The magic is back! (2nd try at CMSF)

[istari]

To the Battlefront team -

To set some context, I've been a member of the Combat Mission fanclub since the late 20th century :>. Back when we had forum #s, I was ~1500. I remember participating eagerly in the fierce grog debates and spending hundreds (or possibly even thousands) of hours enjoying CMBO, CMBB and CMAK. They provided some of the most memorable experiences in 25 years of tabletop and computer gaming and triggered a deep dive into WWII history that continues to this day.

I preordered CMSF last year, partly out of gratitude for all the good gaming from CMx1, partly because I had complete confidence in the Battlefront team. That's why I was totally appalled when CMSF finally arrived. I couldn't believe this was the same company! I won't revisit all the complaints and concerns (which have been rehashed here many times), but suffice it to say that I tried CMSF probably half a dozen different times and in every case, gave up after 1-2 scenarios. It just wasn't fun. It lacked that special "magic", that combination of great game mechanics and authentic simulation that just pulled you for all-nighters of "one more turn".

Since finally shelving CMSF last summer, I've popped into the boards every month or so to see how the game was progressing. To be frank, I didn't have high hopes since the problems seemed so widespread and perhaps even fundamental to the design (Real-Time emphasis over WEGO), but I hoped there was a chance that CMSF could be redeemed with enough patches.

Last week I dropped in and started reading all the praise for 1.10 and the new Marines module. Yes there had been improvements in previous patches, but something in the tone of the forum posts here seemed qualitatively different. So I gulped and paid my $25 for the Marines module last night. Yes I could have just downloaded 1.10, but I still feel a debt to Battlefront for how many hours of gaming I enjoyed from CMx1, so I figured that even if CMSF: Marines was awful, I was basically paying them for all my good times with the original games :>

Last night around 9pm, I reinstalled the game and updated with the Marines module. Set the Options and fired up "Al Hawl" as a test scenario, figuring I'd give it an hour before shelving it again.

As my bleary eyes glanced over at the clock and blinked in astonishment at the clock showing 3:30 AM, I knew that THE MAGIC IS BACK! :>

That special something that has been missing all along is finally back with this new release. It was an utterly gripping battle between my Strykers and an advancing Syrian armored company supported by BMPs with mech infantry. Even though the AI still showed signs of problems (most of the Syrian armor never moved beyond their starting zone), I still had an absolute blast. The infantry now move intelligently, making use of cover and mutual support by fireteams. The frame rate is finally smooth. The spacebar dropdown helps overcome the awkward UI tabs. Both sides open up with appropriate weapons and seem to react with reasonable actions on contact....

I could go on, but I just wanted to offer a big congratulations to the Battlefront team on this new release. I was doubtful CMx2 could ever recover the glory of CMx1, but after my marathon session last night of "just one more turn" (:>), I'm now looking forward to diving into the rest of what the game has to offer.

If you've been waiting on the fence like I was, give 1.10 a shot. There are a number of incremental advances and improvements, and there's clearly still more work to be done, but CMSF has finally recovered that indefinable "fun" factor that made the CMx1 series such a riveting wargaming experience.

Chris

[Bahger]

Very helpful endorsement, istari, many thanks. I'll be giving it another whirl, too, partly thanks to your vote of confidence.

[Elmar Bijlsma]

Pfft, judging by the story, clearly you only like CMSF due to the sleep deprivation at the time of playing. Fanboi!

If you have continued difficulty with the hotkeys, may I suggest searching for Other Means's hotkey file. It's pretty much Old Skool Combat Mission command hotkeys.

[kipanderson]

Istari, hi

I am glad you like the game… it is like having ones favourite toy back… …

I too now find that nothing breaks the immersion which is the test that counts…

BTW… now that the intrusive bugs are gone it is the ability of the scenario designers to script the AI which makes for such good human v AI games.

All good fun,

All the best,

Kip.

PS.. lots and lots of great titles and modules in line to came our way… Normandy, Arnhem, Bulge and Bagration plus more…

[Other Means]

My hotkeys file is here. Backup your default set (in the Data folder) and put this one in if you fancy giving it a whirl.

