AR and NR

[Corily]

Such formulae are unlikely to be anywhere near the real thing. What you get is simply mimicry of what the actual increase in landrate is pr. value, but the whole thing is still chance based, so in the end, you're not better off knowing that there may or may not be a 84% chance of you landing your next hit than you were knowing only that there's a pretty good chance that your next hit will land.

Alright, wtf is wrong with you?
Look, alot of us made big efforts here to get this working, you could start by at least taking a decent look at things first.

This graph/formula can be very handy, if you use it the right way. Should you be using it in a duel? ofcourse not. Should you use it to make a setup suiting you? Yes.
It shows wether it would be worth it perking NR1. It shows wether it is a good choice to go for CSS insead of CS. In fact, every endgame setup is ALL about balancing things out. It will be alot easier with this graph/formula.

If you don't want to know the graph because you think it's worthless, well then go ahead but don't post in this thread either.

[Ebag333]

There are more unknown factors in AO than there are in the reducable number of chances in a cardgame. They are far as I can tell rather different sizes on the surface, even if they're both games of statistics at the lower level.

There's actually less unknowns in AO because you have a very good rough idea of your opponents skills.

If an NT is trying to nuke you, would you assume that he has the same nanoskills as an NR2 agent?

If you're trying to land a debuff on a raging enfo, would you assume it would be as easy as on a soldier?

You have no idea what cards your opponent is holding. It could just as easily be an Ace as a 4. At the very least with AO you have a range for your opponent, with games of chance the cards are wide open to anyone.

As for the number of variables, there are three. AR Skill, Def Skill, and the defenders Agg/Def bar setting.

You should know at least one of those three for sure (if you don't, you have bigger troubles than I can help with ), and you had better have at least good guesses for the other two.

In a card game, you have 52+ variables, and you know only a handful (depending on number of players, decks, and game chosen). For a classic poker game (on the first hand), out of 54 variables you know 5. Not great odds.

[Ebag333]

Here is my initial Nano Skill AR calculator. Feel free to use it, but I make no claims as to it's accuracy beyond that the land rate should be accurate +-5%.

http://pvp.aodb.us/ARDEF/nano_calc.html

[Manaas]

wow, just shows how pretty worthless NR can be unless you can really push it up to the max like some profs can.

[srompu]

Something else that should be tested is whether a nano landing is calculated at the beginning or the end of a cast. If I get time I'll try and do some. It should be pretty easy to see just by swapping Agg/Def to either extreme.

[Ebag333]

wow, just shows how pretty worthless NR can be unless you can really push it up to the max like some profs can.

I wouldn't say worthless. I think a little bit of NR makes a much bigger difference than most people expect.

It's somewhat deceptive, because 50% Agg Def % (where AR = Def) doesn't sound like much. But it's much easier to get NR than it is to get nanoskills. Plenty of profs can get 2500+ NR....but most folks don't have 2500 nanoskills.

Something else that should be tested is whether a nano landing is calculated at the beginning or the end of a cast. If I get time I'll try and do some. It should be pretty easy to see just by swapping Agg/Def to either extreme.

Pretty sure it's at the beginning, but worth testing. Go for it.

[Corily]

Something else that should be tested is whether a nano landing is calculated at the beginning or the end of a cast. If I get time I'll try and do some. It should be pretty easy to see just by swapping Agg/Def to either extreme.

Should be at the beginning indeed. I noticed it many times when trying to SS myself under UBT procs and making a mistake on timing the aggdef bar.

[Ebag333]

So since I seem to have figured out Nano AR and NR, I've moved back to weapon AR.

And while plotting some of my data, I noticed something quite curious.

Attackers Agg/Def setting doesn't seem to matter.


I had previously thought that Attacker's Agg/Def did make a difference, but a relatively small one. However, I have too many tests where the attackers AR doesn't matter to believe this anymore.

This also makes sense because I know that with Nano AR, the attackers Agg/Def didn't make a difference, and both Nano AR and Weapon AR follow very similar curves.

This is a big bonus for me, because not only does it simply the equation (I just need one curve to calculate the effects of the agg/def bar, instead of two), I can now use all my different test scenarios for different attackers Agg/Def, and combine them all, effectively greatly increasing the number of test scenarios in each group.

[Ebag333]

So here's my first pass at weapon AR calculations.

http://pvp.aodb.us/ARDEF/weapon_calc.html

I'm actually pleasantly surprised at how well it came out. I added up the difference between the test scenarios and the calculated (for -100 agg/def), and ended up with a total of -1.54%.

So if one test was 1% higher than the calculated, and another was 2% lower, the average of the two would be 1% lower. Having an average of 1.54% lower is phenomenal.

[erozc]

So here's my first pass at weapon AR calculations.

http://pvp.aodb.us/ARDEF/weapon_calc.html

I'm actually pleasantly surprised at how well it came out. I added up the difference between the test scenarios and the calculated (for -100 agg/def), and ended up with a total of -1.54%.

