Hi,

I am a theory crafter at heart and definitley knows what happens to well formed plans when they get kicked in the balls by reality, but despite that I continue to tinker away with numbers and only spend a little time actually playing. Each to his own I guess.

So I need you guys to help me with some ideas and estimations and such, preferably from people who plays alot and have seen it all.

I have put together a little program that test two squadrons agains each other and it runs smoothly enough to try 100 000 jousts in a couple of seconds. That is enough data to get the accuracy level down to parts of a percent and that means that I can estimate what modifications and stats are worth cost wise in a 100 pnt build.

There are ofcourse some very serious limitation to my little program and I will get to them, but first this is what it can do (at the moment - I have a long ToDo list).

Each side can have three different ship types and total number of ships on each side is limited to 10.

I have no database of actual ships, so I have to put in the stats manually ( I have a few default values so it is pretty quick). The advantage of this is that I can try any type of stats on the ships.

I have a few upgrades coded, like Focus (for all), Target Lock, evade, Stealth Device, Upgraded Shields, etc. Just some of the easy ones. Upgraded Shields could just be simulated by increasing the actual shield attribute of course.

Other things can be worked around, with a little theory crafting. An ability that gives me an extra attack die once every three times or so could be simulated by adding two ships with the basic attack and a third copy of the ship with the increased attack, etc. This is a workaround until I have implemented that particular ability properly.

The first goal has been to be able to simulate basic shipbuilds without any modification and that is almost doable.

Once all the ship type's stats are typed in, and the number of each ship type I have two guestimated values:

FocusFire%:

This simulates the ability to attack the first ship in the enemy list (should be the highest value target). If this check fails it will apply half the FocusFire% chance to aquire a secondary target, then a third, etc until we run out of enemy ships and end up not having a target to shoot at.

At the moment I have this value at a default 50%, but it can be changed to simulate 360 degree arcs, or bad manouverability, etc.

Stress%:
This is the chance that the ship will be stressed every turn. If this happen it will not take an action, but will automatically get rid of the stress. This is set at 25% at the moment but can be adjusted to make up for different abilities, etc. A ship that is stressed will have it's FocusFire% halved for that round, since it pretends to make a green manouver by default to get rid of the stress.

There is no actual movement, since that is waaaaaaay to hard to simulate in a good manner without a very good AI module. Instead the squads just follow a range list that begins with range 3 then alternate between range 2 and 1 until it starts over after 8 rounds.

There is an action priority list with some simple rules like: If a ship doesn't have a current target, but is the target of another ship it will take an Evade action if it can, etc. Usually ships ends up taking Focus actions alot, but Target Locks were more useful than I thought at first.

Die rolling, etc are done according to the rules. Crits are handled correctly, but the effects of the crits are a basic 34% chance of causing an extra point of damage. No other ill effects from them.

So after adjusting the guestimates (I usually keep them the same for both squads since that will usually clearify the difference between the other settings) I start the simulation and get back a win rate for squad one in percent. Draws are excluded from the average. I also get the average number of turns each match up took.

I will say it again: I am not simulating any manouvering which means I am not simulating real games in any way or form. My goal is to see the effects of small adjustments and get answers to questions like:

What happens with my fictive win rate if I add a shield to one ship in my 8 ship TIE Swarm, instead of an extra hull? Everything else being the same.

How long does Howlrunner need to stay alive to justify her extra cost in a TIE Swarm. And how much will the number of ships effect this? This takes some theory crafting, tweaking and several runs to get an answer to, since I haven't directly implemented her ability yet.

What is 3 agility worth instead of 2 in a basic X-wing build?

How does 3 attack instead of 2 on one ship effect the outcome of a match with basic ships jousting.

What ships are the best and the worst in their basic configurations, not taking manouverability into account?

What ships are worth their basic price, not taking manouverability into account?

Etc...

So this is pretty much where I am now with this little project. So now my questions is, where to go from here? What abilities are the most important to simulate? Any ideas on how to "simulate" better manouverability besides the FocusFire% mechanic described above?

I will be happy to simulate any simple scenarios for you if you have specific questions or just want to test two pretty basic squadrons agains each other?

I will be posting some of my findings, and adjust them when my simulation improves. And answer any questions or requests of course.

