Post news RSS LOUD AI - Dec 2017 - Notes from Sprouto

A mid-December 2017 update on the development of the LOUD AI Project for Supreme Commander: Forged Alliance. Comments on setting up a competitive game with the LOUD AI.

Posted by on

December 2017

Lots of unit updates this month. A very long list of tweaks, corrections and changes do a number of things for the game, particularily in the water, where we did a full audit of torpedoes and
torpedo defense systems. The result is a more satisfactory water game and one that isn't quite as hard on performance. While the magnitude of the torpedo game is down, it's still very
busy, so we're still watching and perhaps thinking of cutting the overall number of torpedoes in the water at any given time even further than we already have.

A review of the transports was also a big part of this pass and they too, in general, behave in a much more reasonable fashion. Just in short, they're marginally slower, more expensive, a
trifle less robust, and don't pack quite the punch against ground targets. While not a popular change, this puts them in better balance in their tier, especially since we saw players opting
for transports over gunships. By contrast, T2 gunships got a very minor perk-up.

A whole bunch of other general minor fixes are included this month -- refuel and transport pool behaviors for ALL aircraft should improve the self-preservation of the AI aircraft. The starting
bomber, fighter and gunship platoons have had their lower start limits decreased to one, the idea being that almost all aircraft should be in the air, doing something, at all times, regardless
of which base they are at. This should apply to reinforcement platoons as well.

The distribution of engineers, and importantly Sub-Commanders, has been revised and LOUD will now aggressively distribute his SCU to other bases and DPs when available, so that
important works (like anti-nukes, Experimentals, etc) are being reliably performed at all his bases.

There will be another UNIT pass in the near future, with some minor balance and performance tweaks, and we are looking forward to some new structure upgrade options for T2 Airpads, T2
TMD and T2 Torpedo launchers, to make them more useful and combat worthy. We're always looking for comments on those upgrades we've made available to the standard units
(factories, T3 SAMS, etc.) so please let us know your ideas.

Good progress on the Naval changes. Exciting work and new ideas and, as always, it's nice to see things come together. Perhaps the biggest change is one that isn't so visible, but really
shifts gears for LOUD. Focus. Until recently, when LOUD was putting up new bases and DPs, he was not staying focused on his attack plan goal, that is, moving his forward position ahead
in order to beat you. Now that he has two independent primary bases (on water maps), it became clear that LOUD should only put up new DPs if they were better located than his existing
one. This means that LOUD isn't wasting time and effort putting up listening posts in far-flung places on the map where he has little interest (like behind him). It also means that when
LOUD AIs are teamed up, they don't get in each others way so much. In general, they concentrate more effectively, and with the addition of water DPs, the amphibious attack takes on a
dangerous new angle given that it can often be staged a lot closer to the goal.

It's become clear that the Naval DP has really moved LOUD forward, and of course, it has also clearly revealed some shortcomings. As a result, his build priorities have been tweaked, some
factional capabilities have been accounted for, platoon compositions and sizes have been adjusted, and quite importantly, LOUD can now effectively produce Sea Experimentals, not only at
his Naval bases, but at his Naval DP's.

A secondary topic these last couple of weeks has been scaling. Several of our regulars like to make use of the Resource Rich mod. This mod greatly increases the overall mass and energy
output of the game. LOUD doesn't quite understand these huge bonuses and in a nutshell, he cannot make effective use of it - mostly because he has some upper limits on how many
engineers he can make use of. The answer - scaling those limits. This is still new but the results are good - so far, as LOUD will respond more dynamically to resource, map size and unit
cap conditions, not only increasing the number of engineers, but the number of simultaneous upgrades he'll do, and decreasing the period between new upgrades.

On the subject of 'getting' in each others way, one of the most frequent questions we see is;

"What is the best way to set up a game with LOUD AI ?"

Well, we study so many simulations with LOUD that there are several answers.

For performance, and a game that usually has a shorter timeframe, use less AIs but with a slightly higher cheat. Our top end players routinely play 4 or 5 against 2 AIs on 40k maps, with a
cheat level around 1.25. LOUD's ability to quickly populate a map combined with a massive economy allow him to explode (since this configuration gives him a substantial increase in his
unit cap versus the humans) and the focus he can have on one player is often overwhelming. We do note that using 3 vs. 1 AI is a good trainer for building teamwork, which is an essential
element in defeating LOUD.

For longer games, that are not quite so devastating, equal numbers of humans versus the AI will give a good result, unit cap management often becoming an issue for a LOUD AI. There is
a small performance hit for having so many AI's but that's greatly impacted by the slowest machine in the game.

Just some global guidelines - Map size makes little difference on performance. The number of units is everything. Mass heavy maps don't make for a faster game but they do allow the
game to get crowded very quickly.

Post a comment
Sign in or join with:

Only registered members can share their thoughts. So come on! Join the community today (totally free - or sign in with your social account on the right) and join in the conversation.