So if one test was 1% higher than the calculated, and another was 2% lower, the average of the two would be 1% lower. Having an average of 1.54% lower is phenomenal.

It looks very nice. And I am impressed by the approach you've taken. I guess you've got custom log analyzers by now, in addition to a perpetual pecking bird hammering away on the '1' key on the keyboard (or a software alternative )

So far I'm puzzled at what your 'adjusted AR' and 'adjusted DEF' are. I imagine this is the % check of nanos, or weapons.

Finally, either it's just me or something is up with the weapon AR calculator. AAD and AAO seem to have ludicrously low effect on the diff AR, and consequently on the results. I smell a fish, but it might all be on your head - but still, 2000 AAO or AAD not having a noticeable effect at typical TL7 values of attack and defense skills?

[lordran]

So here's my first pass at weapon AR calculations.

http://pvp.aodb.us/ARDEF/weapon_calc.html

I'm actually pleasantly surprised at how well it came out. I added up the difference between the test scenarios and the calculated (for -100 agg/def), and ended up with a total of -1.54%.

So if one test was 1% higher than the calculated, and another was 2% lower, the average of the two would be 1% lower. Having an average of 1.54% lower is phenomenal.

Jees, if this is correct I might just need to delete my fixer. 1800 dodge/evade and maybe 2000 AAD and still only 48% chance to miss.
And that is when someone has 1700 AR. My 172 enfo can get over 2000....

[Mereditche]

AAD has no effect when you use his calculator. Something seems wrong in there.

[Gridpain]

1000 attack skill, 1000 def ... -100 agg/def ...

0 AAD and 1000 AAD have less than 1% difference in hit chance ?

are you sure ?

[Ebag333]

It looks very nice. And I am impressed by the approach you've taken. I guess you've got custom log analyzers by now, in addition to a perpetual pecking bird hammering away on the '1' key on the keyboard (or a software alternative )

Quit looking at my bird like that! He's mine! MINE, I tell you!

So far I'm puzzled at what your 'adjusted AR' and 'adjusted DEF' are. I imagine this is the % check of nanos, or weapons.

Not sure what you mean by my adjusted AR/Def. I do apply the % checks right away, so that might be what you're thinking of. Or do you mean the AR Dif %? That's just a simple formula I came up with to ensure that I would always have a number between 0 and 100 to describe the difference between the attacker and defenders skills.

AR Diff % = AR/(AR+DEF)

In my testing I found that 100 vs 200 gave the same results as 400 vs 800, 500 vs 1000, 1k vs 2k, etc, so I needed a way of boiling down all the examples above into a single number.

Finally, either it's just me or something is up with the weapon AR calculator. AAD and AAO seem to have ludicrously low effect on the diff AR, and consequently on the results. I smell a fish, but it might all be on your head - but still, 2000 AAO or AAD not having a noticeable effect at typical TL7 values of attack and defense skills?

Jees, if this is correct I might just need to delete my fixer. 1800 dodge/evade and maybe 2000 AAD and still only 48% chance to miss.
And that is when someone has 1700 AR. My 172 enfo can get over 2000....

AAD has no effect when you use his calculator. Something seems wrong in there.

1000 attack skill, 1000 def ... -100 agg/def ...

0 AAD and 1000 AAD have less than 1% difference in hit chance ?

are you sure ?

Sorry guys, stupid programming error. Evidentially AAD/AAO had 1/10th the effect it should have.

It should be fixed now. I've also added in a check for AAO/AAD as I've only tested with 100% check weapons, so I'm not sure if the attack/def check effects that or not. That's on my list of things to test (though I'm pretty sure it doesn't actually effect it).

(Plus I'm lazy and that was the fastest way to fix it. )

[Ebag333]

Some of you might find this useful. I went ahead and created a comparison page between my real life tests and the calculated results for weapon AR.

http://pvp.aodb.us/ARDEF/display.php

Note that if a calculated result is higher than 100%, the difference is not included in the total at the bottom. I have no way of seeing real life test results of 114% land rate, so those differences would throw off my average, so I count those as a 0 difference.

The total simply adds up all the differences and basically tells you how far (on average) my formula is off.

5.06%. Not bad.

(If anyone requests it I can do the same for my nano tests.)

[erozc]

Quit looking at my bird like that! He's mine! MINE, I tell you!

Just tell your bird I have some tasty bird seeds right over here

Not sure what you mean by my adjusted AR/Def. I do apply the % checks right away, so that might be what you're thinking of. Or do you mean the AR Dif %? That's just a simple formula I came up with to ensure that I would always have a number between 0 and 100 to describe the difference between the attacker and defenders skills.