So my first larger test:

I put together a tournament for basic ship lists. I tried to buy as many ship as I could afford for around 100pnts and padded one side or the other with an extra hull or the like to make the cost about the same. Then I sent them into the ring and got the results below.

There are a few very strong reasons as to why some of the ships placed higher or lower than expected. The Lambda for example, it has a terrible manouver dial (no k-turn and no 360 fire arc), but in this test that wasn't a factor (it "flew" just as good as any other ship)... The YT or Fire Spray didn't get a bonus for their increased fire arc. Ships with lots of possible modifications went in to the ring without any of them in use, etc. Cloaking ability wasn't a factor, etc.

So the only thing that one should take away from this is the following lesson: Two comparable shiptypes where one placed high and the other low on the list, pick the high placing one. If you pick the low placing one, make sure you use their modifications slots, cool abilities and synnergies or their better manouver dial if they have one or you will be playing with a sub par ship type. :-)

This listing will most likely change when I improve the simulator. This is mostly a teaser and not my final word on the matter. :-)

1. Lambda

2. TIE Fighter
3. TIE Interceptor
4. Z-95

5. FireSpray
6. X-Wing
7. E-Wing
8. A-Wing

9 TIE Bomber
10 TIE Phantom
11 B-Wing

12 TIE Advanced
13 TIE Defender
14 YT-1300
15 Y-Wing
16 HWK-290

And again do not take this too seriously. :-)

First off; you sir have pinned the needle on the nerdometer! Lol!

So if I get your findings right, a YT is not a great ship..it's the upgrade shenanigans that make it work? What does the program think of the current YT lists running X or A wing support? I'd be curious how it fares in your arena vs a 4 ptl interceptor build.

Yup, my nerdometer is pretty much busted at the moment. :-)

YT is a great ship, but in my current simulation there are three problems.

1. No upgrades, so running three naked Outer Rim Smugglers might not be the best idea. (Or four against, say 9 naked TIE fighters.)
2. 360 degree fire arc makes a huge difference, no need for k-turn stressing and no chance of enemies barrel rolling out of you arc. I haven't included any adjustment to the fireFocus values in the above list. When I try to simulate your request I will.

3. Manouver wise the YT is an odd duck and that is ofcourse not taken into account either.

So to your question:

I need a little more to go on. Many of the YT builds include alot of special mods and those are hard to simulate at the moment. I can create work arounds for some of them, but not all and not at the same time.

But give me two of the more simple builds and I will do my best.

I tried this set up:

2 X-wings
1 YT
2 A-Wings

vs.

5 Interceptors

Point wise it was like 103 vs. 90 and I adjusted for the 360 fire arc of the YT.

It ended up a fairly clear victory for the Interceptors.

I haven't implemented PTL that increases the chance for an extra action at the cost of stress and worse manouverability yet.

A naked YT isn't that good.

When I switched to 2 YT, X-wing, Y-wing the win rate dropped from 40ish to 16% so that was even worse. THe value in points were pretty close around 90 though for both teams.

So as long as PTL on the interceptors give at least equal value as the goodies on the YT I put my credits on the Interceptors.

Back to the coding! :-)

I've been wondering how long it would take for somebody to try to write code to find the best build possible!!! . I just didn't know if it was feasible!

I hope you are successful in this endeavor and am excited to see your findings! I think many of us will be pleasantly surprised with the results!

I have to say this is pretty cool!

I think there will be quite a few people interested in more data and hard numbers.
this is very interesting

Nice!

I have done some similar work, using average damage numbers averaged across the entire meta, to get jousting values for the ships. Some day I'll do a battle simulator using probability density functions (rather than the Monte Carlo simulation that you are doing).

http://community.fantasyflightgames.com/index.php?/topic/100360-using-lanchesters-square-law-to-predict-ships-jousting-values-and-fair-point-values-work-in-progress/

My action economy assumptions need some updating / validating, but it's still pretty solid. Your results should match up pretty evenly with mine if we use similar assumptions.

