German navy

[newpaintbrush]

@froodster:

okay goys like sewiously…

instead of all this talking, let’s see this played out in practice… I’d like to see this discussions carried out in actual play, as in “here’s my play, how do you counter THAT”.

For analysis purposes, it’d be best to do this with No Luck, so you can see what the expected result would be.

No Luck does NOT give expected results.

Flip a coin 4 times.  What’s the odds of 2 heads and 2 tails?

According to No Luck, it’s 100%.

According to actual probability - 37.5%.

A system that indicates a 100% likelihood of something that is actually only 37.5% likely to occur in acutality is not an accurate predictor of expected results.

God, kids these days and their “average results”!

If I eat 80 pounds of food every day over the course of a year, and 19 other people eat 0 pounds of food every day over the course of a year, should I say that on average, we each had four pounds a day?  SURE, because it’s TRUE, isn’t it?  But of course what REALLY happens is you have one extremely fat owl, and nineteen people dead of starvation!

D*** statistics anyways!

–

Things you get in mail order should always have a warranty and a return policy, and preferably a 100% satisfaction guarantee.

Since I do not believe mail order brides fall into that category, I will not order a mail order bride at this time.

However, if a company should exist that does offer those features for mail order brides, I will order, um, twelve.

[JamesG]

I’m not going to comment on if I think a German navy is a good or bad idea, because frankly I’m not sure.  But I will comment on this

@froodster:

instead of all this talking, let’s see this played out in practice… I’d like to see this discussions carried out in actual play, as in “here’s my play, how do you counter THAT”.

and this

@Imperious:

you play germany and follow the script. Ill post the counters or others can do it instead. Remember you have to follow the plan to its logical conclusion. Buy the 3 AP and 1 CV on turn one, and goto sz 6 and attack on G2

Also your medd fleet has to take gibralter and post at west medd. on G2 it has to come out together at #7

This exercise would prove nothing.  The game isn’t played that way in reality.  The German player does not say to the Allied player “This is exactly my plan, and I’m going through with it no matter what, try and stop me!” in actual play.

In actual play, the German player makes his G1 move, and then does his G2 move based on the Allied counter moves.  If the Allied counter moves makes the fleet unification a bad move, the German player does something else on G2.  Does that mean setting up on G1 for a unification that never happens on G2 was bad a G1 move?  Not necessarily, it all depends on what the Allies gave up doing in order to stop the G2 unification.

[newpaintbrush]

@JamesG:

I’m not going to comment on if I think a German navy is a good or bad idea, because frankly I’m not sure.  But I will comment on this

@froodster:

instead of all this talking, let’s see this played out in practice… I’d like to see this discussions carried out in actual play, as in “here’s my play, how do you counter THAT”.

and this

@Imperious:

you play germany and follow the script. Ill post the counters or others can do it instead. Remember you have to follow the plan to its logical conclusion. Buy the 3 AP and 1 CV on turn one, and goto sz 6 and attack on G2

Also your medd fleet has to take gibralter and post at west medd. on G2 it has to come out together at #7

This exercise would prove nothing.  The game isn’t played that way in reality.  The German player does not say to the Allied player “This is exactly my plan, and I’m going through with it no matter what, try and stop me!” in actual play.

In actual play, the German player makes his G1 move, and then does his G2 move based on the Allied counter moves.  If the Allied counter moves makes the fleet unification a bad move, the German player does something else on G2.  Does that mean setting up on G1 for a unification that never happens on G2 was bad a G1 move?  Not necessarily, it all depends on what the Allies gave up doing in order to stop the G2 unification.

I approve this message.

[Guest]

–

[frood]

I think I agree with James G there - the fleet unification play of building TRNs in the Baltic is probably more useful for what it forces the allies to do than for actual fleet unification. Hence, it prolly shouldn’t called a play but a threat.

An in response to NPB: Of course No Luck does not demonstrate actual percentages. However, from experience with my Sim, 9 times out of ten the No Luck result is the result that is most likely to occur out of 10,000 runs.

Now, if you want to test out a strategy, you have to follow it through multiple turns. To do that you have to choose one particular outcome for each turn so you can move on to the next. I’d say that the most common result is the one that should be followed (or perhaps one that is slightly worse, to account for a bit of bad luck).

I know that A&A is a wild and woolly game (esp. after my just concluded game with Switch) and you’ll rarely get the exact result that “No Luck” suggests. However, on average, your results will be within a certain range of those results.

How else would you test out a strategy somewhat objectively? By ranting about it in a forum? That’s not how the game is played either.

