Solo Altitude using the App

**Derrick** · 08-17-2020 20:40

Hey, sorry for the repost from another thread, but I was inspired tonight to give this another shot after I wrote out my Altitude Rules for Solo. I do only play 4 levels.

Here is what I posted in the other thread:

---------------------------------------------------
To understand where I'm coming from, here is my ideal state when playing altitude:

1. Enemy in arc/range
2. Not in enemies arc/range
3. At same altitude as enemy
4. Have as many climb counters as possible (unless at highest level)

So for AI. The first time per turn a straight is revealed, follow these questions. If yes, stop and do the thing. If no, continue.

(a) Is 1, 3, and 4 met? Straight (or sideslip - see below)
(b) Is 1 met and 3 not met? climb/dive in direction of target
(c) Is 1 and 3 met and 4 is not met? Climb
(d) Is 2, 3, and 4 met? Straight (or sideslip - see below)
(e) Is 2 and 3 met and 4 not met? Climb
(f) Is 2 not met and 3 and 4 met? Flip a coin. Head climb. Tails dive.
(g) Is 2 and 4 not met and 3 met? Dive if able.
(h) After all checks above is 3 not met? Dive/climb in direction of target.

I have a few issues still. One is immelmann. So what I do if I dove or climbed right before is go dive/climb, straight, immelmann, acrobatic pilot. Lol.

This is my rough draft of it. Still working on it.

-------------------------------------------

Tonight, I decided to try this out but without evaluating on a straight, but instead using the first revealed side slip. It actually worked better. Less thinking during the Immelmann. And took care of another issue of evaluating so often. It's not that the AI would even climb so much, but because you are doing this part outside the app, it's nicer if you're not evaluating so often, but yet, still have the planes use elevation.

My biggest problem right now is how the checks are done is very similar to how I play against players. So, the AI and I stay at the same level a lot of the time. Usually only thrown off by Immelmanns.

The longest streak of us being on different elevations was two turns.

I feel pretty set on looking at using the first side slip as the climb/dive, so now I'm going to move on to some other tests.

1. I'm going to try to throw it off. I'm going to do crazy elevation, and not play my style, and see what happens.

2. I'm going to play planes with different climb rates. Which right now, I've been testing with two climb rate 3's.

3. The final test will be adding more planes and see if I've perfected it.

Figured I'd post my progress here for those interested.

**flash** · 08-18-2020 02:40

Interesting project Derrick.
Think you need to insert 'straight', or, 'sideslip', whichever you decide to use, rather than 'nothing' for clarity.
Sideslip would seem the way to go for you. It doesn't complicate the Immel (and you can't play a climb/dive in the turn you play the immel card if you follow the rules)
Ending up at the same height is very often the result of using any alt rules, which allows longer range shots, so what you're finding is not unusual.
Now, going through the options I don't understand this one:
Is 1 and 3 met and 4 is not met? - Climb.... if enemy is in your arc/range at the same level why would you climb to gain a climb token and possibly compromise your shooting range ?
Introducing altitude becomes a complicated beast the more you think about it.

**Derrick** · 08-18-2020 05:44

Thanks Flash. I edited my post for clarity as you suggested.

If 4 is not met means if you are a climb rate 3, you have at max 1 token. So if 4 is ever not met, most the time it is climbing. That's just the way I would play it. But also, the AI is not ever considering how many tokens you have. Just if you are not at the same level. So this gets it prepared for moving levels if it has to. Yes, an immelmann throws it off. But I haven't found that to be an issue yet. Maybe just not enough tests.

One intereesting thing to note. The side slip method only caused one collision. And one time, replacing the sideslip prevented collision.

The straights seemed to cause more collisions.

**flash** · 08-18-2020 08:00

The point I'm stuck on Derrick is that if the AI has it's enemy just in range and arc, condition 1 is met; if they are at the same altitude, condition 3 is met. Your metric says if not at max climb counters - the AI must now climb....
The climb card is the length of a stall; if the enemy is moving away from the AI & goes straight, or, sideslips the chances are it will move out of range of your AI's guns in this instance.
In a similar position you, I & everyone else would forego the climb to take the shot - I see no reason why the AI would not do the same.
Maybe there needs to be a closing/moving away consideration in these decisions ?