It's pretty self explanatory:

<E>F // move fast

<E>Q // move quick

<E>M // move normal

<E>S // move slow

<E>H // move hunt

<E>A // move assault

<E>R // move reverse

<E>B // blast

<E>N // mark mines

<E>T // target

<E>Y // target light

<E>V // target arc

<E> // NOT CURRENTLY USED... Do not assign a key to this function! It will eventually be the hotkey for Clear Target, but not yet properly activated

<E>O // face

<E>D // deploy

<E>P // pause

<E>U // hide

<E>L // dismount

<E>B // bail out

<E>- // vehicle open up

<E>I // pop smoke

<E>G // acquire

[Huntarr]

Glad to hear it, ishtari.

Just because we're all holding hands doesn't mean we'll be taking warm showers together until the wee hours of the morning.

[Sivodsi]

Your post is an interesting contrast to Grumbling Grognard's post. http://www.battlefront.com/community/showthread.php?t=84287 Are you sure you guys aren't working together?

[istari]

Hi all -

Glad the post found some resonance. I spent much of yesterday immersing for the first time in CMSF, and I can safely say that it's holding up well to the first impressions. I'm having a blast, and while scenario quality varies (as with CMx1), the best scenarios are every bit as memorable and nail-biting as the original.

There are a number of small niggling items, but agree with kip's comment that none of them are immersion-killers (unlike CMSF 1.0!). I'm also coming to appreciate some of the ways in which CMx2 has really taken a quantam leap beyond the limitations of CMx1.

Bahger - hope you have a similarly positive experience!

Elmar - yes, I am sleep-deprived and probably not thinking with total clarity after my gaming marathon, but the game is worth it. :>

OtherMeans - thanks for the Hotkey list! Will check it out, since that remains one of the major irritants in the UI.

Sivodsi - haven't read Grumbling Grognard's post yet. Have to finish my workday first, then I'll check it out. Cheers!

Chris

[istari]

Actually, Grumbling Grognard's post wasn't that long, so I can respond quickly.

My experience to date has been slightly different than his. Clearly there are AI problems. The vast majority of the Syrian armored force in Al Hawl just sat there or diffidently creeped forward without any real urgency. So that critique is fair. However, I did run into Syrian infantry within the town that were pushing ahead, so at least some of the troops were advancing. The AI wasn't entirely brain-dead.

Just a few more off the top of my head...

In USMC Bad Moon Rising, the T-90s are clearly moving forwards across the terrain and engaging my M1s.

In Allah's Fist, the AI does seem to just appear and then sit there in the reinforcement zones. This is definitely disappointing, as they could have really mauled me had they aggressively moved to my flanks.

In Ambush Tutorial, the Syrian uncon troops are clearly pushing ahead with urgency and drive, forcing a really tough battle.

Etc etc. Overall, the AI seems pretty passive in most of the battles I've fought. Not totally brain-dead, but not putting me under much pressure. This is disappointing and I hope that future scenarios I try will have better AI "plans". Some of my best memories in Steel Beasts were conducting desperate defenses under terrific Red pressure, and I'd love to duplicate that in CMSF.

I wouldn't go so far as to call it broken, and I'm still having alot of fun (the ultimate test! :>), but I can see where the game might peter out much sooner for me if the AI isn't posing a challenge. Yes, I can play RT against other humans, but my gaming schedule is pretty erratic and I really enjoy the pace of SP WEGO, so I'm hopeful that the AI will improve as scenario designers become more proficient with CMx2 and as the series itself evolves towards Normandy, Bagration et al.

My 0.02,

Chris

[Cpl Steiner]

The AI definitely needs some work. If we analyse it rationally, how does the AI work and what is it capable of?

At a tactical level, AI-controlled units will make some basic self-preservation decisions, just like the player's own troops, such as moving away from enemy fire towards nearby cover. In effect, this is the same for all units, player controlled or AI controlled.

At a strategic level, the AI doesn't really exist, in the sense that it just reads the scenario designer's script of where it is supposed to move to, with what units, and when. It takes no account of what the player is doing whatsoever, as far as I can tell.

