AI for MWiF - USSR

[Shannon V. OKeets]

ORIGINAL: Froonp
To add to the picture I posted earlier, and to try to have this defense better.

What do you think if the troops defending Pskov, Novgorod and Vitebsk were used along with those in Riga Kaunas, Vilna and Minsk to apply the strategy outlined in point 6 of my precedent post (#16).
[In the north, defend with some strong unit stacks, always in the woods. Build a ZoC network, and withdraw when he advances. Do not compromise the Red Army for a flipped unit, let it die. ]

Those forward units would hope to survive more or less the first ground strikes (they would be placed in the woods), and they would try to retreat each time the German advance, to try to make the German loose time, and to try to save units.

One point I may have forgotten to stress in my previous posts about Germany & Russia is that the objective of the German Army during the first blow to Russia is more to destroy Russia units than to gain territory. Knowing that, the Russian should try to save his army att he expense of territory if necessary.

Cheers.

Though I am not always successful, I try to restrain myself from inserting comments as a player. It isn't that I am uninterested, or have no opinions, but rather that I have to keep focused on the task of creating the software.

So, I leave your questions here unanswered by me.

[Shannon V. OKeets]

ORIGINAL: buckyzoom
One thing the the Russian AI should have is a pact breaker calculator. I've seen two situations that come up in WIF games (other than Germany trying to break the pact). One is...

Packing the Border
One is that the Russian player packs the border. What this means is that the put enough troops on the border so that the Germans cannot break the pact until '42. This is mathematically possible and a computer should be really good at this.

The AI would need to take into account the quality of their chit picks. Perhaps if the standard deviation for their picks is consistently on the postive side for x turns a 'pack the border' strategy is adopted.

The other is...

Bluffing
I've seen/been the Russian player in late '41 and early '42 starts turning chits face down and move his forces forward (if not at war with Japan, then Russia declares war on Italy). This indicates that the USSR player believes he can break the pact. This can force the German player to declare war prematurely.

Other thoughts...

Spain or London
If Germany declares war on Spain or invades the UK this is an indication the USSR has until '42 before Germany attacks. This opens up options like,

...invade Persia
...invade Iraq
...invade Bulgaria
...demand Finnish borderlands

Japan
There has been some pretty pointed arguments about when and even if Russia should ever attack Japan. I won't comment, but declaring war on Japan allows some pretty interesting options for the Russian AI. Since they are active,

...they could help garrison other Allied territories or countries. These could include France (generally speaking the Italian border mountains) or the Suez if a route is open. Other options are possible.
...it would allow other countries to bring troops into Russia. It's not uncommon for the Commonwealth to garrison Murmank or even bring land troops in to attack Persia from the USSR.

I like all of these except the last bit about moving Russian troops to other countries and vice-a-versa. It is the need to include HQ units that bothers me (Composer99 pointed this out earlier). To my eye, those units could do more in direct confrontation with the Axis on their own fronts instead of being moved overseas where they would have limited support from their own country (and have to coordinate with nearby allied troops). All in all a nightmare for the AIO to keep straight.

The discussion of the invasion of Bulgaria has come up a couple of times. I assume this is after conquering Rumania?

[c92nichj]

The discussion of the invasion of Bulgaria has come up a couple of times. I assume this is after conquering Rumania?

No this is done by invasion. Set up the tarnsport and good Div's in the balck sea area region and you will have a good chance of getting intop Bulgaria during '39, goal is not to conquer her, but rather to deny germany her resource and the possibility tpo rail home the turkish resource.

[c92nichj]

I like all of these except the last bit about moving Russian troops to other countries and vice-a-versa.

A CW garrision of murmansk is not that bad. If Russia is hardpressed and she will be if Germany focus all of his attention on her, just two corps more defending russia and not defending Murmansk will mean a lot.
Remember that the game is won or lost in Russia, and I cannot see many things that the CW could do with Gort and two corps? Gort only needs to stay until the corps have landed and only need to be for a turn in Russia.

[buckyzoom]

ORIGINAL: Shannon V. OKeets

I like all of these except the last bit about moving Russian troops to other countries and vice-a-versa. It is the need to include HQ units that bothers me (Composer99 pointed this out earlier). To my eye, those units could do more in direct confrontation with the Axis on their own fronts instead of being moved overseas where they would have limited support from their own country (and have to coordinate with nearby allied troops). All in all a nightmare for the AIO to keep straight.

