp and Turnhout.

I live near by Tilburg. If you need any help/advise in planning, just let me know.

What is there to do in and around Tilburg?

What kind of things do you like?

Hello Klaashaas,

would you like to visit the Olympiad for an afternoon

or evening? We should let you know the schedule (via

LG forum) so that you really see Ewn in action.

In which of the other Olympiad games would you be interested?

Cheers, Ingo.

Yes, I would like visit the Olympiad. Is the exact schedule still to be decided? (I can't find it online). I'm also interested in Go bots (and ready to be impressed!).

There is only a tentative schedule for Go:

9x9 on 21 and 22 (Mo and Tue)

13x13 on 23 (Wed)

19x19 on 24 and 25 (Thu, Fri)

Tiebreak probably on Saturday (November 26)

No times fixed, yet.

COnference slots on Sunday - Tuesday in the mornings.

Ingo.

A first version of Gambler that plays EWN is now registering here for games. Highly experimental, taking only about 1-2 seconds per move. I'll operate manually so I hope I do not make mistakes.

It seems I'm not completely off-track with how I've implemented Gambler for this game. And I have some improvements coming up in a few days (like a proper search instead of pure minimax). Nevertheless, to be able to plan my spare time until the tournament I'm wondering how the games in Tilburg will start: With a populated (random) board like here at LG, or with an empty board where the programs will need to master the art of finding the optimal way of setting up the stones?

A second question is whether rolling the dice will be open for the opponent as well so that pondering on the opponent move (even how little sense it will make due to having to roll the dice once again for your own move after the opponent played) can be implemented a little more efficiently (which is not possible here as rolling the opponent dice is blind)?

What an evolution in Ewn bots - unbelievable.

Some want to come with complete 7-stoners on harddisc,

others want to do pondering despice the dice levels...

I try to give my (hopefully balanced) opinion:

\* Only computers in the tournament hall should be allowed. No remote machines.

\* Bots are operated manually.

\* Thinking times should be at most 10 seconds for each single move.

(I hope, we can do that constructively without formal control with

external clocks.)

\* Series "best of k" are played. My original idea was to choose k=13.

But YHW prefers an even number for k, to allow for draws (which would

give more refined results: win - draw - loss).

\* I opt for public dice rolls. (But Hanfried_c, to which I belong, will not

do pondering.)

\* Originally we had the idea to use LittleGolem for recording the moves (and

also for providing the dice rolls). But now I am not sure any longer, if

this is a practical idea. If I remember correctly it is not possible

to play LG games where both sides use the same IP address.

In any case, I will have true boards, stones, dice and "dice cups" with me

in Tilburg.

Any other comments and discussions are welcome.

Ingo.

I forgot: starting positions should be

determined adhoc at random.

I agree with most of Ingo's suggestions. I have a different preference on 2 of his points.

I think 10 secs per move is a bit fast and would make things hectic. Since I (and most of us?) will be entering our moves manually this would mean we would be spending the vast majority of our time communicating and making moves. I would like to have a little time during the game to actually sit back and see what is happening in the game.

At a very selfish level I still prefer we use LG to play our games. My interface is designed completely around LG and so anything different will require some (considerable?) effort to get going. Also, I like the idea of the games being visible here as they are being played.

P.S. I am not going to be doing any pondering either.

Gambler said: "... rolling the opponent dice is blind".

Is that true? Or maybe I misunderstand. As soon as I make my move I think I see the current position along with the current die throw for my opponent.

I definitely do not have the time to produce databases, but on a 10 sec time control pondering (knowing the opponent dice) probably gives virtually about 1-2 sec extra. And that is probably a bit easier to do. Regarding time control, 10 sec is probably a lot after move 5 or so, but not so much in the initial position. So what about 2min per game, allowing for some dynamic time allocation (and giving some incentive to move quickly on easy moves)? Own time keeping after entering the dice can probably be trusted ...

Perhaps dice are not blind in EWN here. I just assumed so as they are blind in Street Soccer.

"And that is probably a bit easier to do." What is the "that" you are referring to?

Re time control, I think there is a balance between getting the most games in and being (somewhat) relaxed while playing. Though the more games, the more accurate the result of course, but I still lean towards being more relaxed. Especially for those of us (Richard, too! :) ) playing in multiple events.

Three comments:

\* With "10 seconds per move" I had meant

"10 seconds of thinking time after entering the outcome of the dice roll"

\* Dice are not blind on LG in the sense that I can see the dice roll

for the opponent immediately after I entered my move.

\* Using LG is fine for me if the IP address problem can be resolved

in Tilburg. So, what is the situation: Does LG allow games between players

which use the same IP?

> \* With "10 seconds per move" I had meant

> "10 seconds of thinking time after entering the outcome of the dice roll"

Maybe it's not as bad as it first sounded. As implied in your comment, there will be no obligation to enter moves at breakneck speeds.