What the Strat-AI really needs is some way of reacting to the player mid-game, in the same way that a chess computer reacts to its human opponent's moves. One way this could be done is with "triggers" - i.e. points on the map which if crossed by the player's units trigger some reaction from the AI. In fairness to BFC, they have said that this is what they would like to add in some future iteration of the game engine. As of yet, however, no timeframe has been put forward for when this might happen. Let's hope it gets done for CMx2: Normandy.

[Redwolf]

The AI definitely needs some work. If we analyse it rationally, how does the AI work and what is it capable of?

At a tactical level, AI-controlled units will make some basic self-preservation decisions, just like the player's own troops, such as moving away from enemy fire towards nearby cover. In effect, this is the same for all units, player controlled or AI controlled.

At a strategic level, the AI doesn't really exist, in the sense that it just reads the scenario designer's script of where it is supposed to move to, with what units, and when. It takes no account of what the player is doing whatsoever, as far as I can tell.

What the Strat-AI really needs is some way of reacting to the player mid-game, in the same way that a chess computer reacts to its human opponent's moves. One way this could be done is with "triggers" - i.e. points on the map which if crossed by the player's units trigger some reaction from the AI. In fairness to BFC, they have said that this is what they would like to add in some future iteration of the game engine. As of yet, however, no timeframe has been put forward for when this might happen. Let's hope it gets done for CMx2: Normandy.

That opens the question why the CMx1 AI was "better".

I choose "better", because it was 1) autonomous, didn't need the scenario designer to pre-plot and 2) even with pre-plotted moves (or because of them?) CMx2 is still less realistic and less challenging.

[Cpl Steiner]

That opens the question why the CMx1 AI was "better".

I choose "better", because it was 1) autonomous, didn't need the scenario designer to pre-plot and 2) even with pre-plotted moves (or because of them?) CMx2 is still less realistic and less challenging.

This has been gone over time and time again - the old debate about CMx1 AI vs. CMx2 AI. Well, I remember when I first got CM, I played nothing but random QBs against the AI for several months, simply because at that time I just wanted a quick tactical fix without having to read through lots of briefing text, so I had plenty of experience of the AI.

The AI for CMx1 didn't need AI plans to do something. However, to conclude from this that it was better than CMx2 seems a bit of a leap of the imagination. I'm sure BFC could have used the exact same system for CMx2 and saved themselves a lot of work if this was the case but instead they chose to use the AI plans to make the AI behave MORE intelligently. This was because CMx1 AI was hopeless at attacking. It would just charge blindly towards the flags no matter what, to the extent that scenario designers started chaining flags across the map to guide the AI to where they wanted it to be.

In effect, BFC have copied this approach by allowing the scenario designer to add "invisible" flags that the AI will move towards - in the form of AI orders. The system has even been enhanced by the addition of timings of when the AI will move. CMx2 AI may still be pretty poor at attacking but you can't blame BFC for trying.

[hoolaman]

I agree.

The CMSF system is not perfect, but all the talk about it depending on the scenario designer is still as true as it ever was. Unfortunately the CMSF release scenarios were built when the people building them did not have quite the understanding of the editor that they do now, and those old scenarios do show badly by todays standards.

The times I have been surprised and beaten by a CMSF scenario are when the AI comes out much harder than you expect from a computer AI. CMx1 and most other games have AIs that are fairly passive and seem to keep a middle ground without extremes of manouvre, like a one size fits all kind of system must do.

Some CMSF scenarios I've played have a strong flanking move or quickly running up a strong unit to challenge your forces, things that would NEVER happen without a human hand on the AI forces. These moves may not be perfectly in synch to your moves, but the unexpected nature of it has more in common with playing a human player than a crappy generic AI.

[Battlefront.com]

We could have written a "generic" AI for CMx2 that could act on its own, but it would have taken more time than we had to invest. The existing AI code was not reusable, in part because the CMx1 game environment was more simplistic and far easier to code AI to handle. For us it was like the difference between writing AI for checkers vs. chess. Same basic concepts, but entirely different needs for each.

The most simplistic part of CMx1, from the AI's standpoint, is the whole issue of "objectives" and how they interact with terrain. In CMx1 the AI either had to retain or take Flags, that's all. Not only that, but those flags were positioned in a large 20x20m grid made for a lot fewer terrain combos and types. The less the AI has to consider, the easier it is to make.

