Second Front first impressions (Part II)

Second Front’s five-scenario playtest build has what it takes to divert, surprise, and charm a discerning digital wargamer on the lookout for the Next Big Thing, but what happens when that discerning wargamer swaps his smoking jacket for a lab coat, and his (rose-tinted?) spectacles for a microscope? This week, with the help of the bundled map and scenario editors, I train a forensic eye on the AI in Jo Bader’s eagerly awaited Steel Panthers-like.

Before I share the results of my hexperiments, it’s important to stress that right now SF is an unreleased work-in-progress with scant documentation. I imagine one of the purposes of the playtest build is to gather information that can be used to improve the code prior to launch (whenever that my be). Hopefully, the following will prove useful to Jo and interesting to anyone who, like me, appreciates artful artificial opponents in their tactics titles.

Although the AI in all five pre-built scenarios is no push-over, complicated battlefields make serious analysis of its skills difficult. Keen to discover whether I was tussling with a genuinely cunning foe, or a lucky one who needed a numerical edge, a carefully crafted map, or advantageous deployment zones to prevail, earlier this week I built a series of simple challenges to test its abilities.

Designed to illustrate appreciation of cover ‘Conifer Corridor’ tasks a computer-orchestrated US force of Shermans and GIs with taking two victory locations defended by little but German AT guns. A human player looking at the American starting positions would, I’m sure, quickly realise that the best way to tackle the situation would be to send infantry squads (initially clustered at A and B) along the central spine of woodland, and attack the VLs from the sides or rear. By keeping to the middle of the strip of trees, the infantry with their relatively short weapon ranges (maximum 12 hexes) could approach their goals out of LoS of the two Pak 40s (D) and single ’88’ (E).

Sensible tank tactics are more problematic, especially when you bear in mind that SF’s turn structuring doesn’t allow ‘shoot ‘n’ scoot’ tactics. If the Shermans wished to support the infantry attack, they would need to expose themselves to weapons capable of bringing about their instant demise.

So how does Second Front’s AI tackle Conifer Corridor in mid-July 2022? Well, there’s good and bad news. The way it keeps its tanks huddled in the trees at point C shows its keenly aware of the danger posed by the Pak 40s and 88. That’s great (although as I’ll explain later, that caution may be too pronounced at times). Also encouraging is the way the computer quickly brings into play the one weapon – other than the main guns on the M4s – with the range to trouble the Germans from the eastern map edge – a 60mm mortar. Within a couple of turns of the start, detonating mortar bombs are showering the 88 with dirt and shrapnel. If the mortarmen get lucky and de-crew or KO their target, the Shermen, appreciating the coast is now clear, speed into the southern field and join the attack.

What disappoints me is the way the CPU uses the ten infantry squads at its disposal. Instead of pushing some or all them towards that handy belt of timber, it opts to inch them west across totally open ground. If they’re lucky the crawlers survive a few turns of German HE before losing their nerve and hurrying back the way they came. I could force the grunts to use the woods during their advance by placing them circa C in the scenario editor, but that feels like hand-holding. In other titles, weighted pathfinding routines produce arborophilia dynamically.

In Snowhere To Hide, another one of my test scenarios, the CPU’s Soviet attackers have no armour and bugger-all chance of approaching their goal – a two-storey building occupied by a German MG 42 team – unobserved. They do, however, enjoy a substantial numerical advantage. Ten leader-boosted infantry squads plus two MMGs and a heavy mortar confront that one well-appointed Axis HMG.

When I run the scrap for the first time I’m hoping to see the Ivans site and use their support weapons sensibly and I’m not disappointed. The range of their MMGs is slightly too short for the AI to leave them on the edge of the forest, so they advance a little further to the single hexes of trees at X and Y before opening fire.

The mortar team dashes to the central copse (Z) and immediately starts lobbing smoke bombs at my HMG. In SF white phosphorus rounds don’t always play ball, but on this occasion the smoke screen is perfect. Blinded by the murk, for several turns my machinegunners struggle to target the knots of scampering attackers.

For a brief moment after the smoke clears, it looks as though my tripod-mounted Maschinengewehr might get a chance to show what it can do, but when a second salvo of obscurant restores the veil, I realise it’s not going to be my day. But for some unnecessary MMG movement mid-way through the battle, the AI doesn’t put a foot wrong. Impressive!

The second time I run Snowhere To Hide, SF’s slightly strange approach to special ammo (there’s a percentage chance of running out of smoke rather than a specific number of rounds) knees the AI in the nuts in turn #2, but, with the help of parabolic HE and those MMGs, the CPU still finishes up occupying the VL.