**Derrick** · 08-18-2020 08:45

That is a good point. It has not come up. But that is because, typically, the AI had been loaded up before that scenario happens. On a climb rate 3, one immelmann gets you halfway there.

So I think I need to play with planes with bad climb rates to see how often this happens.

I will say, there are times when the solo plays sub-optimally just using 2D rules. Here yeah, might not be great but they are prepped going forward. But they could've just as well made a bad move.

I really need to try different climb rates to see how it works. Maybe I'll try that next.

I did realize though from analyzing #4 that I need to say unless at highest level. Because AI could immelmann themselves too high. This will give them buffer and keep the player from "correcting" AI's course. Which correcting course is not new. Because the edge of the mat.

Edited my original post to add that part.

I'll be watching for your scenario with more attention Flash and give a report on what I find.

**flash** · 08-18-2020 11:09

You'll only get there by thorough play testing Derrick, using all types, in all attitudes of attack/defence. I'm sure this will be watched with interest by users of the app and solo charts.

**Flying Officer Kyte** · 08-18-2020 12:24

Thanks for putting me on to this thread Derrick I can now follow it without us disturbing the flof the other thread.
Rob.

**Derrick** · 08-18-2020 21:11

First, I edited the first post by putting a letter before each step to easily tell you how far down the chain the eval went. But now, I kinda want to keep testing until all evals are done with sideslips. It just happened more often with straights.

So now on to the stats of the 2nd test with sideslips triggering the eval.

I flew a Nieuport 11 with Higher Machine Guns. Because I like Higher Machine Guns. (Climb Rate 4)
AI flew a Fokker DrI. (Climb Rate 2)

We went 17 turns. And only one sideslip to eval the Altitude. It went all the way down to (h). The Fokker Dr1 dove to get on the same level as me.

The Fokker won. 10 cards equaling 11 damage.
The Nieuport went down with 6 cards equaling 10 damage.

It would not have been close if not for the Higher Machine Guns which triggered twice.

There was only 1 collision.

The story of the night was Immelmanns, the Fokker Immelmanned 8 out of 17 turns. That... is a lot of tokens, especially for a climb rate 2. As the Nieuport, I pretty much kept at my altitude: S-Splitting and Immelmanning, climbing, hovering with 2-3 tokens most the game. I had to eventually come up, because the Fokker kept going up.

The longest streak of being at different levels was 4 turns. But we were at different altitudes for a total of 9 turns. We were never two levels away.

The Fokker missed 2 key and obvious shots because it Immelmanned instead of S-Split, which yeah, I don't have S-Split set up for AI. However, if you wanted, you could do two things to make this work a little more intelligently:

1) If you use the charts, you see the Immelmann, just switch it out with an S-Split if #3 would be broken or the target is below you.
2) If you use the app, if you see a straight, you have to reveal the next maneuver until you don't see a straight and take note of them. If an Immelmann pops up, play the straights up til the needed S-Split if # 3 would be broken or the target is below you.

I don't use S-Splits for AI because I like how the app reveals itself. When playing several planes, it helps me keep track where I'm at. It didn't super matter for the Fokker. The Fokker did win afterall.

The major oddity here is that the Fokker only pulled one sideslip. 1 turn out of 17. Because of this and all the darn Immelmanns and being a climb rate 2, the Fokker did Immelmann with max level and tokens.

So new rule: If a plane does Immelmann with max level and tokens, replace Immelmann with a dive. This does take the plane way out of the fight, so I'll have to see how it works with more planes, but it works, because the plane kinda resets when it comes after you.

I was tempted to tweak this a little further by saying eval on the first sideslip OR stall, but that would've only been one more time. But the Fokker would've Dove and spared the corrected maneuver on the Immelmann discussed above. But, I can't do that until I try it with a Phonix, because that thing will be using altitude like every turn lol.

That's what I have for now. If you have questions, let me know, I took pretty detailed notes.

Also, I'm saying it now, my measure of success will be if I can play 2 planes on 2 plane photo recon with altitude, and really feel like it works.

**Derrick** · 08-19-2020 16:54

Gonna have to do this one again... I was playing around with the app and just testing to see how many side slips show up. Then I hit D+, and I'm confident this would have played better with D+.