No. If you follow a strategy through and it works out based on average results, then I think it’s fair to say that it is a sound strategy.

Nowhere did I say that the No Luck result was a 100% certainty. But it’s the best outcome to follow in terms of analyzing the mid-point between good and bad luck and how a given strategy will play out.

After all, this is how casinos and insurance companies make money - they evaluate the probabilities of certain events. The fact that some smokers live to 90 and some vegans cash out at 30 does not mean they are wrong to set higher premiums for smokers.

Suppose you were gambling on the outcome of four coin tosses, and you had to stake $100 on one specific outcome. While two heads and two tails is nowhere near 100%, it would still be the best bet.

Here’s the breakdown:
all heads: 1/16
1 tail, 3 heads: 4/16
2 h, 2 t: 6/16
1 h, 3 t: 4/16
4 t: 1/16

Where would you stake your $100?

[newpaintbrush]

The single most likely result is not the MOST LIKELY result.

When you say that something that happens a good 6% of the time should be used as a general predictor of strategy, you end up looking bad.  Or you end up with a bad strategy.

If casinos and insurance companies used your brand of logic, smokers would have LOWERED insurance rates, because any particular smoker is EXTREMELY UNLIKELY to develop lung cancer, so therefore NO smokers develop lung cancer.  And casinos would be able to give 10,000 to 1 odds for the “0” on the roulette wheel, because since the “0” comes up so FEW times, it must THEREFORE really come up NO times.

[newpaintbrush]

It’s Springer time!

[squirecam]

@Imperious:

I dont play online. Can you just play it out yourself…

Remember we allready went this route and got not takers. I guess that was an idea that needs no testing… like most scientific theroy it does not need to be proven ( note hidden sarcasm)

Edit: Another flame

jamesG is absolutely correct. You cannot “test” this by having 100% premonitions. You dont KNOW what your opponent will do, and so you must move guessing. If your opponent sees an opening based upon your moves, he will CHANGE his strategy.

Also, the “expected” result, if 37% is the top result, means that 63% of the time the expected result WONT OCCUR. So you cannot take a result as being “expected” if 2 out of every three times it doesnt happen.

But feel free to play something out amongst yourselves if you feel it validates your position.

Squirecam

[frood]

Bring it on!

So now I look bad, do I? Oh yeah! Let’s frickin get it on like King Kong!

(only here do people trash talk each other about stats & probabilities)

Okay, NPB, I’ll try to make it simple so you understand. If you run my sim 10,000x, you’ll see a graph of all the results that occurred and how often they happened. That graph usually forms a bell curve. The fattest part of that curve may only be a 6% probability, but the most likely results will all be grouped around that possibility. And granted, the length of that graph is also an indicator of the “volatility” of a given battle. A battle with a wider range of results is harder to predict, and it is less likely that the median result will occur in a battle with a wide spread of results.

HOWEVER, how do you propose to test a strategy? By simply playing it out and letting the dice fall where they may?

If you want to analyze a strategy for future use in future games, what is more reliable? The way it ACTUALLY worked the last time you tried it, or the mathematically demonstrated result with the highest rate of incidence?

Granted, my sim can only really show a range of expected results, not THE expected result. However, for planning, is there anything better to use than the result that occurs in the mid-point of that range, where 50% of results are better and 50% are worse? That represents what will happen with medium luck, and that’s how you have to evaluate a strategy.

If you want to predict what will likely happen, would you rather rely on one sample, or 10,000 samples?

Maybe we’re talking about different things… is NPB saying that there is no meaningful way to analyze a strategy in the abstract based on median results?

[squirecam]

@froodster:

Bring it on!

So now I look bad, do I? Oh yeah! Let’s frickin get it on like King Kong!

(only here do people trash talk each other about stats & probabilities)

Okay, NPB, I’ll try to make it simple so you understand. If you run my sim 10,000x, you’ll see a graph of all the results that occurred and how often they happened. That graph usually forms a bell curve. The fattest part of that curve may only be a 6% probability, but the most likely results will all be grouped around that possibility. And granted, the length of that graph is also an indicator of the “volatility” of a given battle. A battle with a wider range of results is harder to predict, and it is less likely that the median result will occur in a battle with a wide spread of results.

HOWEVER, how do you propose to test a strategy? By simply playing it out and letting the dice fall where they may?

If you want to analyze a strategy for future use in future games, what is more reliable? The way it ACTUALLY worked the last time you tried it, or the mathematically demonstrated result with the highest rate of incidence?