Your range bin method is interesting. It results in [37.5% 50% 12.5%] for ranges [1 2 3]. I would suggest just randomly selecting a range bin for each ship each round, having a different weighting for the first round (emphasizing longer range), and then the standard range bins after. You can check my thread for a baseline of ranges, I recorded all the shots in the 2013 Worlds Final Match. Again, my range & action economy assumptions need some updating / verification, but they should be pretty good.

I would also suggest trying to run it with a simpler action economy and see what happens: if a ship has an action, always just focus. Or, better yet, make the option selectable so one side can use "focus only" option and the other side can do something more complicated. Taking an evade action and then assuming that the ship will get shot at anyway is not a fair assumption though. A player may decide to focus fire down a different ship. Especially if you have multiple 3 dice attacks going against a single TIE, it is still not unlikely to kill a full health TIE anyway. so if you want to allow for Evade actions, you have to write additional logic to possibly let the attacker focus on a different ship.

It doesn't look like you break out 3/1/8/5 YT-1300 (named) vs. 2/1/6/4 YT-1300. They are essentially different ships. The ORS efficiency is ~62% and the named Falcon is ~70%.

The results look a little funny for a few ships.

• The B-wing should be WAY higher. It's ~20% more durable than the X-wing. It's jousting efficiency is almost as good as a TIE Fighter.
• Likewise the TIE Bomber is also very efficient if all you look at is its stat line.
• The Y-wing should also be much higher. It should at least be better than the TIE Advanced.

These could be due to:

• the way that you are matching up ships with each other and ranking them (most likely),
• not having enough samples (highly possible - see below)
• weird assumptions somewhere (maybe)
• or a latent coding bug (less likely).

If you run pure squad vs. pure squad, then you have 16! different scenarios to simulate: factorial(# of ships - 1), and you have 2 YT-1300 stat lines. That's about 21 trillion different battles. If you wanted to run each 10 times, that's 210 trillion simulations. So, the brute force method of comparing all possible matchups and then somehow ranking them isn't really feasible. I'm not sure what method you are using, exactly.

Also, margin of victory is very important, and the final ranking should ideally have some numerical weighting to it. For example the top 4 ships may be very tightly clustered, and the #5 ship is way less efficient (theoretical example).

How do you handle pilot skill? You either need to directly account for it by letting ships shoot first, or else back it out of their cost and let that affect the ranking.

Sorry if that all comes out like splatter painting. I don't mean to be critical, just providing constructive feedback.

I recommend reading up on my thread technical details, at least the Lanchester's section derived from differential equations. Especially important is the concept that doubling the number of troops increases your power by 4x, and conversely the cost of a unit should be proportional to the square root of its durability times its damage output.

Good luck!

Gah, just wrote an answer, even longer than your post MJ, did some simulations to confirm your ideas about the Bomber and the Y-Wing, etc. And then the computer ate my homework. :-( I guess I took too long in reply edit mode or some such.

Shorter posts from now on.

I will get back to this tomorrow, but I just wanted to say thanks for the links. Alot of interesting stuff to go through. I will try to steal as much as I can, like the range statistics and such data. I will also try to implement upgrades and pilots in the order of the most commonly used first according to your stat lists.

Your post was truly appreciated, but now I am swearing loudly at my computer and this forum's application coders or just lost sessions in general. :-( I will come back tomorrow after some sleep.

Oh, the Bombers performed better that the Y or adv. the Adv better than the Y's and the Z's beat the Bombers. IF anything the X-wing should switch places with the Bombers. The Y's looks like they should do better, but they don't.

Oh, the Bombers performed better that the Y or adv. the Adv better than the Y's and the Z's beat the Bombers. IF anything the X-wing should switch places with the Bombers. The Y's looks like they should do better, but they don't.

Stupid dog computer eating your homework!!!

So:

Z-95 > Bomber > TIE Advanced > Y-wing

In the specific matchup of TIE Advanced vs. Y-wing, the TIE Advanced is at a relative advantage at having 3 agility. 2 attack dice has a hard time punching through 3 agility. So even though in general the TIE Advanced is worse than the naked Y-wing (albeit not by much), in this specific matchup it skews the balance as far as it can in favor of the Advanced. It could explain why the Y-wing loses.

Did you try X-wing vs. TIE Bomber? If you really want to make it even points, try 21 TIE Bombers and 16 X-wings, and see what happens.

