Well, that's interesting

I'm a little surprised Glacier didn't drop more than it did, considering it's relatively "dumb" during the 1v1 at the end of each match.

Makes sense to me that SandboxDT rose quite a number of ranks in this context. Interesting that Logic, Coriantumr, and YersiniaPestis rose notably.

I think the key is to look at the KNPPBI i.e. blue line for a bot in a regular melee, if bot is specialized on a hard bots, which seems to be in above cases, then it will be placed higher in MeleeTop30. Glacier seems to be better at the mid level bots, though it hard to judge for a top 10 bot.

The rumble doesn't allow one to see this type of information, but for melee I would say that bots specializing against other groups of bots by strength, is not as directly relevant as how much they specialize at a certain points in time during the match (i.e. when the field is packed full, moderately full, or only a couple left).

Changing to relatively small group of bots drastically changes the distribution of how many bots specialize in what portion of the match.

What I was trying to say that some bots, for example EvBot, does not have a fancy targeting system, nevertheless it is good enough to score against simple bots to be in top 30. But when it fight faces only top bots, which presumably are good in 1on1 as well, it just stand no chance and slides down in the rating. Since it is not design to fight against the top bots.

Gigarumble is great at spotting how much bots rely on prey-on-the-weak strategies. The more they rely, the more they go down. And which bots follow more defensive/generalist strategies. They go up.

Best is Diamond 1.8.28 Because Of 100% PWIN

Well, while that is partly true, there is nothing stopping a bots that follow a more general strategy from preying-on-the-weak when the opportunity presents itself. They may just recognize that few of the robots it's facing are "weak".

May be we can generalize and say that most of the bot have very weak motion algorithm, i.e. they do not dodge even a circular gun. Older version of EvBot got to the top 30 melee only using circular gun with reasonably good dodging motions.

I still remember my shock of this discovery: circular gun is good enough.

The circular gun though completely inadequate in 1on1 to get anywhere high.

The top bots authors are usually on top in both 1on1 and melee, so they can evade such simple gun, and then they still dominate top position.

So I think it is not preying on a weak, it is more of not able to hit a strong one.

In general melee the chance of meeting the top bot is quite small, so you score on others and simply yield to the best.

I'd disagree with "the movement brings you in the top-30, the gun into top-10 of Gigarumble".

I'd say targeting plays a huge role in getting into the top-30. A few of us with high ranking melee bots currently, started out developing our melee targeting using the movement from HawkOnFire. Between GlacialHawk and the current version of Glacier, I only managed to improve the movement by 1 APS worth, and well... HawkOnFire ranks 47 places below Glacier currently.

(1: Granted, HawkOnFire's movement is exceptionally well-tuned for a relatively simple low-codesize movement)

(2: Also, I'm pretty sure the 5 bots ranked higher than Glacier in the main MeleeRumble mostly outrank Glacier on the the basis of movement, rather than targeting)