It’s only when I remove the 82mm mortar and the MMGs from the scenario that I start witnessing Soviet tactics bordering on ‘bumbling’. While the assaulters are more than happy to charge towards my MG 42 when its pinned, blinded, or routed, when it’s fully operational they insist on crawling towards it painfully slowly. The make-like-a-snake approach might make sense in a more cover-rich environment, but on this map it’s calamitous. Long before the majority of Soviet grunts have got close enough to their goal to use their SMGs (range: 4) and rifles (range: 10), they’ve been ripped to shreds by the diabolical death dispenser on the second floor. The infamous “crawl of death” did Close Combat no favours. Fingers-crossed SF won’t ship with the same behavioural quirk.

I’ve fashioned and run several versions of Tigerphobia, the last of my test scenarios, over the past week. The maps are very different, but all of these tactical Petri dishes…

…ask the same question. In SF how do packs of AI-controlled Allied medium tanks go about neutralising heavily outnumbered Panzerkampfwagen VIs?

Right now the answer seems to be “Not very imaginatively or efficiently”. I was hoping to see troops of computer-guided Shermans converging on big cats from all directions… T-34s using their superior mobility to flank and crowd isolated heavyweights. Instead, more often than not, I witness either excessive reticence or piecemeal frontal attacks that only succeed if the dice gods are in an extremely mischievous mood.

Once they engage, enemy AFVs seem very reluctant to disengage (panicky reversing never occurs) or pop smoke, however faint their hope of perforating the Tiger. Last night I watched a Greyhound armoured car knowingly enter the killzone of a motionless Tiger, then sit there trading shots despite a minuscule kill chance.

Bizarrely, the Greyhound’s recklessness actually paid off in the end. Thanks to SF’s slightly odd interpretation of bocage (In the game, Normandy’s bermed hedges hide hulls but not turrets), and at times questionable penetration mechanics (In reality could a Greyhound ever puncture a Tiger from the front?), the canine ended up KOing the feline!

The absence of reversing… AFVs sometimes choosing inappropriate hull directions… the improbable survivability of unbuttoned commanders… crews exhibiting ridiculously gung-ho behaviours after bailing out… extremely generous movement allowances… over-modelling of duds and gun malfunctions… there are a host of small things that individually are of little concern but considered en masse alongside an armour AI that currently isn’t especially imaginative or dashing, leave me somewhat conflicted about tank vs. tank combat in Second Front v0.864.

(TO BE CONTINUED. Not next week though, as an A2Z is due.)

8 Comments

    • Hi Jo. These were stress tests rather than attempts to create credible or enjoyable scenarios. All in all, I think your AI coped with them pretty well. In Part III I plan to showcase some of my less sadistic creations!

    • Hey Jo, for what it’s worth I think most wargamers are aware of how tough it is to produce good AI and will recognize Tim’s experiments here as pushing the boundaries beyond what most games could handle without scripted AI plans or carefully designed scenarios. It’s fascinating to see (and the results reflect well on the game in my opinion) but I know a typical scenario isn’t likely to encounter situations like this. I’ve already wishlisted the game!

      I can’t think of any game off the top of my head that would handle Tigerphobia very well, for instance.

      @Tim – great article, love to see a deeper dive into the AI workings, and very much looking forward to part three!

  1. I uploaded a scenario “that’s all volks” to the Workshop where the AI is attacking.
    It puts the player in the defending role and it is very difficult to win.
    AI attacking scenarios are more difficult to design as it is more difficult and risky to attack.
    The design of a scenario is intended to be in favor of the AI not against it.

  2. Thank you Tim for putting your lab coat on and dissecting the frog for me. The AI does interesting things, but the AFV behaviour is a bit :m

    After a decade of playing games/sims with realistically scaled weapons, tanks and soldiers, I think I am not able to go back to Steel Panthers-like, impressionistic, si non é vero é ben trovato, tactical wargames. But I am also sure that those who weren’t able to wrap their heads around CMx2 or Graviteam games, find Lock and Load turn structure fiddly and counterintuitive, are bored by Valor and Victory heavy die rolling, or put off by The Troop very limited scope, will their find solace in Second Front.

  3. I think the Greyhound commander is being unfairly disparaged here.
    “You fool!” cry the accusations.
    “You’re going to die!” remonstrate onlookers.
    “You’re wasting our precious armour!” plead the sensible.

    “You.. won?” mutter the surprised crowd.

    The hatch opens, a beret gets straightened and, after a short glance, the commander jumps down from the vehicle, a hexi burner and kettle in hand.

    His driver follows, tin mugs swinging casually from his relaxed fingers. He winks at the crowd, calls out to them, “Got any biscuits?”

    • Someone explain to me why competitive tank dueling isn’t a real sport I can watch every day.

      Preferably not sponsored by RedBull but almost certainly sponsored by RedBull.

Leave a Reply