...I wonder. Could this be used as a way to rate lists?

If so, is there a way to account for turns and so on?

Hi again guys, back from a good nights sleep. :-)
Cptn this could definitely be a way to rate lists, but only basic ones at the moment.

First I need to implement more upgrades and some of the more common pilots or find work arounds to simulating their powers and abilities by adjusting other factors. And secondly I need a benchmark squad that I test everyone against to get a base line.

And yes I do take Turns into account, I even get the average amount of turns each fight takes. Some match ups lead to very quick fights and when two defensive ship types battle it out it can take 20+ turns on average. I usually run 100000 jousts to get rid of most of the variance due to lucky rolls. My guestimates and assumptions doesn't improve though, but since they are equally wrong over both sides and all the jousts, I can still find the difference between adding a shield over a hull to one ship out of 10. And better detail than so is not necessary.

Today I will try to implement Barrel Roll and Boost, by using some sort of OutManouvering value that gives each ship a chance to sacrifice their action and in return not get shot at by an enemy (but they still get to shoot if they have a target themselves). The chance of this happening should be fairly low and then increase with the access of the BR and B action. I was also thinking of steealing MJ's dial groupings and implement them in the same fashion. The better your dial type the better chance of escaping an enemies fire arc.

When this is done I can simulate all of the basic ship types and will redo my Jousting list based on winning percent against a couple of basic opponent lists. This should give way better results than an elimination tournament where they randomly face off against each other and there is no difference in winning by 51% or 85%.

Now back to some of the points that MJ made yesterday:

At the moment the range bin, as you call it, is a list (set in the preferences) of ranges for ten turns, and then it starts over. At the moment the default list is 3212321212 (and round 11 uses the first value again) which is now weighted similar to your torunament findings. Asteroids have a 10% chance of showing up on every shot taken, no matter the range. This % is also set in preferences.

And I also found a bug where range 1 actually gave +1 dodge die instead of +1 attack. 8-)

When I implement secondary weapons, range will become more of an issue.

I will not go for a simpler action economy. Rather the opposite actually. I try to do it as realistic as possible. The rules I use are good ones, I will use the same ones for my face to face games. Barrel Roll and Boost are not yet implemented but I will try to find a reasonable bonus to the Outmanouvering chance so they will show up as an asset in the win/loose margin for ships that have them. And then compare that value with annecdotal evidence from real games to see how well they correlate. IE How often do people take a BR or Boost and slip out of a fire arc, and still keep a valid target in their own sights and the enemy ship can't just shoot at some other ship instead. I don't think it happens that often, especially in dogfights with many ships. I will only use the BR and B when these conditions are met. Slipping out of a fire arc and then not be able to shoot your self or save one ship from taking fire just to have another of your ships fired upon instead might be valid tactics in real games, but in a simulation it is too many variables. So BR and Boost will be used less frequently in the sim, but have more of a guaranteed effect.

So far the actions Focus, Evade and Target Lock are all usefull, and all of them have value, more equal than I thought (at least TL and Evade are similar - everyone have Focus ofcourse).

SO a quick comparison of TIE Bomber and X-wing. The current number of ships on each side is limited to 10 (might fix that later but large numbers skews the results more than giving un underdog an extra hull or shield).

3X vs 4Bombers = 42% Win for the X-wing, so the Bombers are stronger and with more ships, 6 vs 8 even more so, only 35% win rate for the X-wings. This is probably due to the 1,3,6,10 power relationship, but also due to more ship equals more rolled dice and that favours the favourite, since luck will be less of a factor in each match up.

Wise from previous experience I cut this post short and continue on a new one. :-)

MJ: I am so new to the game and have played so few matches that I didn't even realize the two YT was actually two different base types. I will look into that. I have only played once against them and never with them. I have probably spent 100 times more time on theory crafting and another 100 times more time on coding than actually playing, hehe. I only have like 10 games or so under my belt. Hell I've seen more alot more games on YouTube than I have actually played myself. :-) That doesn't stop me from having fun with the game in my own way hehe. My X-Wing dealing friend is coming back to town tonight so it might be room for a real game again, yay!