Granted, my sim can only really show a range of expected results, not THE expected result. However, for planning, is there anything better to use than the result that occurs in the mid-point of that range, where 50% of results are better and 50% are worse? That represents what will happen with medium luck, and that’s how you have to evaluate a strategy.

If you want to predict what will likely happen, would you rather rely on one sample, or 10,000 samples?

Maybe we’re talking about different things… is NPB saying that there is no meaningful way to analyze a strategy in the abstract based on median results?

No, but this is evidence that “one game” wont show/prove anything. The dice being snarky and such, there can be wild variations in the plan.

But you dont need “a game” to see your “expected” result. Just math it out yourself…

Squirecam

[DarthMaximus]

But with the expected result you get +/- both good and bad.

So even if the expected result is 37%, meaning 63% the expected doesn’t happen.

This post:

@squirecam:

Also, the “expected” result, if 37% is the top result, means that 63% of the time the expected result WONT OCCUR. So you cannot take a result as being “expected” if 2 out of every three times it doesnt happen.

Yes but of the 63% that don’t occur (“expected”), roughly half are going to be BETTER and half worse, give or take a few % points.

You’re looking at a battle that has 65-70% acceptable results for you.

Even a slight roll down may be acceptable pushing this number to say ~75%.

[newpaintbrush]

@froodster:

Bring it on!

So now I look bad, do I? Oh yeah! Let’s frickin get it on like King Kong!

Jer-ry!  Jer-ry!  Jer-ry!

(only here do people trash talk each other about stats & probabilities)

Okay, NPB, I’ll try to make it simple so you understand

Ooo!  If you really want to make me read it enough to understand it, you should put it on a bikini, and put that bikini on Jessica Alba.  No wait.  You should PAINT it on Jessica Alba.  Yeaaaaa.

If you run my sim 10,000x, you’ll see a graph of all the results that occurred and how often they happened. That graph usually forms a bell curve. The fattest part of that curve may only be a 6% probability, but the most likely results will all be grouped around that possibility. And granted, the length of that graph is also an indicator of the “volatility” of a given battle. A battle with a wider range of results is harder to predict, and it is less likely that the median result will occur in a battle with a wide spread of results.

So we agree.  That Jessica Alba is hawt.

HOWEVER, how do you propose to test a strategy? By simply playing it out and letting the dice fall where they may?

The FIRST way to test a strategy is just to read it (or listen to it), and to listen to critics of that plan.  The SECOND way to test a strategy is to see what allowances the strategy has for When Things Go Horribly Wrong (if the strategy doesn’t provide for contingencies, it sucks).  Also in this step is to see what allowances the strategy has for When Things Go Really Right (because if the strategy doesn’t provide for unlooked for luck, it STILL is probably not that great).  Etc. etc.  Really, it’s a matter of thinking it out and discussing contingencies.  A few dozen games wouldn’t hurt either.

If you want to analyze a strategy for future use in future games, what is more reliable? The way it ACTUALLY worked the last time you tried it, or the mathematically demonstrated result with the highest rate of incidence?

I stick with what works.  Lol.

Granted, my sim can only really show a range of expected results, not THE expected result. However, for planning, is there anything better to use than the result that occurs in the mid-point of that range, where 50% of results are better and 50% are worse? That represents what will happen with medium luck, and that’s how you have to evaluate a strategy.

WHO can visit the most violence on deceased animals of the equine persuasion?  I said my bit.

If you want to predict what will likely happen, would you rather rely on one sample, or 10,000 samples?

That’s my point.  But I think I’m the one that’s talking about the 10,000 samples and that you’re the one that’s talking about the one sample.  Of course, I’m sure you have rather a different understanding of the matter.

Maybe we’re talking about different things… is NPB saying that there is no meaningful way to analyze a strategy in the abstract based on median results?

My god, that is kind of what I’m saying, isn’t it?

Russia has a number of possibilities to choose from first turn.  One of those possibilities is an attack on three German territories.  Each single one of those attacks SLIGHTLY favors the Russians.  If even ONE of those attacks fails, it really sucks for Russia.  But - if you want to go by the SINGLE most likely outcome, which is what you will get in NoLuck or LowLuck, the Russians should ALWAYS execute those slightly favorable attacks.  That is, the Russians will WIN 100% OF THE TIME.

However, in REAL life, dealing with REAL probabilities, USUALLY what happens if Russia tries this, is that at LEAST one of those battles fails quite spectacularly.  And when that DOES happen, it REALLY sucks for Russia.

I think it works out to 52%, 54% and 56% or something like that for Russia winning any particular one of those battles, but for Russia to win ALL of those, the probability needs to be 52% * 54% * 56%, or about 16%.