Not sure what D+ is. Gonna play a full game with it next.

**hokusai** · 08-20-2020 03:03

Your work progresses well!
Just one question. Once you find the optimal rules do you plan to add unpredictability by allowing the AI to do "wrong" moves? Or do you think the app does already enough in this respect?

**Derrick** · 08-20-2020 06:27

Originally Posted by hokusai

Your work progresses well!
Just one question. Once you find the optimal rules do you plan to add unpredictability by allowing the AI to do "wrong" moves? Or do you think the app does already enough in this respect?

Thanks for the question. It is predictable in the sense that you know it will climb more than it dives. However, sometimes it's climb is not advantageous, so it can catch you there. This AI plays like me for the most part, so like, I know what I would do... lol.

It was more predictable when I was using straights. Because it evaluated more often. But with sideslips, you don't know WHEN it will climb/dive.

This last match was super predictable, but because immelmann turns. I'm going to play with D+ and watch out for predictability. In the last game, I wanted to hover just below it. And easily achieved it.

It would be less predictable while still being reasonable if I randomized Immelmanns and S-splits. If you are playing charts just flip a coin everytime it lands on immelmann. However, I'm sticking with Immelmanns because I want to scale it, and the app helps me track better than the charts, so I flip the cards as they should be and not all at once.

You don't want it too random as the altitude can really make planes get away from each other. This AI suggestion comes from a BGG post I wrote kinda instructing people how to play altitude and have it be fun. (Because I had been hearing hate on altitude.)

I love altitude so much that I'm also writing up a variant altitude rules set for T&T, but that will now take forever to finish because I need two people for that.

**hokusai** · 08-20-2020 10:12

Originally Posted by Derrick

you don't know WHEN it will climb/dive.

Right. Was not aware of this.

In fact your work inspires me to try something similar for my variant of WW2 altitude-house rules which I have no chance to play test with players. I could run the WW2-app in no-altitude mode combined with my externally defined rules....hmmm...need to think

**Derrick** · 08-21-2020 06:27

Played the same match up. And I don't know what the D+ is, but it is not meant for my Fokker. It tried to make some illegal moves and had cards I didn't have. So I subbed the next best thing which actually helped it.

For what it is worth, here are the stats.

13 rounds.
Evaluated for altitude 7 times: h, h, e, h, h, g, b

It could not dive on g so it used the sideslip

The b was not what I now call "Flash's Scenario" the planes were a bit more perpendicular. I believe the climb actually helped it as the sideslip would've overlapped it. I should have taken a pic. It is probably hard to imagine it.

No collisions. No higher machine gun shots.

Longest streak of being at different levels: 3 turns

Total turns at different levels: 5 turns

Fokker had 9 cards for 11 damage.
Nieuport had 6 cards for 11 damage.

This method of evaluating mid turn might make the AI too smart. Especially, in the beginning. Unless you close your turn with the climb/dive. Because of this, I'm starting to think I should just do charts of my own... but I wouldn't know where to start... can I just edit someone else's? I figure on the Blackronin/Peter L. charts 1 could equal near in 12 or 1. And 2 could equal not this combo: closing, near not in 12 or 1. Then people would only have to answer 2 more questions. Or I could use the very basic charts and have separate charts for each of the questions above (a, b, etc.)

Another side of me says that it is good it has an advantage, because when planes are equal, I usually win if I don't pull explosions.

**flash** · 08-21-2020 06:52

The AI should be as smart as you Derrick, as you should put the I in AI and do what is best for the AI machine in that moment.
My approach is much simpler, if the enemy is above the AI will gain Alt, if below it will lose Alt, especially to get off a shot. Makes eval much simpler !

**Derrick** · 08-21-2020 08:05

Originally Posted by flash

The AI should be as smart as you Derrick, as you should put the I in AI and do what is best for the AI machine in that moment.
My approach is much simpler, if the enemy is above the AI will gain Alt, if below it will lose Alt, especially to get off a shot. Makes eval much simpler !

I will be the I to make sure it does not go off the mat, fly higher than is allowed, and to redirect its course to an obvious move during a photo recon. But otherwise, some "bad" moves makes it unpredictable.