So back to some of your questions. Pilot Skill. I looked over the tournament rules and try to go with their version on how to handle the turn order (random not imperial has initiative). At the moment I shoot in the order of PS. If they have the same skill the eliminated pilots get to shoot back. When it comes to "movement/action" I do it in a random order, since it is just a bunch of formulas/rules anyways. First they aquire a target, then basen on that, each ship takes actions. This favors low PS ships, since in a real game they would not know if they would actually be targeted by higher PS pilots when they must choose their action, but in the sim they do know. On the other hand there is no advantage to low PS pilots being able to move and force a collision from a higher skilled pilot either, so I let those two advantages cancel out. And more often than not you can figure out if you will become a target or not even if you move first.

PS will be hard to put a price tag on, since there is no actual difference between 2 and 12 if the enemy has 1. I will try to give the base line opposition ships different PS so it will give some weight to it in the jousting lists.

And then again I will take your advice MJ and skip the elimination tournament format when I create the next list and instead go by a set opposition 1-3 standard/mixed squads should do it, and let all of the pure lists fight them and compare the win% instead. I hope to get a new list together later today after the Barrel Roll Boost addition (and maybe Dial-typ adjustment).

Back to the coding!

And again thanks for all your numbers in your posts, a wealth of information. I read some of Vorpal Sword's stuff as well. I am glad there are other people with actual math knowledge to keep me grounded. *Bow*

Hmm, I found something strange. First I thought it to be a coding bug, but now I am not so sure.

I tested a ?,3,3,3,3 TIE Def versus two 2,2,2,6 Bombers. And varied the Defenders PS up and down to see how much difference it made in the win/loss ratio and it showd that shooting first was actually a disadvantage by quite alot!?!

When I increased the Stress% to 100, no actions were taken and the result shifted back to what was expected, about a 1.5% increased win ratio, per step, going from PS1, to 2 and then to 3. The defenders still loose, but still, with a smaller margin.

So this means that it is bad going first if you get to take an action, and that must be due to Focus being used on attack instead of defence. So there must be some sort of rule on when to boost your attack over saving it for defence. Something like:

Use focus if you can go from x hits to x+y hits where x and y depends on the relation between expected return fire odds and current defender numberOfDice. So if you are about to get creamed by the return fire, do not use focus to go from 0 hits to 1 hit against an agility 3 ship at long range behind an asteroid.

At the moment my ships freely spends their Focus if they can. The only limit is that they will not spend a Focus token on defence if it is not needed. But when you roll your attack you don't know if it is needed or not, since the defender hasn't rolled their dice yet.

Any thoughts on a rule of thumb that I can apply to this? Like only boost an attack if you can get to at least two hits+crits (if you are targeted by a ship that hasn't had their shot yet). Or three rules depending on if your own attack is higher, equal or lower than your agaility or the enemy's attack or agaility?

This is not an insignificant question. Equal and then Higher PS can give 1.5%+1.5% bette win rate. But misuse of your Focus is worth like twice that!

Summary, under what circumstance, if ever, would you change a zero hit roll to a 1 hit roll with a Focus token? A one hit roll to a two hit roll?

Oh, I certainly agree that fully modeling the action economy is better than simply "taking focus" every round. I was just suggesting doing some 100 point squad simulations with focus as the only action, as a comparison against my method, since those were the assumptions I used.

SO a quick comparison of TIE Bomber and X-wing. 3X vs 4Bombers = 42% Win for the X-wing, so the Bombers are stronger and with more ships, 6 vs 8 even more so, only 35% win rate for the X-wings. This is probably due to the 1,3,6,10 power relationship, but also due to more ship equals more rolled dice and that favours the favourite, since luck will be less of a factor in each match up.

4 bombers = 64 points and 3 X-wings = 63 points, so that's about as close as you can get points wise. The X-wing jousting efficiency is somewhere around 92%, vs. 97.5% for the TIE Bomber, so I would expect the bombers to win more often. This is very interesting because it puts numbers to it though, and shows how much a ~5% difference in jousting efficiency makes. The stronger ships (X-wing in this case) lose their advantage relative to square root law when the ship count increases, so it also makes sense that the X-wings lose more when you double the ship count.

Hmm, I found something strange. First I thought it to be a coding bug, but now I am not so sure.