> \* Using LG is fine for me if the IP address problem can be resolved

> in Tilburg. So, what is the situation: Does LG allow games between players

> which use the same IP?

It's hard to imagine it would be an issue. Why would LG even track where the moves are coming from? When I originally tested, there were OneStone vs. lorentz games played from the same machine and there seemed to be no trouble.

If I remember well, and for some period, gamnes where 2 palyers had the same id, resulted in no change in the ratings between them. To avoid rating manipulation. I think it's not an issue but you can always check with a Lan and 2 userids.

Or just 2 userids from one box.

To me, using LG holding the tournament over LG would be convenient but certainly not necessary. The one proposed rule that I don't like is the requirement to use hardware that's local to the tournament. If the idea is to put a restriction on the allowed hardware then we should do that explicitly. But I don't want to have to either put my computer on the plane or scrounge together an entirely new one for Sybil once I land.

Also, if we do decide on using LG to run the tournament, we should have ourselves a backup plan. Though uptime's been pretty good lately, this site still isn't 100% dependable.

> Or just 2 userids from one box.

That does definitely not work. As soon as you start

some move activities, all open LG windows on that machine

are switched to the same userid. (Just tested with my

new account "MaWaldorf".)

On remote computing:

We will run in problems when allowing that - reliability

of connections is one; the danger of a hardware race another one.

When the fear is that some people can bring stronger machines

to Tilburg, we can limit the computing power for machines used

in Tilburg.

I can live with almost any limit. The machine for Hanfried will

be a single-core 2.2 GHz notebook.

On reliability of LG:

That is a crucial point. In some days/weeks LG had very slow

response times. Currently I do not see a way to guarantee fast

response times for the Tilburg week. As I wrote before:

I will have true boards, stones, dice and "dice cups" with me

in Tilburg.

So:

- 10 sec pure thinking time per move

- boards generated external to the programs

- pondering allowed

- local machine required (in that case not a fast machine for me either)

I will probably not have an automated interface to LG (very little time, still a lot to do) so manual operation with real boards and real dice is ok with me. I guess you already thought of a way to generate the boards in that case ... ?

@Ingo: This is a side-effect of cookies. You can either:

use the cumbersome emthond of logging in and off between every move, or

use 2 separate browsers for the 2 userids (just tested with Chrome and Firefox, it works), or

use separate OSes from same box.

So, just checked, you can definitely use LittleGolem for play. I'll check if you can play rated games as well from same IP.

I agree that connectivity to a remote machine could be an issue. If all that's needed, though, is the ability to SSH from one box at the tournament site to one box back home, the issue is likely to be a minor one.

Do we not trust competitors to follow the hardware guidelines we set?

About the LG question: Sybil and I play our games from the same box. LG figures out which of is is which by the cookies that we send. One can play around with this by accessing the site with two different browsers.

> so manual operation with real boards and real dice is ok with me. I guess

> you already thought of a way to generate the boards in that case ... ?

The boards have already been produced. They are in my shelves. We even

can write coordinates on them for those who need it for orientation.

At the end of the tournament each participant can take one such board home

(perhaps with the signatures of the competitors...).

Testing done, you can actually play rated games from same IP.

For rated games:

- You could set private tournaments, if you want all-vs-all tournaments. Or private tournaments with 2 users for knock-out phases.

For unrated games:

- Use invitations.

LF will not support structured best-of-EvenNumber matches. If that's what you decide you want, you'd have to do it as the combined score of two matches, unless you go even further in that direction. The extreme would be having each game separately set up.

sybil_c wrote:

> I agree that connectivity to a remote machine could be an issue. If all

> that's needed, though, is the ability to SSH from one box at the tournament

> site to one box back home, the issue is likely to be a minor one.

I have been tournament director in two computer chess world championships

(back in Mainz, 2005 and 2006) - and connectivity became a serious problem,

although before the event "everybody" said "no problem at all". If I remember

one of the pairings affected in 2005 was "Shredder vs TheBaron" where I finally

had enough and give the players a 2-minutes penalty each.

> Do we not trust competitors to follow the hardware guidelines we set?

Looking at history in other games, it was a problem from time to time.

One prominent example: the long debates after the Computer Chess World

Championships back in 1986 in Cologne: Cray Blitz played remote, and after

its win over Hitech in the last round the Hitech team complained about

human intervention.

Concerning Tilburg, participants have at least two options:

\* bringing with them a "normal" notebook

\* asking the organizers for a loan PC (you can click on that in the

application form)

> The boards have already been produced.

I should probably have said: 'I guess you already thought of a way to generate the setup in that case ... ?'

Pondering should be easily implemented in all programs. But can pondering really help to find the best program/ideas or is only the one favored with the fastest hardware?