In CMx2 there is no such nice, clean, and neat set of parameters to base the AI's behavior off of. There are different types of Objectives which require entirely different operational plans. The Objectives are even weighted differently, which means the AI has to be capable of assessing each Objective within context, including things like terrain considerations, probable position of enemy units, etc. That's a lot of stuff to mess around with for the AI to even come off as well as it did in CMx1 (which, by wargame standards, was great. By Human standards it was poor).

Given a choice between a HUGE investment in an AI that likely wouldn't do very well, then present us with diminishing returns from then on, and a system that was capable of being directed by a Human mind... it was a "no brainer"

We will enhance the AI over time, though I can not say how much and how soon. AI always has been, and always will be, the least best place for a wargame maker to invest in. It will never be accepted as anything more than "passable" by the customer, yet to get even to that level a lot of effort has to be put into it. Effort that isn't put into things that customers will appreciate if present or very much miss if absent.

Scripting is the way to go, overall, but that doesn't mean way it is scripted will remain unimproved over time. It will be.

Steve

[Battlefront.com]

As for the original scenario's AI... that's definitely the problem of our poor testers having to learn an ever changing game system under deadline conditions. There simply wasn't enough time for them to really get the hang of it before we had to finalize stuff. Some found that even after spending time doing it that they didn't have the knack for it. Compare what people have been able to do since then to what was done before and there is a big difference. Even when comparing the early and recent work of the same scenario designer!

Steve

[Mishga]

I look back over some of my early scenarios and recoil in horror at what I see.

[istari]

Question about Defensive AI in CMx2:

Is there capability for broader AI plans when defending?

I ask because I just completed "USMC Bad Day for Allah", where you're tasked with assaulting a town held by a Moktada Al-Sadry style militia and take the Madrassa.

Great scenario and tense nail-biting gameplay until I finally broke into the main town and occupied the Madrassa. The rest of the AI didn't react at all. They still stayed in their original locations scattered around the city and I had to hunt them down one-by-one. That was a big surprise and broke the immersion for me - if the Madrassa is the religious center, the symbolic heart of the town, for the AI to not react at all and just sit there to be ground down piecemeal... very dismaying.

I was proud of my break in from an unexpected angle and thought I'd caught them by surprise before they could reposition to plug the gap, but now I'm thinking that I don't have to worry about counterattacks in CMx2. I realized that the scenario appears to be just going in and dismantling a static defense piece-by-piece. No surprise counterattacks, no sudden shift of forces.

That's a big shift from CMx1, where I was overwhelmed more than once by the AI rushing back to defend the flags. Yes it was predictable, but it was still difficult to counteract sometimes when you'd been bled dry trying to GET to the flags in the first place :>.

Is this just a quirk of this scenario, or is it true that there aren't really any AI plans beyond the local TacAI of each individual unit when AI is defending?

Thanks!

Chris

[Dakuman]

I think if "triggers" of some sort are introduced to the AI plans it would make for a much more interesting game against the computer. Such as if a certain casualty threshold is reached units could fall back or if an objective falls the AI could mount a counterattack.

Any chance of seeing something like this for the WWII Normandy game?

[MarkEzra]

Question about Defensive AI in CMx2:

Is there capability for broader AI plans when defending?

I ask because I just completed "USMC Bad Day for Allah", where you're tasked with assaulting a town held by a Moktada Al-Sadry style militia and take the Madrassa.

Great scenario and tense nail-biting gameplay until I finally broke into the main town and occupied the Madrassa. The rest of the AI didn't react at all. They still stayed in their original locations scattered around the city and I had to hunt them down one-by-one. That was a big surprise and broke the immersion for me - if the Madrassa is the religious center, the symbolic heart of the town, for the AI to not react at all and just sit there to be ground down piecemeal... very dismaying.

I was proud of my break in from an unexpected angle and thought I'd caught them by surprise before they could reposition to plug the gap, but now I'm thinking that I don't have to worry about counterattacks in CMx2. I realized that the scenario appears to be just going in and dismantling a static defense piece-by-piece. No surprise counterattacks, no sudden shift of forces.

