Matrix Games Forums

Has anyone ever noticed how the system rolls some incredible dice rolls against you on more than a few occasions?

Case in point. I'm currently playing Northern Fury #14. Played over two hours so far. Fired 188 total AAMs at the enemy, only 79 enemy aircraft shot down. Now here's my gripe about this: I look at the dice values used and I tell you, this is baloney. Of the 109 misses, 41 were spoofed with each at a 20%. The rest were misses due to non-spoofing parameters. Even when I had high percentage shots of 80%, darned if I didn't suffer a natural miss. I saw a LOT of dice values of 95+%.

Cheating AI? I think so.

To give a perspective:

In a span of less than 75 seconds, I had SIX consecutive missiles spoofed by chaff. Rolls of 20, 7, 3, 14, 9, and 16. This is not including those that were jammed at a less frequent rate.

What's up with this?

It's frustrating, yes, but most likely this is just a case of observation bias. You'll find that if you average the results across many scenarios, the die rolls are all balanced fine. AFAIK, there is nothing in CMO that would rig the randomness against the human player.

The other point is play in the editor and see what's happening on the other side for die rolls.

Sorry about my ranting, it was quite infuriating for a moment. Or six.

I'll chalk this up as "War Ain't Fair, Dude" and move on.

I have played probably hundreds of wargames since online forums were a thing. I can't think of one where someone at some point didn't come on an either question random number effectivity or AI cheating on randomness.

Every discussion about RNGs should IMHO reference this: https://unherd.com/2021/07/what-warhamm ... bout-life/

The felt difference between missing one time in every five engagements and a 80% chance of hitting is incredible. Human brains are extremely bad at statistics.

I'm 90% sure you're right and 20% sure you are wrong.

thewood1 wrote: Mon Jan 29, 2024 10:47 am The other point is play in the editor and see what's happening on the other side for die rolls.

Just another observation, but in the same set of scenarios (Fury) I rarely see NATO "spoof" any missiles, while PACT seems to exceed their 15% rolls constantly. Who knows.

I mean, here's the thing: In theory you can just test this, since the game gives you the hit percentage. Take all the missiles fired on both sides and tally up the numbers.

maverick3320 wrote: Thu Feb 01, 2024 12:38 am

thewood1 wrote: Mon Jan 29, 2024 10:47 am The other point is play in the editor and see what's happening on the other side for die rolls.

Just another observation, but in the same set of scenarios (Fury) I rarely see NATO "spoof" any missiles, while PACT seems to exceed their 15% rolls constantly. Who knows.

Exactly, who knows. I mean, without showing specific examples in their full context (ie. a pre-engagement save file) anyone can claim anything.

RNGs are interesting. I had a hand held craps game to learn how to play and studies the probabilities and got good at it. I then went to Reno and played using real money and real dice and the flow and results are far different that what I experience using the microchip version using 0s and 1s.

I guess same can be said for trading platforms like FOREX, options and stock trading programs that lets newbies trade using paper accounts to get a feel for how things work, but once real money is on the line its a much different ball game.

Perhaps an option to have dice rolls displayed like other games notably games like many board based games ported to the computer would be something to consider. Some games allow the player to tilt the dice rolls too in the options menu.

I have to say that I still believe that there is something dodgy about the combat algorithm. Maybe not deliberate bias but perhaps a tendency to roll streams of similar numbers. In my current playthrough of Hormuz Hoedown. I fired a dozen Patriots at a group of Floggers at 20nm range (before a later upload of v1.06.1328.14 made the weapon completely useless - see AAR). Two were spoofed and the other ten all missed on 46-76% chances, which is simply outlandish. I then replayed it and scored four hits with four shots.

Also, in the same playthrough, I've lost eight aircraft and every single one generated a Downed Pilot on what I think is a 60% chance (correct me if I'm wrong). Again, highly unlikely.

This sort of thing happens to me a lot. Unfortunately, I can't supply logs or screenshots because, for some reason, my laptop does not seem capable of taking such from CMO.

I have a theory that luck is a force of nature and different people are given different amounts of it. In general, a person's luck is often (but by no means always) in inverse proportion to their ability (I've often been lucky against opponents who are brighter than I am but tend not to be in the reverse situation - and the AI in CMO is pretty dumb). So it could just be me, rather than a problem with the game.

RNG getting stuck is not unheard of.