The discussion of the invasion of Bulgaria has come up a couple of times. I assume this is after conquering Rumania?

The idea of using Russian troops to garrison France is in the Strategy Notes of the game. I personally don't like it, but if you're open to early war with Japan it can work and the USSR troops magically teleport home after France is conquered.

As the USSR I've toyed with the idea of invading Portugal in late '40 if it looks like Germany is setting up for a 'Close the Med'. Portugal has no land troops so its not tough tough if I can get basing worked out. (Once again it only makes sense if you are planning on being active.) Once its conquered the English and US (once they are in the war) can start building up troops there and Germany can't do a thing about it until they declare on the USSR.

Remember major powers don't need HQ's in minor countries or territories. Foreign troop commitment applies to minors in minors (or majors) and majors in majors.

I believe others have answered your other questions.

[buckyzoom]

Without providing any proprietaryinformation can you provide an outline of how the decision tree will look for the AI? It may help how I organize my thoughts on this subject.

For example, will it mirror the turn sequence?

1. Roll Initiative
2. Roll Weather
3. Impulse
3.1 Impulse Declaration... and so on until

17 Partisan Phase

Maybe this has already been answered. If so please refer me to the appropriate post and I will happily dig into it!

[buckyzoom]

What kind of AI are you planning on having for production? (Again refer me to a thread or post if this has been asked/explained.)

I was wondering whether you were taking a template approach based on a high level strategy (for example I mentioned packing the border and an AI around this could optimize builds for garrison value) or is there a distribution model where based on the country an emphasis will be placed on unit class and type?

Thanks,

b5

[Froonp]

ORIGINAL: c92nichj
Remember that the game is won or lost in Russia, and I cannot see many things that the CW could do with Gort and two corps? Gort only needs to stay until the corps have landed and only need to be for a turn in Russia.

While this may be often true in WiF FE, I definitely think that the CW has better to do to help Russia that send precious CW corps in Russia.
There is a second front to open, and even if the CW is not strong enough alone to do it, the CW must begin the Job. Clear eastern Africa, conquer sardinia asap, conquer Tripoli when possible, defend its mediterranean dominions.. All this help Russia tremendously, because everything the CW force Germany to do is good for the Allies. Forcing Germany to come helping a pressed Italy in the early Barbarossa is a very good thing.

[Shannon V. OKeets]

ORIGINAL: buckyzoom
...
As the USSR I've toyed with the idea of invading Portugal in late '40 if it looks like Germany is setting up for a 'Close the Med'. Portugal has no land troops so its not tough tough if I can get basing worked out. (Once again it only makes sense if you are planning on being active.) Once its conquered the English and US (once they are in the war) can start building up troops there and Germany can't do a thing about it until they declare on the USSR.

Remember major powers don't need HQ's in minor countries or territories. Foreign troop commitment applies to minors in minors (or majors) and majors in majors.

Aligning and declaring war on minors is part of what I am looking for as part of a strategic plan. These ideas fall under that category for the USSR.

I will eventually gather all of the individual ideas into strategic plans for each of the 8 major powers. Optional bits (like Portugal) add seasoning.

[Shannon V. OKeets]

ORIGINAL: buckyzoom
Without providing any proprietaryinformation can you provide an outline of how the decision tree will look for the AI? It may help how I organize my thoughts on this subject.

For example, will it mirror the turn sequence?

1. Roll Initiative
2. Roll Weather
3. Impulse
3.1 Impulse Declaration... and so on until

17 Partisan Phase

Maybe this has already been answered. If so please refer me to the appropriate post and I will happily dig into it!

The thread "Artifical Intelligence for WIF" is the best place to start. The conceptual design for the AI opponent (AIO) will not use decision trees. Instead, it uses a fixed hierarchy of 8 decision makers each of which makes primarily autonomous decisions based on information provided by the others. Each decision maker gathers information, processes that information, and transmits the results of its analysis to the others.

Each item you listed (1 -> 3.1 ...) has been assigned to a decision maker as part of his task list. When a decision needs to be made during the game ("Ask for a reroll?"), the decision maker with the responsibility for making that decision, does so. How the decision maker makes that decision is generally written in the form of rules: "If ..., Then ...". This means that they are based on satisfying a conditional statement prior to an action being taken. At the end of a long list of conditionals there is always a final clause that says "If none of the above is true, Then do such and such."