That's a big shift from CMx1, where I was overwhelmed more than once by the AI rushing back to defend the flags. Yes it was predictable, but it was still difficult to counteract sometimes when you'd been bled dry trying to GET to the flags in the first place :>.

Is this just a quirk of this scenario, or is it true that there aren't really any AI plans beyond the local TacAI of each individual unit when AI is defending?

Thanks!

Chris

SPOILER INFO

*

*

*

*

*

*

Actually a quirk of the scen. The Madrassa, itself, is but one specific objective. The Syrian Units are holding areas of the town, surrounding hills and the road net. All are valuable to them and to the Marines who are tasked with clearing the entire area. Additionally: Syrian units need inflict 25% casualties to receive 200 pts while Marines must reach 80% Syrian casualties to get 100. The values of the objectives are only 50 pts. The reason for all that is the Marine Player needs to search and destroy the enemy. The Syrians must simply hold ground and avoid catastrophic casualties. There are some reinforcements with pathing orders but those orders do not include the town. During play testing I quickly saw the hopelessness of reinforcements attempting to re-take the Madrassa. The Overwhelming fire power of the Marines just didn't make that very much fun. So I gave them some blocking positions.

One final note: There are 5 different attack and defend plans. You will not always meet the same forces and threats at the same spot. Something TOTALLY unavailable in the old game editor. BFC has stated an interest in adding "event triggers" at some point. With a few simple triggers added to the multiple plans, CMSF's AI will become a monster for the player to contend with.

[istari]

Mark -

Thanks for the detailed reply. Great to hear from the scenario designer himself! :> Props to you for an excellent scenario. Along with "Al Hawl", it's the best I've played among the dozen or more I've tried so far.

*** SPOILERS ****

I ultimately attacked across the fields to the L, overran the trench on that side and broke into the main compound from there. Now that the scenario's over and I can see all the technicals and RPG waiting for me I had come up the main road... nice defensive plan!

So you intended the scenario to be a die in place mission, ala Fallujah in 2004? Bleed the Marines white and claim victory? That makes sense.

I'm curious about the reinforcements that arrived - I saw some technicals along the road to the N, but they seemed to just kind of mill around aimlessly once they'd appeared. Does the new CMSF AI allow you as a designer to thrust them towards an objective or otherwise direct their actions once they arrive? I ask because I saw similar behavior in Allah's Fist, where the reinforcing AI just seemed to sit in the deployment zones and get picked apart.

Final question for you... I agree that with the addition of triggers, the AI could become a real monster. That's a good thing, and I can't wait! However, in the current version of CMSF, is there any way for a defender to respond dynamically to an attacker? Or will all my future missions in the current CMSF/CMSF Marines end up being static defenders being dismantled piece-by-piece? It makes sense given this particular scenario, but I'm concerned if that's the general experience of all attack scenarios in the future.

Thanks

Chris

[Battlefront.com]

Mishga,

I look back over some of my early scenarios and recoil in horror at what I see.

Heh

istari,

However, in the current version of CMSF, is there any way for a defender to respond dynamically to an attacker?

Unfortunately, no. That's what Triggers do and exactly why we need to add them. The thing is that the attacker shouldn't know enough about the defender's positions, units, and their own intent to have any assumptions about what the AI may, or may not, do. The problem with CMx1's AI was that you knew the defender would hang onto the flags, though not exactly how they would go about doing it. So on the whole I'm not sure which one offers up the best fight when averaging 1000 games. I suspect the scripted CMx2 system simply because the designer can script some really fun stuff in there.

In one Campaign scenario I was advancing on the enemy's main line of resistance. I had cleared my flanks, but I hadn't left them guarded. The AI moved a platoon of T-62s through my left flank gap and to my rear. I lost a couple of AAVs before I realized the fire was coming from behind and dispatched my Abrams to take care of them. They did, with one immobilized well out of the fight (grrr). I also had a Mk19 mounted Humvee that managed to get around to the rear of one T-62 and brewed it up. This sort of thing would probably never have happened in CMx1.

All in all I think CMx2's AI is slightly better than CMx1. By that I mean CMx1's AI was generally decent, but often was quite bad. After a while it was fairly predictable (which is the main failing of "dynamic" AIs). In CMx2 the AI ranges from really excellent all the way down to dismal. One thing it isn't is predictable.