But too many players jump into a game and claim AI bias with no thought that it might be them. They especially make no effort to bolster their claims.

fitzpatv wrote: Wed Feb 07, 2024 5:31 pm and the AI in CMO is pretty dumb

Dimitris wrote: Wed Feb 07, 2024 7:51 pm

fitzpatv wrote: Wed Feb 07, 2024 5:31 pm and the AI in CMO is pretty dumb

I mean, compared to a human player, it really is dumb, regardless of how complex the AI is from a development perspective. Just a curated list of things I regularly see it do that well-informed humans wouldn't:
* decide to turn around and refuel when on RTB even though there is plenty of fuel to return all the way to base (resulting in being shot down by chasing fighters)
* chase a fighter for as long as possible, then just turn around to RTB and ignore the fighter (resulting in being easily shot down by said fighter when it too turns around)
* make poor weapon allocation choices (e.g. firing SM-3 at an SRBM in the boost phase, losing the track once the missile climbs, and thus wasting the interceptors) (WRA is not granular enough to override this)
* make awful evasive maneuvers (e.g. the ship torpedo evasion logic just swims in circles at flank speed and accomplishes close to nothing while maximizing the amount of noise emitted)

And that's excluding the many buggy behaviors that are created and patched with each release (currently: manual refueling orders being ignored while breaking all other behaviors; previously: CAP patrols attacking AAMs 800nmi away; strike escorts flying in the wrong direction; the list goes on)

I don't want to downplay the work by the devs, but the AI ultimately suffers from a lack of contextual awareness that makes statements like "the AI is pretty dumb" ring true. The intent is not to criticize the game but rather to make a point about the limitations of a computer player in such a complex simulation.

Totally agree with what caelunshun says here and have experienced most of the behaviours mentioned. It's worse when the game imposes them on the player and stops them correcting it. It all emphasises that developer effort should focus on fixing the many bugs with the game at the moment, instead of aiming to introduce even more new features.

Incidentally, was the Mad Hatter supposed to be the AI or myself?!

caelunshun wrote: Wed Feb 07, 2024 9:05 pm I mean, compared to a human player, it really is dumb, regardless of how complex the AI is from a development perspective. Just a curated list of things I regularly see it do that well-informed humans wouldn't:
* decide to turn around and refuel when on RTB even though there is plenty of fuel to return all the way to base (resulting in being shot down by chasing fighters)
* chase a fighter for as long as possible, then just turn around to RTB and ignore the fighter (resulting in being easily shot down by said fighter when it too turns around)
* make poor weapon allocation choices (e.g. firing SM-3 at an SRBM in the boost phase, losing the track once the missile climbs, and thus wasting the interceptors) (WRA is not granular enough to override this)
* make awful evasive maneuvers (e.g. the ship torpedo evasion logic just swims in circles at flank speed and accomplishes close to nothing while maximizing the amount of noise emitted)

And that's excluding the many buggy behaviors that are created and patched with each release (currently: manual refueling orders being ignored while breaking all other behaviors; previously: CAP patrols attacking AAMs 800nmi away; strike escorts flying in the wrong direction; the list goes on)

Are all of those reported on Tech Support? Also, pretty sure most of these have been fixed?

I don't want to downplay the work by the devs, but the AI ultimately suffers from a lack of contextual awareness that makes statements like "the AI is pretty dumb" ring true. The intent is not to criticize the game but rather to make a point about the limitations of a computer player in such a complex simulation.

So Command's AI is still outsmarted by humans. Okay, I can live with that

fitzpatv wrote: Thu Feb 08, 2024 12:55 pm Totally agree with what caelunshun says here and have experienced most of the behaviours mentioned. It's worse when the game imposes them on the player and stops them correcting it. It all emphasises that developer effort should focus on fixing the many bugs with the game at the moment, instead of aiming to introduce even more new features.

This is a common "looking from outside" view, and one that the creator of Rails answers better than myself:

The only reliable, widely used way to ensure impeccable software quality is to write less software that does less stuff, and then spend eons honing that tiny lot. Such an approach, however, is very rarely compatible with commercial success or even programmer motivations (despite what many may claim).

How do you think the market would receive [the next iPhone], if its headline improvement was cutting 1/3 of the features to shrink the code base so it’d have fewer bugs? Yeah, I thought so. While people may get excited in concept by “stop the train, we need to fix the tracks” directives for software development, it’s not what they would buy.