[Shannon V. OKeets]

ORIGINAL: buckyzoom

What kind of AI are you planning on having for production? (Again refer me to a thread or post if this has been asked/explained.)

I was wondering whether you were taking a template approach based on a high level strategy (for example I mentioned packing the border and an AI around this could optimize builds for garrison value) or is there a distribution model where based on the country an emphasis will be placed on unit class and type?

Production decisions are dominated by the strategic plan. There is a small amount of leeway within the strategic plan dictates, but not a lot. Again rules are being used. Simply think in terms you would use when talking to another player. That is the advantage of rules; they are very similar to normal communication between experts in the field.

[composer99]

ORIGINAL: c92nichj
Remember that the game is won or lost in Russia, and I cannot see many things that the CW could do with Gort and two corps? Gort only needs to stay until the corps have landed and only need to be for a turn in Russia.

While this may be often true in WiF FE, I definitely think that the CW has better to do to help Russia that send precious CW corps in Russia.
There is a second front to open, and even if the CW is not strong enough alone to do it, the CW must begin the Job. Clear eastern Africa, conquer sardinia asap, conquer Tripoli when possible, defend its mediterranean dominions.. All this help Russia tremendously, because everything the CW force Germany to do is good for the Allies. Forcing Germany to come helping a pressed Italy in the early Barbarossa is a very good thing.

Agreed. Gort & 2 corps are immensely helpful for the CW and Allied cause in many places throughout the world; they are only marginally helpful defending Murmansk. If they are in Burma or India, they are defending those countries from Japanese attack or advancing in the British counter-offensive. If they are in the Med, they are helping defend Egypt or Gibraltar, or they are helping clean the Italians out of the Med. If they are in England, then Gort is probably assisting the strategic air war by re-organizing bombers or escorts and the 2 corps, if there are spare AMPH about, are forcing the Germans, if they want to play smart, to put extra units into defending France, the Low Countries, and Denmark.

The best way for the CW to assist Russia is to pull German fighters, land units, HQs, and attention away from the Eastern Front. This means the strategic air war, pressing on Italy whenever possible (and preferably threatening to invade Italy from 1941 on), launching tactical invasions into Western Europe to indispose German naval forces or kill land units (and then pulling back out), and doing the utmost to sink convoys in the Baltic Sea.

The Soviet Union can survive getting thrown off the European map if enough of the Red Army survives, so its main job is to preserve that army. The less Germans are in Russia, and the more often Germany has to call actions that are not optimal to fighting in Russia, that is, anything other than lands during the summer, the better chance the Red Army has.

[c92nichj]

The less Germans are in Russia, and the more often Germany has to call actions that are not optimal to fighting in Russia, that is, anything other than lands during the summer, the better chance the Red Army has.

I cannot see a German player will do anything but land actions during the summer of a '41 Barbarossa, and I can hardly see what the CW could do to prevent it, except maybe to put all his convoys unguarded so Donitz cannot resist to come out to play.

An invasion of Italy,Denmark, France would require lands for the three railmoves, and under normal circumstances the CW couldn't make an impressive invasion during '41 anyhow.
Defending Suez or the far east during summer '41 in a barbarossa should not require Gort, the axis will not be able to do anything there in that timeframe if they are serious about a barbarossa. So that leaves the stratbomb campaign which will atmost affect germany's forces in russia during '42 '43 hardly anyhelp to a russian on the ropes.

Losing murmansk, and probably Vladivostok means that the only way left for lendlease into russia is through Persia and that can be a dangerous play.
With the terrible supply situation in the murmansk region a single militia or Garrision goes a long way as far as protection comes.

-Nicklas

[buckyzoom]

The Russian AI is going to be much simpler than the other AI in some ways. Much of their decisions can be made independently of other nations.

Assuming a Global War Scenario the first choice for the Soviet Union is whether it will be active or passive.

An active Russia will be conquering minors and possibly attacking majors. How active is a separate question that I will provide in a later post.

A passive Russia builds up for the eventual Barbarossa.