I tested a ?,3,3,3,3 TIE Def versus two 2,2,2,6 Bombers. And varied the Defenders PS up and down to see how much difference it made in the win/loss ratio and it showd that shooting first was actually a disadvantage by quite alot!?!

When I increased the Stress% to 100, no actions were taken and the result shifted back to what was expected, about a 1.5% increased win ratio, per step, going from PS1, to 2 and then to 3. The defenders still loose, but still, with a smaller margin.

So this means that it is bad going first if you get to take an action, and that must be due to Focus being used on attack instead of defence. So there must be some sort of rule on when to boost your attack over saving it for defence. Something like:

Use focus if you can go from x hits to x+y hits where x and y depends on the relation between expected return fire odds and current defender numberOfDice. So if you are about to get creamed by the return fire, do not use focus to go from 0 hits to 1 hit against an agility 3 ship at long range behind an asteroid.

At the moment my ships freely spends their Focus if they can. The only limit is that they will not spend a Focus token on defence if it is not needed. But when you roll your attack you don't know if it is needed or not, since the defender hasn't rolled their dice yet.

Any thoughts on a rule of thumb that I can apply to this? Like only boost an attack if you can get to at least two hits+crits (if you are targeted by a ship that hasn't had their shot yet). Or three rules depending on if your own attack is higher, equal or lower than your agaility or the enemy's attack or agaility?

This is not an insignificant question. Equal and then Higher PS can give 1.5%+1.5% bette win rate. But misuse of your Focus is worth like twice that!

Using focus on offense is huge. It is statistically more likely to cause more damage on offense by using a focus than the damage you prevent on defense by spending a focus. But that doesn't meant that you should not spend your focus on defense! The conditional probability of how much damage you will do with a focus on offense, vs. the guaranteed damage you are preventing with defense, will be different. It's almost certainly still worth spending your focus on defense. Unless you are at range 1 attacking with 4 dice, then saving the focus on offense might still be worth it... you would have to run some average expected damage numbers. In general this is an area with an open question that is worth looking into. It's always a core question... spend the focus on defense or hold onto it? My philosophy is that if spending focus helps you, always spend it, because you don't know if you will be able to use it later. But that could be sub optimal in a few situations.

The real issue here is that you are not simulating a battle at 100 points. The main advantage of higher PS is the ability to destroy a ship before it fires, so if you only have 1 v 2 this is less of an issue. If you are 8 v 3, then the Defenders can probably knock out a TIE every round if they are higher PS. That immediately reduces the opposing side's damage output, which is huge. It becomes a compounding snowball effect. Granted, TIE Defenders are pretty bad at pure jousting value, so they are not going to compare favorably against pretty much any ship, unless you can somehow model what a white K-turn does for them.

So basically, I would try and run simulations with close to 100 point battles as much as possible. Weird things happen at lower point values compared to what we are used to seeing. For example Luke + R2-D2 is extremely strong against 2 TIE Fighters (even at equal point cost), but in a 100 point game he can get focused down quickly.

Back at the key board. Spent the night gaming, but alas, not X-Wing. I guess you can't have everything, hehe. A Study in Emerald was interesting as well. :-)

The range bonus bug might definitely have made low attack squads vs high agility squads have a problem. With attack one or two you really need that range one boost. But at least I found it hehe.

I am still implementing alot of new nifty functionality, so testing, verifying and double checking everything hasn't been on top of the list. At least I know have all the basic ships hardcoded so I can chose them from a list, etc. Simulations will take less effort on my part so I might run more different variations. It got tedious in the long run to type in every value manually before each test run.

Top of the list at the moment is to finish the Outmanouver implementation combined with your values for dials. I put them into five categories and then I will give the white K-turn and the lack of a k-turn their own Bonuses (lie a free upgrade for those ships that have it or lacks one). A white K-turn will for example lower the chance of stress more than it will outmanouver an enemy, etc. And I also gave every ship a basic chance of avoiding enemy fire, so that I can raise and lower that value when I start to implement other modifications. I haven't seen how it will turn out yet, but I will get back to you with a few comparisons to see if my guestimates are ok.