AR Diff % = AR/(AR+DEF)

In my testing I found that 100 vs 200 gave the same results as 400 vs 800, 500 vs 1000, 1k vs 2k, etc, so I needed a way of boiling down all the examples above into a single number.

The diff formula was very easy to understand. However, in the next bit you talk about 'post adjustment' which I found to come out of the blue so to speak. See the bolded bits. I would imagine that it indeed is the % checks, but would like to know for sure?

L = Land Rate
D = Diff AR %
A = Attackers AR (Post adjustment)
N = Defenders Def (Post Adjustment)


L = 3*(A/(A+N))^2.4


Or perhaps a simpler way of look at it:

D = A/(A+N)
L = 3*D^2.4

Sorry guys, stupid programming error. Evidentially AAD/AAO had 1/10th the effect it should have.

It should be fixed now. I've also added in a check for AAO/AAD as I've only tested with 100% check weapons, so I'm not sure if the attack/def check effects that or not. That's on my list of things to test (though I'm pretty sure it doesn't actually effect it).

(Plus I'm lazy and that was the fastest way to fix it. )

I knew it! And I wonder why software people give testing and testers a bad name

For a less scientific (but more user-friendly) approach, you might want to cap the positive percentages at 100% and negative percentages at 0% in your calculator, too. Also, it would be really cool if the calculators had those 'points of uselessness' calc options for evades and NR. Both for what AR your evades will be useless for (100% hit) and what your AR will always hit.

But seriously great job there. Your approximation of the formula seems very valid, considering the circumstances. Of course, we could all just feed you lots of data points and you could manage a much more complex approximating function that is essentially the same, yes?

[Ebag333]

The diff formula was very easy to understand. However, in the next bit you talk about 'post adjustment' which I found to come out of the blue so to speak. See the bolded bits. I would imagine that it indeed is the % checks, but would like to know for sure?

Indeed it is the % checks. I know, not the best way of wording it, but I wrote that up in about 30 seconds.

For a less scientific (but more user-friendly) approach, you might want to cap the positive percentages at 100% and negative percentages at 0% in your calculator, too. Also, it would be really cool if the calculators had those 'points of uselessness' calc options for evades and NR. Both for what AR your evades will be useless for (100% hit) and what your AR will always hit.

I seriously considered that, but in the end decided against it as I find it more useful to see how far over (or under) you are. So if you're at a 114% land rate, you know that you could bring it under 100% with some (relatively) minor changes. If you're at 160% land rate...well, might as well forget it.

In reality the vast majority of people play at full def, and the AR difference isn't big enough to reach a 100% land rate.

Points of useful/lessness is a good idea. Shouldn't be too hard to calculate backwards.

But seriously great job there. Your approximation of the formula seems very valid, considering the circumstances. Of course, we could all just feed you lots of data points and you could manage a much more complex approximating function that is essentially the same, yes?

I could refine the formula a little more, but to refine it a lot more would require a very complex formula.

I'd rather have a simple formula that is off +-5% that is easily understood, than have a 30 variable one that is accurate +-3% that no one can understand or replicate.

Plus...I can't really imagine the dev's making it that complex. My biggest fear was that it'd be a tiered style of result, but fortunately it's just a simple curve.

Now if you want to feed me more data I'd love to get more so I can see about refining it even further.

[mr_road]

still a bug... the check box for AAO and AAD should probably default to 100 (and be in the same table row)

to make it work... put 100 in the check boxes below AAO and AAD.

from the formula... 1 AAD = 1 evade skill


how much variation was there in the amount of AAD used to give the same 'defense' skill ?

was there any pattern between the tests where the difference between calculated and observed hit rate was negative / positive ?

the cluster around Diff AR % of about 40% is a little worrying.

[Ebag333]

still a bug... the check box for AAO and AAD should probably default to 100 (and be in the same table row)

to make it work... put 100 in the check boxes below AAO and AAD.

I believe that they should just be 100 always, but I was lazy and put the box back in (since I copy/pasted from my nano calculator). But yes, it should default to 100.

from the formula... 1 AAD = 1 evade skill

how much variation was there in the amount of AAD used to give the same 'defense' skill ?

was there any pattern between the tests where the difference between calculated and observed hit rate was negative / positive ?

Yup, tested quite heavily. There were tests with pure weapon skill (no AAO), tests
where evade items were swapped for AAD items, etc.

Now all these were tested with a 100% check weapon, so it could be slightly different for a non-100% check weapon. But there's not many (any?) out there. (0% check weapons/nanos are special cases, by the way.)

the cluster around Diff AR % of about 40% is a little worrying.

What do you mean "cluster"?

[Lheann]

Ebag333,

What formula are you using for calculating Evade total for weapons. Over the years so much on how to get to the true totals has been discussed.

As you indicated most people fight at full def these days. I have been on and off working on some data for evades and I had been focusing on just the total def rating portion of the math.

Any insight you have would be great as to what that formula looks likes.