(Something to think about is conditions that will change Russia from active to passive and vice versa. E.g. - the US has entered the war and there is no longer US entry hits for declarations of war. China is in trouble and it might be saved by a Russian declaration of war on Japan. The WAXIS are executing a close the Med so there is a year to gobble minors before Barbarossa. If I attack Finland it could cost two US entry chits and there is only one in the European chit pool.)

Active or passive will help answer the question of how to setup.

Passive is easy.

On the Nazi/Soviet border you always setup to defend against the foolish, but still possible, 'Russia First' gambit. You can generalize this assumption to include a Rumania heavy setup in case Germany declines the Bessarabian land claim (assuming you make it). In an oil/gas game another generalization is to setup every strategic bomb factor you can within range of the Rumania oil fields to discourage the Axis from denying the land claim.

On the Manchurian border you setup based on whether you plan to try and fight hard if the Japanese choose to take Vlad.

After a passive Russia has setup there is only the matter of determining whether you will pack the border or defend back. I won't go into the details of defense here, but the build schedule shouldn't be too tough to make rules based.

When packing the border you base your builds on the build point to garrison value ratio. (You're not going to be building planes or armor.)

When defending back (with or without speed bumps) a couple of simple rules you can establish are around a point of attack concept. You create a schedule so that all possible Russian INF and ARM class units are on the spiral for a May/June '41 Barb (the point at which they are likely to attack). If the May/June Barb does not materialize then reset the point of attack date to May/June '42.

(A more cautious Russian player would plan for March/April if a speed bump defense is used. May/June '42 is statistically the second longest turn of a year. If the German's can clear the bumps in March/April and crack the Dnepr in May/June it is likely that Jul/Aug will see fun times for the Germans.)

I'll try and post more about active Russia on Monday if time permits. Gotten run work beckons...

[Shannon V. OKeets]

ORIGINAL: buckyzoom
The Russian AI is going to be much simpler than the other AI in some ways. Much of their decisions can be made independently of other nations.

Assuming a Global War Scenario the first choice for the Soviet Union is whether it will be active or passive.

An active Russia will be conquering minors and possibly attacking majors. How active is a separate question that I will provide in a later post.

A passive Russia builds up for the eventual Barbarossa.

(Something to think about is conditions that will change Russia from active to passive and vice versa. E.g. - the US has entered the war and there is no longer US entry hits for declarations of war. China is in trouble and it might be saved by a Russian declaration of war on Japan. The WAXIS are executing a close the Med so there is a year to gobble minors before Barbarossa. If I attack Finland it could cost two US entry chits and there is only one in the European chit pool.)

Active or passive will help answer the question of how to setup.

Passive is easy.

On the Nazi/Soviet border you always setup to defend against the foolish, but still possible, 'Russia First' gambit. You can generalize this assumption to include a Rumania heavy setup in case Germany declines the Bessarabian land claim (assuming you make it). In an oil/gas game another generalization is to setup every strategic bomb factor you can within range of the Rumania oil fields to discourage the Axis from denying the land claim.

On the Manchurian border you setup based on whether you plan to try and fight hard if the Japanese choose to take Vlad.

After a passive Russia has setup there is only the matter of determining whether you will pack the border or defend back. I won't go into the details of defense here, but the build schedule shouldn't be too tough to make rules based.

When packing the border you base your builds on the build point to garrison value ratio. (You're not going to be building planes or armor.)

When defending back (with or without speed bumps) a couple of simple rules you can establish are around a point of attack concept. You create a schedule so that all possible Russian INF and ARM class units are on the spiral for a May/June '41 Barb (the point at which they are likely to attack). If the May/June Barb does not materialize then reset the point of attack date to May/June '42.

(A more cautious Russian player would plan for March/April if a speed bump defense is used. May/June '42 is statistically the second longest turn of a year. If the German's can clear the bumps in March/April and crack the Dnepr in May/June it is likely that Jul/Aug will see fun times for the Germans.)

I'll try and post more about active Russia on Monday if time permits. Gotten run work beckons...

Here is what I wrote to start these threads on strategic plans. Your post gave good answers to some of them but left blanks on others.

(1) Victory cities to be taken and/or defended.
(2) Which major powers to declare war on, when, and any associated conditions concerning same.
(3) Which minor countries to declare war on, when, and any associated conditions concerning same.
(4) Which minor countries to align, when, and any associated conditions concerning same.
(5) Expected areas of conflict: with whom, where, type of combat (land, naval, air, convoys).
(6) A master production plan by unit type and/or gearing limits. This does not have to be detailed, just a broad outline.
(7) A time line for the strategic plan’s major milestones.