After that I will create ~100pnt squads (pad them with some extra hull or shields if needed) and run them all against three basic diversed squads and collect the win%-margin. Help on the composition of those testbench squads would be appreciated. 8 TIE swarm should probably be one of them, and then two mixed squads, one defensive and one very aggressive would probably be the best benchmarks. I will also vary their PS. These three test squads need not be correct rules wise, they might even be completely fictive ships as long as they give a reasonable balance.

And the next step would be to run all pure squads against each other... Need an excel sheet for that I guess. Every such fight would probably take at least 1 minute to do, 30 sek to enter values, 15 seconds to run and 15 seconds to enter the result... So it will probably take 5-6h. Hmm... that might be overdoing it though. I will probably test a few and see how fast I can run each match up.

Other than that I will try to implement some sort of approximation for Cloak, since it is a basic ship ability. And fix the Focus-use question. After that I will start on the modifications and pilot abilities, one by one in the order of how often they got to see use in your statistical data. Phew...

Somewhere along the line I should probably start to figure out the relative values, cost in points, of everything, but I hope to get some help with that. The simulator is pretty user friendly as it is now. I need some more tool tips and such and I can let the forum help with the calculations and the tests. :-)

After that I will create ~100pnt squads (pad them with some extra hull or shields if needed) and run them all against three basic diversed squads and collect the win%-margin. Help on the composition of those testbench squads would be appreciated. 8 TIE swarm should probably be one of them, and then two mixed squads, one defensive and one very aggressive would probably be the best benchmarks. I will also vary their PS. These three test squads need not be correct rules wise, they might even be completely fictive ships as long as they give a reasonable balance.

And the next step would be to run all pure squads against each other... Need an excel sheet for that I guess. Every such fight would probably take at least 1 minute to do, 30 sek to enter values, 15 seconds to run and 15 seconds to enter the result... So it will probably take 5-6h. Hmm... that might be overdoing it though. I will probably test a few and see how fast I can run each match up.

How long it takes depends on how many pure squads you want to compare against each other. If you want to do every single possible permutation, that is 20,922,789,888,000 combinations.

If you run pure squad vs. pure squad, then you have 16! different scenarios to simulate: factorial(# of ships - 1), and you have 2 YT-1300 stat lines. That's about 21 trillion different battles. If you wanted to run each 10 times, that's 210 trillion simulations. So, the brute force method of comparing all possible matchups and then somehow ranking them isn't really feasible.

I was thinking more along the lines of only 100pnt pure squads against each other and that would be (16x16)-16 (no need to test the squads against them selves. And the first test would be 16 x 3(base squads).

My program is not set up to create permutations on testing lists and then test them and then record the answer and compare many different results. When I am done being able to test one set up at a time that might be the next step (like random set up paired off against a benchmark squad and keep trying to find the best one). But that has never really been my goal. I just wanted empirical tests to get some data for real cost and a basic understanding about what makes a good ship good. Then I want to use that info to build efficient squads and test those against each other on a case by case basis.

But then again a brute force approach would be nice hehe.

PROBLEM:

I found a bug of sorts, that got me stuck most of the day. :-(

I am beginning to think it might not be a bug at all, just a very strange side effect of how the game plays out.

I have tried around with different coding rules for when you should save your focus token during your attack for better defence later.

So far it goes like this:

An agressive ship with lots of attack face off against a copy of the squad of about 5 ships.

It is definitely better to have the higher pilot skill and shoot first.

A defensive game where both sides are equal, pilot skills seems to be reversed.

The higher your agility and the enemy's agility is, the better it is to wait with your focus use and use if for defence instead. Sure three eyes on attack are probably always worth it, but one or two eyes (with out without other hits or crits to complement it)? Not so much, it turns out.

The only reason for this must be that using a token on defence is a sure deal, you know if your expenditure will yield less damage to you. While on attack, you know that the damage potential can be increased by x amount. But you never know if the enemy defence roll will come up all blanks and you over kill him or if he will dodge everything anyways. So the offensive use of the focus token ahs a high variance to it while the defensive spenditure is 100%.

Depending on how the focus token is used, the match can easily tip 5% or more either way. So this is kind of an important questions. It doesn't matter much if two equal squads face off against each other, but one defensive vs an offensive one or one low PS versus a hight PS and points are wasted because of sub par handling of the focus token.

