I spot a nice one!

[rangzen]

Did you spot a nice replay?
An interesting strategy?

I start the topic with this one : http://ai-contest.com/visualizer.php?game_id=4811861
Clash of the titans with an exchange of the side!

[RebelXT]

nice game indeed! position swaps are so rare when bots play..

Finally top level bots are starting to play like humans play Galcon.

[Evgenius]

Very interesting game, I also noticed it.

And since we are talking in Strategy forum, let's have a discussion. Do you think players were doing the right thing by waiting and saving ships? Do you think they were making the right moves when action started (around turn 56)?

As for me, I guess dmj111 could attack a 5-growth neutral planet with 90 ships in the right bottom corner (at least).
And when the action started, it's really hard to analyze anything, but let's look at how dmj111 plans an attack. There are two fronts: upper and lower. Dmj111 attacks small growth-1 planet with a large ships pack instead of neighbouring bigger planets. And then dmj111 instantly attacks all neighbors leaving them no time to plan defence. And on the upper front dmj111 concentrates power on the 4-growth planet just upper the center. After that also follows an immediate all-directions attack. I think it was a key to success.

[dmj]

Evgenius,

I agree... my bot probably should have taken that neutral in the corner. My bot is overly conservative as far as neutrals go. If I figure that out, it will be the next iteration

nice work starting the strategy discussion too!

Dave

[RebelXT]

I also tried to analyze this game.

McLeopold's bot was ahead both on production and ship count until it launched attacks on 68 and 71 planets. I'm not sure what triggered McLeopold to attack 1-growth 71 planet since payback time is 71 turns, and it was not ahead by 71+ ships at the time. This gave dmj's bot a chance to attack and come back.

McLeopold should have won this game had he not over-expanded during turns 54-59, wasting too many ships, and also letting dmj steal the 45 neutral on the right.

dmj, you should stop tweaking your bot for a week or two to let us catch up! j/k, it's nice playing you, McLeopold and other powerhouses on the TCP server.

-rebelxt

[rangzen]

Do you think they were making the right moves when action started (around turn 56)?

I sadly have to say that actually I've no portion of code to "wait", I really don't understand how the bot wait for and why
Impressive.

I guess dmj111 could attack a 5-growth neutral planet with 90 ships in the right bottom corner

Same here, I hardly imagine what kind of loop can wait for instead of taking this planet.

I don't understand also the size of some attack. They are far too heavy and I don't understand if there is a purpose or if it's a calculation not so fine … The 45 neutral is taking back with 248 ships but in the same time there is a 297 attack in the opposite direction?

Very impressive bots, guys

[DanielVonFange]

Heavy attacks allow you to immediately continue the attack onto nearby less defended planets. Currently bots don't defend well against this.

[RebelXT]

I sadly have to say that actually I've no portion of code to "wait", I really don't understand how the bot wait for and why
Impressive.

One example when it makes sense to sit tight and wait is when you are behind on total number of ships by a little bit while your total production rate is higher. What this means is that you will eventually get ahead in a number of turns, and you don't want to continue expanding by capturing neutral planets as long as enemy is not expanding.

Bots have to balance their priorities between defense, offense, and expansion. Top level bots definitely change their strategy based on various metrics such as number of ships, growth, number of turns remaining, positional advantage, etc.

It's pretty amazing how deep this AI challenge turned out to be!

[dmj]

I guess dmj111 could attack a 5-growth neutral planet with 90 ships in the right bottom corner

Same here, I hardly imagine what kind of loop can wait for instead of taking this planet.

Yeah.. my code is far too conservative when it comes to neutrals. At first, it was good, because I don't waste ships on a neutral. But it misses shots like this that should be obvious.

I don't understand also the size of some attack. They are far too heavy and I don't understand if there is a purpose or if it's a calculation not so fine … The 45 neutral is taking back with 248 ships but in the same time there is a 297 attack in the opposite direction?
Very impressive bots, guys

And here it is a bit too aggressive. The aggressive / conservative bits of logic in my bot are working out ok, but when I tweak either at home.. things get ugly.

There are a lot of impressive bots out there. I imagine the leaderboard will change a lot as more games are played.

@RebetXT -- I haven't been on tcp much because my internet connection has been intermittent (at least that is what Verizon calls it).

[McLeopold]

My bot tried to bite off more than it could chew. It was attacking the enemy and expanding at the same time, which wasn't the best thing. When the action starts, dmj moves a large fleet up north to his own planet, possibly for better positioning? The key is that dmj's fleets choose short destinations and mine choose long destinations. That gives dmj much better reaction time to take enough growth away from me and turn the tables.

My bot is now getting it's butt kicked, which I think proves fast growth is only one part of a good bot now. (And foo on anyone who thinks "number of turns code" is good strategy.)

Here is another great side swapping game.
http://ai-contest.com/visualizer.php?game_id=4796346