You can take as a given that there will always a lust on my part for more detail (items 2, 3, and 7). Though items 2 and 3 have been covered by others, in their posts on the USSR strategic choices. You are the first to add elements to item 7, and more detail there would be helpful.

I am particularly interested in item 1. As a looooooooong time player of many different games about the German-USSR conflict during WWII, I have my own ideas on how to define the relative importance of the victory cities. However, for that precise reason I am interested in the opinions of others, so I am not too locked into my own view.

[buckyzoom]

I'm just getting started

I was going to provide an overview first and then dive down into the detail. I probably shoud have mentioned that at the beginning of my post huh!

I was thinking in terms of a high level flow where you pick a scenario, decide a basic strategy, set up your units to support that strategy and the the game happens and you change your strategy...

Playing WIF (speaking for myself) is in many ways pattern matching, for example if Player A is doing X and Player B is doing Z then I can respond with Q or W kinda thing. I was going to describe mid grain patterns and responses before a more detailed dive, but I'm flexible. I'll provide some victory city info next.

What do you mean by victory city? Is it in the WIF context? When I play Russia I first play to live, and if I live, I then play to crush. I'm wondering the context of how this will be used in the AI? As Russia defending a victory hex because its a victory hex can be counter productive to winning. (IMHO)

To me there are key places to hold that are more important than victory hexes. For example keeping Germany from activating Turkey. Or my personal favorite, Gomel which I think is critical to hold for as long as possible. What info do you want on these kind of concepts?

[Shannon V. OKeets]

ORIGINAL: buckyzoom

I'm just getting started

I was going to provide an overview first and then dive down into the detail. I probably shoud have mentioned that at the beginning of my post huh!

I was thinking in terms of a high level flow where you pick a scenario, decide a basic strategy, set up your units to support that strategy and the the game happens and you change your strategy...

Playing WIF (speaking for myself) is in many ways pattern matching, for example if Player A is doing X and Player B is doing Z then I can respond with Q or W kinda thing. I was going to describe mid grain patterns and responses before a more detailed dive, but I'm flexible. I'll provide some victory city info next.

What do you mean by victory city? Is it in the WIF context? When I play Russia I first play to live, and if I live, I then play to crush. I'm wondering the context of how this will be used in the AI? As Russia defending a victory hex because its a victory hex can be counter productive to winning. (IMHO)

To me there are key places to hold that are more important than victory hexes. For example keeping Germany from activating Turkey. Or my personal favorite, Gomel which I think is critical to hold for as long as possible. What info do you want on these kind of concepts?

I use victory cities to record progress because they are precisely defined and apply to all the major powers. Each major power seems to have certain non-victory cities which are key to its offensives and/or defenses. And the importance of those cities are sometimes in the eye of the player. Please add any that you deem relevant for the USSR. At the end of the day though (or should I say the end of the game?) what matters are the victory cities, so the strategic plan should identify which ones the USSR is going to go after (or merely keep).

As to what to work on next, that is entirely up to you. I keep at least 3 different types of activities going in my work on MWIF. I work on creative ideas (usually for the AIO), rigorous analytical tasks (programming and debugging), and rather mindless tasks (like entering scenario data or typing up edits I have made to a printed copy of code or documentation). On that topic, I have finished 4 of the scenarios, which when added to the 3 that Chris did, leaves 4 left to type in. Very tedious stuff, reading from the spreadsheet and translating the fine details into the MWIF scenario setup language - blagh. The reason I have these 3 types of activities is that I find I am most productive when which type of activity I select to work is based on the clarity of my head. I have learned that it is always a bad idea to work on code when your head is on crooked.

[buckyzoom]

Victory Cities

First victory cities should only influence the Russian AI when the USSR goes over to the offense. Before this the only concern is preserving it's army. Do you want generalized rules for when the Russian army should switch from offense to defense?

Here are 17 victory hexes that Russia has a shot at. I've put them roughly in the priority order they are for me when playing Russia.