My method allows it not only to chase you, but force you to chase it in respect to altitude. But as you can see from the stats above, mostly this method is what you are doing.

Edit just to explain why: if it evals to an h, it basically is just going up or down based on your altitude and that evaluation gets hit the most.

**Derrick** · 08-23-2020 14:49

My most recent failed experiment was trying to find a flawless way to give the AI unpredictable S splits.

**Derrick** · 08-27-2020 19:27

Not wanting to test out S Splits just yet, I decided to go

Me: Hanriot HD1
AI: UFAG (climb rate 5)

We were at the same level the whole game. UFAG collected 3 counters.

It evaluated e,c,e in the first 3 turns.

The game went 8 turns. I lost. There was one collision in there.

It kinda went as expected. I wouldn't expect a climb rate 5 to be bouncing all over the place. I was trying to force it into evaluating a f or g. But alas, I was up against a true Ace tonight.

**flash** · 08-28-2020 08:47

I don't know how you can make a Split-S selection flawless and I don't recall how often the Immel comes up in the App/other charts as I don't use them now but they tend to come up when a target is on the AI's six.
I would suggest that if the target is above the AI it could Immel up, (or, as many do for the WW1 Immel back to your start height).
If the target is below the AI then it could make that Immel a Split-S, particularly if the target's moving away.
If the target is on the same level you could flip for it after the target has planned its cards, particularly if the target is closing on it's six.

**Derrick** · 08-28-2020 10:52

Originally Posted by flash

I don't know how you can make a Split-S selection flawless and I don't recall how often the Immel comes up in the App/other charts as I don't use them now but they tend to come up when a target is on the AI's six.
I would suggest that if the target is above the AI it could Immel up, (or, as many do for the WW1 Immel back to your start height).
If the target is below the AI then it could make that Immel a Split-S, particularly if the target's moving away.
If the target is on the same level you could flip for it after the target has planned its cards, particularly if the target is closing on it's six.

In the advanced v2, Immelmanns come up a lot. Not just the 6/7. But I was thinking about randomizing a stall/straight, every time a straight comes up not after a immelmann or S split. It would not even have to be 50/50. You could take 2 straights and 1 stall in it's own deck and draw, shuffle.

The flaw is that this is not as smooth as where I'm at right now. While not absolutely perfect, the way I do it now requires no tracking as long as you consider the AI will do Daredevil on rare occurrences. But with adding this rule, I'd have to track more. Not an issue with one plane. But scaling it will take some of the joy out of it.

This is why I'm now considering doing my own charts. But I don't even know how to start a project like that (images and the logistics).

**flash** · 08-28-2020 12:42

Originally Posted by Derrick

I... But I don't even know how to start a project like that (images and the logistics).

Decide on the dice format that will get you what you want - I went with the D8 to get the options I needed - then spent a winter mapping out the options for each deck (5 per zone, 6 zones per deck) - matching up those with the same decks at different speeds - then spent a year play testing with OTT - then put it out to others - then put it in the files ! https://www.wingsofwar.org/forums/do...p?do=cat&id=63
Introducing RAW altitude will make things more complex for you perhaps.

**hokusai** · 08-28-2020 13:00

Originally Posted by Derrick

You could take 2 straights and 1 stall in it's own deck and draw, shuffle.

Wouldnt a die roll be a little smoother? I know, dice have a bad reputation on this forum but on occasional situations there is no need to be ideological.

Originally Posted by Derrick

This is why I'm now considering doing my own charts. But I don't even know how to start a project like that (images and the logistics)

Good idea.
Not sure I understand the problem. Just start with pencil and paper then you can play-test right away. Once the thing is reasonnably mature somebody proficient with photoshop or gimp may team up doing nice files.

**Derrick** · 08-28-2020 13:52

Thanks for both of your thoughts.

Flash, those charts are great. It inspired me. I think I have to go with a d6 though just to fit everything in.

But before I do any work on charts. I'm gonna nail down this project of using the app. I feel close because I'm not coming up with new moves but modifying existing moves.

Last two tests are Dawn Patrol 2v2 and photo recon.

I have a high interest in this games altitude because I'm more of a kitchen table gamer than a wargamer. I only have room for one mat. So altitude gives me more space to play with.