Anyway, as I said there is room for improving CMx2's AI and so we shall.

Steve

[helm123]

In one Campaign scenario I was advancing on the enemy's main line of resistance. I had cleared my flanks, but I hadn't left them guarded. The AI moved a platoon of T-62s through my left flank gap and to my rear. I lost a couple of AAVs before I realized the fire was coming from behind and dispatched my Abrams to take care of them. They did, with one immobilized well out of the fight (grrr). I also had a Mk19 mounted Humvee that managed to get around to the rear of one T-62 and brewed it up. This sort of thing would probably never have happened in CMx1.

Steve

The scenario designer that designed a move through your lines and into the rear of your position all while you were on the advance has a future in scenario design as well as fortune telling.

[istari]

Steve -

Appreciate the honesty of you just coming out and say, "no, we don't do that right now." I run a medical software company and we try to take the same approach of blunt honesty with customers, but I understand the temptation to avoid saying "no" and upsetting folks. :>

So if I'm understanding correctly, AI plans can be set for either offensive or defensive scenarios. It just requires that the scenario designer anticipate where the player will be attacking (as in the T-62 platoon in your example above).

That's a relief, as I was concerned that all attacking missions would be against a static foe, and only when the AI was attacking would plans come into play. Now at least I'll need to keep an eye out for counterattacks, flanking moves etc as long as the scenario designer hard-coded them.

Agree that nirvana will be AI plans + triggers. Falling back after X% casualties, counterattacking with reserves when Point Y is taken, etc... I'm drooling already. I know you won't be giving any details yet about CMx2 Normandy, but here's one vote to place this feature high in the priority list. The Germans in particular need the ability to execute triggers in defending the bocage.

Chris

[MarkEzra]

Mark -

I'm curious about the reinforcements that arrived - I saw some technicals along the road to the N, but they seemed to just kind of mill around aimlessly once they'd appeared. Does the new CMSF AI allow you as a designer to thrust them towards an objective or otherwise direct their actions once they arrive? I ask because I saw similar behavior in Allah's Fist, where the reinforcing AI just seemed to sit in the deployment zones and get picked apart.

Final question for you... I agree that with the addition of triggers, the AI could become a real monster. That's a good thing, and I can't wait! However, in the current version of CMSF, is there any way for a defender to respond dynamically to an attacker? Or will all my future missions in the current CMSF/CMSF Marines end up being static defenders being dismantled piece-by-piece? It makes sense given this particular scenario, but I'm concerned if that's the general experience of all attack scenarios in the future.

Thanks

Chris

There are 5 different plans. Some have setup positions very near their pathing objective. Others will be actually harder for the human player to spot and have been timed to act as counter attacking units (South hill...just not in the town) The UnCon vehicles become cautious in the face the Marines which is why I gave them shorter routes and their own separate pathing orders.

The response of AI defenders is dependent on the designers use of movement orders and reinforcement orders. Troops under AI control will intelligently move to better cover when attacked or attempt to improve targeting positions but will not move like zombies to a lost objectives ala CMx1....yes, there is another way of viewing that particular action found in CMx1 and Steel panther style games. The idea that a well placed armor unit in overwatch positions decides to run down to the narrow streets of a village and engage in close combat is not such a swell feature, either. I look forward to the "event trigger" to create a far better controlled AI counter thrust.

What I do not want to see is the type of event trigger that are found in FPS... You show up someplace and the Tiger tank always appears. I believe BFC understands that Scen Flexibility is the key to re playability and will continue to evolve the scen editor to enrich the gaming experience for us all.

Speaking of that...Why not try The Orchard Drive scen. @ www.CMMODS.com It's a meeting engagement. You might want to read the designer tip on page 2! The Syrians give as good as they get...it's a personal favorite or get a bit nostalgic and go for "Chance Encounter" or "Last Defense"

[thelmia]

I'd like to say thanks for the good patch, and thanks to Mark Ezra for all the enjoyable scenarios in the Marine module. I've been unpleasantly surprised in many of them, and sometimes it's like fighting a human opponent.