Sverdlovsk (If the Axis control Sverdlovsk I've got big issues...)
Moscow (Automatic victory hex and production bonus with Leningrad and Stalingrad)
Kiev (Production bonus, plus a red factory)
Leningrad (Production bonus, plus a red factory)
Helsinki (Eliminates a minor, plus many winter bonus units)
Bucharest (Eliminates a minor, plus likely will gain three oil along the way)
Budapest (Eliminates a minor, plus a red factory)
Warsaw (Liberates a minor, plus a red factory)
Belgrade (Eliminates or liberates a minor)
Berlin (Automatic victory hex, red factory)
Vienna
Prague
Vladivostok
Bagdhad
Port Arthur
Lan Chow
Istanbul

Priority order could change based on game strategy or a unique opportunity presents itself. E.g. - if Germany goes defensive in '42, and Japan weak in Manchuria then attack Manchuria until rail from China to Russia is open. Once rail is open have China lend all production and excess resources to USSR.

Options can also influence this. If playing Hitler's War the value of Moscow, Leningrad and Kiev change.

What additional information would you like?

[Shannon V. OKeets]

ORIGINAL: buckyzoom
Victory Cities

First victory cities should only influence the Russian AI when the USSR goes over to the offense. Before this the only concern is preserving it's army. Do you want generalized rules for when the Russian army should switch from offense to defense?

Here are 17 victory hexes that Russia has a shot at. I've put them roughly in the priority order they are for me when playing Russia.

Sverdlovsk (If the Axis control Sverdlovsk I've got big issues...)
Moscow (Automatic victory hex and production bonus with Leningrad and Stalingrad)
Kiev (Production bonus, plus a red factory)
Leningrad (Production bonus, plus a red factory)
Helsinki (Eliminates a minor, plus many winter bonus units)
Bucharest (Eliminates a minor, plus likely will gain three oil along the way)
Budapest (Eliminates a minor, plus a red factory)
Warsaw (Liberates a minor, plus a red factory)
Belgrade (Eliminates or liberates a minor)
Berlin (Automatic victory hex, red factory)
Vienna
Prague
Vladivostok
Bagdhad
Port Arthur
Lan Chow
Istanbul

Priority order could change based on game strategy or a unique opportunity presents itself. E.g. - if Germany goes defensive in '42, and Japan weak in Manchuria then attack Manchuria until rail from China to Russia is open. Once rail is open have China lend all production and excess resources to USSR.

Options can also influence this. If playing Hitler's War the value of Moscow, Leningrad and Kiev change.

What additional information would you like?

Very good.

A few questions.

1. You did not mention Teheran in your list.
2. My inclination from your list would then be to defend the USSR cities in the reverse order and give them up: Leningrad, Kiev, Moscow, and lastly, Sverdlovsk. Is that a valid assumption? If not, why not?
3. Though keeping the USSR army intact is clearly important, giving up all the resources, factories, and rail lines also guarantees defeat. Do you have a sense of how that tradeoff between maintaining the existing army and maintaining a production capability (to produce new units) should be judged?

[c92nichj]

2. My inclination from your list would then be to defend the USSR cities in the reverse order and give them up: Leningrad, Kiev, Moscow, and lastly, Sverdlovsk. Is that a valid assumption? If not, why not?

I would not do this first and foremost preserve the russian army, any lost cities can be taken back later.
The key thing for russia is to trade land for time and to trade land for units. Do not trade units for land.

To protect the production rail out factories and demand lendlease. Good spot to rail factories are:
Murmansk - will enable you to get lendlease to the north even if the rail north of moscow is cut.
Baku - if caucasus is cut of from the rest of the factories you can still use your resources here.
The urals - easy to defend.

If the red army is destroyed so is Soviet.
Areas to defend:
Leningrad, put a couple of strong units here together with some saved oil

Sevastopol, the fort makes it hard for the germans to take it so if they want iot they need to fight for it.

Rostov, Also quite easy to defend , if the axis want it make him pay for it.

Caucasus mountain range north of Turkey, try to stop the germans from aligning Turkey. It is also quite easy to defend with all the alpine hexsides.

Your railed factories in the Urals - you need to keep producing units.

Even if you loose pretty much all else on the european map you still have a good chance to come back if your army is intact and you get plenty of lendlease.

Rostov Sevstopol and Leningrad will be good places to place reinforcements and build up your comeback from, be cautious though and don't get on the offensive too early.