OR this problem is an illusion and there is a bug in the code. :-(

Some say it is better to move first, so you can block other ships and force them to crash and loose actions. Maybe it is the same thing with shooting first. If your ships are tanky with high agility, it might be smarter go shoot last and mostly spend your focus tokens in defence boosting the higher number of dice rolls of agility compared to attack? And often you won't need the token for defence, so you can spend it on attack without a second thought, while the enemy do the opposite. They waste tokens on attacks that most of the time won't have an effect against your tanky ships and once you attack back the enemy will be easier to hit. Further more, if you shoot last, you can aim at the ships that has spent their tokens while the enemy didn't have that luxury.

Any thoughts?

Is it reasonable for my simulator to show a win rate of 55-60% with a lower Pilot Skill versus an otherwise identical enemy squad? IF they are both tanky (low attack high agility) or is my code bugged?

I think it has merit! Totally my opinion, but it makes sense!

Is it reasonable for my simulator to show a win rate of 55-60% with a lower Pilot Skill versus an otherwise identical enemy squad? IF they are both tanky (low attack high agility) or is my code bugged?

Have you tried running it with "if using an eyeball is helpful then spend it" and 100% focus fire?

I don't think running head-to-head simulations is an effective indicator of efficacy. There are far more factors at play, especially in a tournament situation.

X-Wing is basically a game of arc-dodging, and it's hard for simulations to reflect that. Sure you can establish the base efficiency of each ship (of which the Z-95/TIE triumphs) but from there forward it's difficult to accurately simulate.

I don't think running head-to-head simulations is an effective indicator of efficacy. There are far more factors at play, especially in a tournament situation.

X-Wing is basically a game of arc-dodging, and it's hard for simulations to reflect that. Sure you can establish the base efficiency of each ship (of which the Z-95/TIE triumphs) but from there forward it's difficult to accurately simulate.

If you are trying to realistically simulate a battle, sure, it is virtually impossible to do unless you can actually create all the rules and make an AI that can play the game intelligently.

But still, a numerical approach like this has merit. Using numerical methods (see my link earlier in the thread), the TIE Fighter, Z-95, and B-wing are predicted to be the most efficient ships. How does that compare to reality? Well, in wave 4 Regionals, the ONLY generic pilots that have won on the final table have been TIE Fighters, Z-95s, and B-wings.

I have tried running it with different rules, in different ways with or without focus fire. I even re coded the the use of FocusFire to be able to both simulate 100% randomness in target selection, 100% focused fire on the priority ship and the "normal" of 1-99% of aquireing the prio target followed by half that due to stress, secondary tries when the first aquired failed and overkill secondary tries when the priority target got shot down before you got to shoot, etc. I even added Focus as a choice so you can remove it from ships, even though all have it as a default.

I have tried a bunch of different rules in using the eye(s) for attack, based on my agility, my agility compared to Attack, targets agility, number of target aiming at me that hasn't shot yet, etc. I ended up with this rule:

Don't spend focus on zero eyes.

Spend focus on two or more eyes.

Spend focus on one eye if you are not targeted by anybody.

Spend focus on one eye if your agility is lower than the number of enemies targeting you. (You will die anyways so better spend it on attack. And with a high agility you can handle more shots than with a low agility and the focus on defense will be worth more.)

On defense I calculate how many dodges are needed and only spend focus if they actually matter. There is no rules about holding back focus when it is only one damage that can be mitigated and I still have multipple enemies aiming at me. So if anything the rules for defensive use of the token is more basic.

And still, it is better to have a low PS if you have a high Agility or tankiness rating. If you are a glass cannony you should definitely go for higher PS though. :-)

So until someone can prove me wrong, I will just go with that result. :-)

And to you Darth Ruin: This is a very limited simulation, and only simulate the factors that has been coded into it. As long as you as a user is aware of the limitations it can give a lot of answers, but then again a lot of questions can not be answered at all and other answers can only be as good as the assumptions/guestimates that they are based on. But in my experience it is always easier to guestimate 10 related but smaller results, than go directly for the big summation. And with some luck erroneus guestimates can sometimes cancel each other out if there are several smaller ones.