So if you have somebody that based a strategy on NoLuck, or even LowLuck, that somebody is going to do something that IS going to succeed 100% OF THE TIME in NoLuck, or LowLuck.  But that SAME strategy is only going to work 16% of the time with real dice.  That is to say, that strategy is going to FAIL 84% OF THE TIME.

[newpaintbrush]

@DarthMaximus:

But with the expected result you get +/- both good and bad.

So even if the expected result is 37%, meaning 63% the expected doesn’t happen.

This post:

@squirecam:

Also, the “expected” result, if 37% is the top result, means that 63% of the time the expected result WONT OCCUR. So you cannot take a result as being “expected” if 2 out of every three times it doesnt happen.

Yes but of the 63% that don’t occur (“expected”), roughly half are going to be BETTER and half worse, give or take a few % points.

You’re looking at a battle that has 65-70% acceptable results for you.

Even a slight roll down may be acceptable pushing this number to say ~75%.

That analysis would be great if there were just ONE BIG BATTLE in Axis and Allies.

What you really have is lots of little battles.

So let’s say you had a big-ass fighter and naval battle that turned out poorly for you (the 30% happened there), while you hosed Russia on trading territories that turn (you got your 70% there).

You just got smoked.

[DarthMaximus]

@CrazyStraw:

This is particularly critical in naval engagements because so many expensive units are typically involved.  In a land battle, deviating +/-2 hits from the mode (calculated outcome) is likely to be a range swing of 12 IPCs (+/-6 for 2inf).  But in the water, once you’re through the fodder you could be looking at a +/-2 hits being a range swing of 80 IPCs (+/-40 for 1car 1btl).

But many of the Naval units don’t need to be replaced.
If you have a major Naval battle (unfication in Sz 7) and the US and UK both lose BB, DD, AC etc and Ger loses the battle even though the Allies lost more IPC say 150-100, they don’t need to rebuy an AC, BB, DD etc.

Then can buy trns only, unless Germany continues to buy navy or air.

The Allies have disposable income, the Axis do not.

I’m not advocating just throwing away high valued units in any battle but there are acceptable risks for taking significant losses.

Even on Land, say Germany attacks Moscow and losses 40 units while Russia losses only 20.  Bad attack, right?
Now Japan attacks and losses 30 to Russia’s 20, but Japan takes Moscow.  The Axis lost 70 units to 40, but they gained a positional adv.

If sinking a unified fleet cost the Allies more, then so be it, but if they gain a positional adv out of the whole process, such as a stronger Russia than that is something an Allied player may be willing to accept in exchage.

[Imperious Leader]

Flip a coin 4 times.  What’s the odds of 2 heads and 2 tails?

According to No Luck, it’s 100%.

According to actual probability - 37.5%.

A system that indicates a 100% likelihood of something that is actually only 37.5% likely to occur in acutality is not an accurate predictor of expected results.

Can you elaborate on this just a little bit?

[DarthMaximus]

@newpaintbrush:

@DarthMaximus:

But with the expected result you get +/- both good and bad.

So even if the expected result is 37%, meaning 63% the expected doesn’t happen.

This post:

@squirecam:

Also, the “expected” result, if 37% is the top result, means that 63% of the time the expected result WONT OCCUR. So you cannot take a result as being “expected” if 2 out of every three times it doesnt happen.

Yes but of the 63% that don’t occur (“expected”), roughly half are going to be BETTER and half worse, give or take a few % points.

You’re looking at a battle that has 65-70% acceptable results for you.

Even a slight roll down may be acceptable pushing this number to say ~75%.

That analysis would be great if there were just ONE BIG BATTLE in Axis and Allies.

What you really have is lots of little battles.

So let’s say you had a big-a** fighter and naval battle that turned out poorly for you (the 30% happened there), while you hosed Russia on trading territories that turn (you got your 70% there).

You just got smoked.

Fine, then don’t attack when 70% is acceptable to you, wait for 80, 90 whatever you comfort zone is.

If I lose a “Must Win” battle where the 30% negative comes out so be it.  But I’ll do the same strat and when it comes to a similar battle I’ll take my 70% chances.  And I’ll do that over and over and over, and out of 10, 50, 100 games I’ll have a 70% win ratio.

You can’t win every game.  I write off 20% right away, I try to win the 80% that aren’t dice skewed in rds 1-2.

If I get smoked by a 30% outcome on an attempted take down in Sz 7, then boo hoo hoo.

It doesn’t make the decision to attack a bad one.  Maybe I can refine it to get a 80-85% chance but that still means screw jobs can happen.

