v1.23.08 Playtest - 1812 Scenario - Part II

[pzgndr]

About a week ago I wrapped up the 1812 playtest. I finished that 1812 game (AI Great Britain won; the victory screen does work btw...) and restarted 1812 to verify changes are good. I found some issues and worked them out. It took me a bit longer than I expected. In no particular order, here's what I was dealing with:

1. French Imperial Guard Corps was still not behaving as I expected. I did a deep-dive on the special units code and scrutinized how guard and artillery is assigned. Performance is much improved now.

2. I didn't provide many details in my previous playtest comments, but was finding the French overpowering Great Britain in Spain/Portugal. The problem was once Spain surrenders and forced access ends, any French corps in Portugal end up being stuck there for the rest of the game. The AI should start retreating for home if in or beyond a forced access country after a surrender. This required a new function and some playtesting. This is working now.

3. The French offensive into Russia was being watered down with spurious off-axis attacks on Russian garrisons. Like, Riga, and then Novgorod/St. Petersburg. And Kiev, and then Odessa. So I duplicated the assign units code to make a first pass for minor units to look at lone garrisons and depots first, and then the second pass would look at all units for all attacks as before. Performance is much improved now, with French regulars more focused on Moscow.

All in all, the 1812 campaign looks pretty good now and I am content to move on to 1805 to verify changes are good for the early game. Below follows my latest playtest today as Austria again, observing the other AI computer opponents on Hard difficulty.

[pzgndr]

June 1812, Russia moves first and retreats. France advances to besiege Riga, Minsk, Vilna, Grodno, and Brest-Litovsk. No surprises here.

[pzgndr]

June 1812. Spain surrendered on first turn, which provided an early opportunity to see France retreating from a forced access nation. Great Britain is still puny and not yet ready to launch any offensives. For AI behavior, this is conservative and I'd rather have the computer opponent err on the side of caution than take unnecessary risks and suffer great losses.

[pzgndr]

July 1812. Russia moves first again and continues to retreat. France continues a cautious advance.

Note: I occasionally see France move first and initiate a major battle at Kovno before Russia retreats. Or, if Barclay only retreats 2 areas instead of 3 and gets caught south of the Dvina River, France sometimes attacks. It all depends...

[pzgndr]

August 1812, Russia moves first again. Usually, France gets in a double move by now, but not this time. Russia continues its retreats, with Barclay only retreating 2 areas this time and Tormassov starting to retreat to Moscow from Kiev. France continues to advance cautiously. I expected an attack on Barclay but it didn't happen.

[pzgndr]

September 1812. France moves first. Napoleon had a good chance to fight Barclay but held back for some reason. (There is still some fog of war and variability with the computer opponent!) Davout focuses on Smolensk. Note here how the French minors are holding back to clean up the sieges in the rear, freeing up the French regulars to focus on Moscow.

[pzgndr]

October 1812. Russia moves first and completes all retreats to Moscow. French main body advances to Smolensk. Minsk continues to hold out but the rest of the rear area is secure.

[pzgndr]

November 1812. Russia moves first. Interestingly, Barclay with a partial force advances out of Moscow towards Smolensk. What will Napoleon do, attack Barclay or press on to Moscow? Napoleon goes for Moscow! Finally, a major battle 106:82, and France picks "Probe" to Russia's "Defend" and loses the battle to Russia. Moscow holds, for now...

At this point, I stopped. This quick playtest verified what I wanted to see happen up to the onset of winter. I think I have run through several dozen iterations of this 1812 campaign over the past couple of months and I've seen things swing both ways. Players should feel reasonably challenged now to play either side against a computer opponent acting a bit smarter on both offense and defense.

Alright, enough of 1812. I want to get back to 1805 for a bit over this next week. Then a couple weeks for some pbem group testing and I'll be sending the v1.23.08 official update package to Matrix. Then everyone can try out the new and improved AI along with all of the other bug fixes and enhancements.

[pzgndr]

Getting back to rerunning 1812 to verify changes made during my 1805 playtests are still good. I continued to see a few minor issues with depot placements, off-axis attacks on distant garrisons, and French bypassing Vitebsk to attack Smolensk. I undid an adjustment for placing depots, so this looks better now with France placing depots at Brest-Litovsk and now Minsk. For the off-axis exclusions, I was checking distances from unit to capital versus target garrison to capital. I was inspired to check total distance from unit to target garrison to capital and compare that. I think this is much better. For Vitebsk, I had an exclusion check for target areas being out of supply and forage value less than 3. Vitebsk is 2, so I changed the checks to include forage values less than 2. The old code just checked for being out of supply, and I've added a little more flexibility if foraging is possible. As seen in screenshot for September 1812, the French advance is progressing nicely.