Match of the Week (3rd in October)

[876584635678890]

AntiAndi (starter-algo-V2.1) vs Isaac (dumb-dumb-base)

https://terminal.c1games.com/watch/173219 (if you do not read the whole post, just be aware that this probably does not feature a high-level strategy, but is interesting exactly because is does not)

This is first of all an interesting match up because dumb-dumb-base is currently at 1338 elo (rank 151) and falling fighting against starter-algo-V2.1, which on the other hand is at rank seven, which equals 1752 elo.

The match

The turn two shenanigans from Dumb-dumb set the tone of this match: Removing all their defenses right at the start seems like a strategy to win a match, especially against a top algo.

Dumb-bases’ variety in attacks, i.e. sending all scramblers every turn actually prevents Starter’s EMP’s from coming through. Our underdog also gave up building defenses after their highly effective move in turn two and because they are sending scramblers each and every turn and Starter cannot get away from sending EMP’s, dumb-dumb-base wins on health after a full length match.

Analysis

When we start to look at how the algos behaved, we will quickly notice that AntiAndi’s algo is only sending out number_affordable EMP’s and thus will never win out against number_affordable scramblers from Isaac. These scramblers were able to deal damage in the beginning because Starter-algo needed its time to set up some destructors and since it only ever sent out EMP’s, it could not match the scramblers.

How

Looking into other matches played by starter-algo-V2.1 (e.g. this, that or that match), we can make out that it is an extremely primitive and linear algo because it always does only one thing: Build up its fixed combination of filters and destructors in the same arrangement and order every time and then send out EMP’s every turn.

You might think that an algo like this should not land at the top of the leader board, but this works against most complex algos because they are building up structures, which get destroyed by the EMP’s and they can deal damage in the “strategic breaks” of more complex algos.

dumb-dumb-base is even more dull, i.e. instead of sending out only EMP’s it only deploys scramblers and these are the perfect counter to the high-cost ranged unit.

Ah

From this we learned that it does not take an overly complicated algo to rise to the top and also that there can be a simple counter to an otherwise working strategy, which will only counter exactly this strategy and thus fall down very fast. In fact, dumb-dumb-base has fallen over three ranks in the making of this and will probably land at the very bottom of the leader board eventually because it only counters one strategy (starter-algo-V2.1 is at the top because it counters many approaches) and will thus lose way more than it wins.

[Isaac]

Awesome awesome analysis. I really enjoyed this :).

I specifically added this algo on a whim (dumb-dumb) and named it such because it literally never makes a base, so it’s dumb, lol. I did not expect it to win any games at all other than against the starter-bot (which is where I tested it) so it is very interesting to see that this can work depending on the situation.

But again, awesome awesome post.

[C1Junaid]

I’ve never seen such BM taunting in any game or sport in my life than destroying your entire base and refusing to put down any defenses for the rest of the match against a top 20 algo.

It’s like a boxer who drops his hands and just dodges, or a jiujitsu master who just sits down in a middle of a match.

11/10 I died laughing.

[RegularRyan]

This is amazing lol, even funnier that its Isaac’s algo. Its a great way to highlight the weakness of a static algo and how effective well-timed scramblers are at defeating them

[KauffK]

@876584635678890 I love your state-of-the-leaderboard updates and spirited analysis of matches. Your enthusiasm is contagious, keep it up!

[RegularRyan]

It looks like CC has a really weak right corner that alot of algos would exploit, and a fairly poorly optimized attack of EMPs in one location. It also leaves its left side open for a while, and only really ‘blocks’ right-moving attackers with its wall. I do see how it would lose to alot of algos, amazing that it beat Mcalgo though

[AntiAndi]

I have to say, that i never thought that my algo will get so high in the leaderboard. Even if i create new ones who wins vs my algo, it loses vs the algos my old algo won against.

When i first saw the match, i never thought about it, that an algo with 0 defence could win, but that is a greate example, that every algo is different and how to counter them!

Also greate analysis

[KauffK]

I have to say, that i never thought that my algo will get so high in the leaderboard. Even if i create new ones who wins vs my algo, it loses vs the algos my old algo won against.

This is my experience as well. The relationship between many algos, even the ones that are adaptive, is very much non-transitive. That’s why ELO makes really good sense as the way to determine the winner for all the competitions.