WiR 3.3 experiences

[Sardaukar]

I started playing WiR again when 3.3 came out as Russians. That was a scary experience, even with only Even settings. AI managed to get next to Moscow and Leningrad before mud and blizzards saved me...Germans were piercing my defensive line just 5 hexes from Moscow in Sept/Oct.
I did pull all my aircrafts to rear area HQs for training, and started to use them massed in few HQs after several months of training.
I had 2 dug in defensive lines I formed in front of Moscow from rear area units and newly formed units. All of that was pierced by **** German spearheads. AI seems to be quite agressive..and Russian performance was abysmal. I gave Germans good trashing during blizzard, but now in spring 42 they are advancing again....dammit...but quite historical.
So, did I do something horribly wrong or was I just unlucky ? I tried to withdraw in front of German onslaught and fight only from entrenched positions. Should I have my air force fighting from start or was it wise to train it first ? I wonder what would have happened if I had Computer Max Help setting I would have probably lost both Leningrad and Moscow

Anyway, seems very exciting..and AI plays bit better than in previous games I've had. I know it's still Artificial Idiot..but I had excitement plenty.

Cheers,

M.S.

[Sardaukar]

One thing, BTW, what I miss in WiR is points scored by casualties like in Pacwar. But on the other hand I'm not sure if it'd suit WiR very well. What do you folks think ?

Cheers,

M.S.

[dtx]

One of the biggest failings of the AI is that it seems to view all hexes as having equivalent value, when it should see the value of Moscow & Lennigrad as crucial. Because of this, often the Russian AI mans the defenses of Moscow with just a few troops and the German AI fails to concentrate its' forces for an all out attack on key points.

Obviously, this is a tough thing to program into AI. In the original Deep Blue supercomputer chess game, one of the programmers did his entire PhD thesis on calculating the value of a knight in the center 4 squares of a chess board. In other words, this type of programming can require incredible effort - even for something like Chess that is tactically simpler than WIR (due to a less number of squares vs hexes, and less numbers of units). Helping the AI to better "see" the board is a key.

To give some extra perspective, although the new Deep Junior can only calculate 3 million moves/second, whereas the older Deep Thought could do 200 million moves/second, Deep Junior is actually a better player because Juniors algorithms help it "see" the board better.

To overcome the WIR problem, it would seem that hardcoding enhanced Russian defense when Germans are close to Moscow or other strategic city and doing the same for the Germans word strengthen the AI considerably.

[Popov]

I find it too easy against the computer

I began a game against german coputer
he advances straight on my strongests units and takes heavy afv losses. His Panzer Korps are stopped near kiev and Mogilev...


Hi didn't avoid and surround my corps with high cv value, and my T34s inflicted heavy losses to his weak Panzers

And it is only the middle of august