You can’t play worry about losing battles where you have 70-90% chances to take.  Now if you consistanly attacking in battles with 55-60% you may want to change things.

[frood]

Okay good point on those Russian attacks.

I was thinking more in terms of testing German naval strategy. There we’re talking about one German build, one UK / US fleet move, then potentially one German attack on the allied fleet, and then one Allied counter-attack to see what’s left at the end of the day.

For those purposes, I think tracking the No Luck result can be useful, just to determine whether in one specific theater with one fairly predictable sequence of moves and replies, can I expect on average to come out ahead?

But even in that situation, I would still want to consider the spread as well. Personally I think it’s better strategy to engage in battles with highly predictable results. Battles with a wider spread (typically where forces are more closely matched) there is too much chance of disaster.

What I am NOT saying is that you should rely on the NL result happening in every single one of your battles.

The trouble in the forum discussions that attempt to do this is that it seems like everyone does the math a little differently, makes different starting assumptions, etc. That’s why I’d like to just see it actually played out - not in a full game, but just in a little exchange like this:

Froodster: Ok on G1 I buy 3 Trns and put them in the baltic, and a bunch of Inf in Germany
NoobGuy29988: zomgzors I am going to stop you from unifying your fleet and a move X Y and Z into SZ 6. (or whatever the opposing theory is)
Froodster: Ok I bring A B and C to SZ 6 and according to the total punch I blow up your X Y and Z and I have A and half of B left
NoobGuy29988: oh yeah well I bring V and W and sink your A and 1/2 B and I have B left according to No Luckzorz
Froodster: Ok Germany is done in the Atlantic - let’s see we each lost M and N IPCs respectively and now the board looks like this…:

I just think that would be a more helpful/interesting way to test out a German naval strat instead of all this pure theory where one person runs through the whole battle in their head. It takes two to tango, so I want to see how a strat would play out between an opponent of the strat and a proponent of the strat, because neither one may fully appreciate all the starting assumptions of the other side.

And no the NL result won’t always happen - but it’s the best “benchmark” to go by. What do you propose as an alternate?

[Imperious Leader]

Darth is right again … as usual…Why are people not looking at the opposite side of the coin?

If something has a 37% clip of working  it also has a 63% clip of not working. Strategy is supposed to have some outcome and my previous label as a “gambit” is even more reinforced. This plan is a gambit. Germany is really making a sacrifice to gain some positional advantage on the western front.

Here is an assesment that probably comes only after many posts and thinking on this…

the fleet unification play of building TRNs in the Baltic is probably more useful for what it forces the allies to do than for actual fleet unification. Hence, it prolly shouldn’t called a play but a threat.

Thats also what i have been saying in the last “German buy thread”  this is not really a viable plan but a direct threat of something requiring the Allies to shift a few turns worth of buys… After all is said and done if the plan is executed it will result in little gain for germany in net IPC, while it will gain them some breathing room on the western front, while to the detriment of the eastern front for about the same period.

[Imperious Leader]

Also, We should look at what the Soviets are doing for the 1st 3 turns while this plan is happening. The German air force is likely tied up for turns 1-2 and partially into turn 3

The Soviet Union has grown stronger… so what are they doing in all this?

[DarthMaximus]

@CrazyStraw:

@DarthMaximus:

But many of the Naval units don’t need to be replaced.
If you have a major Naval battle (unfication in Sz 7) and the US and UK both lose BB, DD, AC etc and Ger loses the battle even though the Allies lost more IPC say 150-100, they don’t need to rebuy an AC, BB, DD etc.

Where that matters in the specific instance of the GUF (German Unified Fleet) is that if you go the battle with a thin margin of winning, you could be in for cascade failure. Â

Say your attack should win for you with a mutual kill.  Instead of getting the mutual kill you hit light by 2 hits on the American attack.  1car 1btl survive; Germany sends 2ftr to refill the carrier.

Now you’ve gone from a situation where Germany would have never been in the water again to a situation where you still have to sink 1car 2ftr 1btl with already depleted Allied forces.

The simple swing of hitting light by only 2 hits is a devastating turn of events that will require many more IPCs of gear to rectify.

Now, here’s where it gets really interesting.  Let’s say the battle goes the other way and Germany is the team that hits light.  Say the US was doing the heavy lifting with air power and the units that survive are 2des.  Does Germany care about that?  Does 2des really change any German activity?  Probably not.  So in the instance I laid out, if things go slightly good for Germany, the Allies suffer dramatically.  If things go slightly good for the Allies, the Germans probably don’t care.

Ah, okay.  That makes sense.