Enmity testing - Still Going!

[aurik]

Based on the data at lvl75, I know the precision of the modifier is at least 10 bit (2^10 = 1024). 9 bit (2^9 = 512) did not provide a value that fell into the modifier range at 75. The issue with this (as far as completeness) is that with 10 bit precision, I cannot find the exact modifier without a cure of ~1024 (my lvl 75 one was 1029), which is impossible at lower levels, so the precision isn't too great.

Unless I'm misunderstanding what you're saying, I don't think that's right, is it? If you wanted to see if it was, say, 744/1024 versus 745/1024, wouldn't a 300HP cure suffice? The former would yield 217 CE, and the latter 218.

Hmm... well this is 1 example I have...

On LVL75, the range is 0.7269 to 0.7279. X / 1024 must fall within this range or 10 bit precision isn't good enough. It turns out the only value that falls int he range is 745 / 1024 = 0.72754. If we look at the LVL70 value, a 418 cure led to a range of 0.7679 to 0.7703. There are 2 values which can fall in this range... 787/1024 = 0.76855 and 788/1024 = 0.76953.

Knowing that there are only 2 possible values, you should be able to construct a cure which will cause the difference in values -> 1 point difference in hate.

[Suiram]

You should find those values diverge at cures of 13, 26, 39, 52, 65, 78, 91, 104, 117, 121, 130, 134, (and there's a lot more than I expected so I'll stop there lol)

[Kaeko]

This is why I sucked at assembly programming.

[aurik]

This is why I sucked at assembly programming.

There is no such thing as good at assembly programming. Anyway, my formulas seem to be accurate enough to predict the value within a few 100ths between 0-50.

Need more data between 51-75.

[Kaeko]

This is why I sucked at assembly programming.

There is no such thing as good at assembly programming. Anyway, my formulas seem to be accurate enough to predict the value within a few 100ths between 0-50.

Need more data between 51-75.

Is there a reason you think there are cutoffs at 0-10 / 11-50 / 51-75 etc.? Or are you just trying to best fit the data with the fewest amount of formulas?

I think someone mentioned that 50 was the original level cap... not sure if that would change anything.

[Suiram]

On a related question, it would be interesting to see the hate generated by a 345 HP Cure to a Lv75 player. If it is calculated as Cured*745/1024 it should yield 251 CE, whereas if it is Cured*0.7275 it would be 250.

[Kaeko]

On a related question, it would be interesting to see the hate generated by a 345 HP Cure to a Lv75 player. If it is calculated as Cured*745/1024 it should yield 251 CE, whereas if it is Cured*0.7275 it would be 250.

I can do this for you tomorrow.

[aurik]

This is why I sucked at assembly programming.

There is no such thing as good at assembly programming. Anyway, my formulas seem to be accurate enough to predict the value within a few 100ths between 0-50.

Need more data between 51-75.

Is there a reason you think there are cutoffs at 0-10 / 11-50 / 51-75 etc.? Or are you just trying to best fit the data with the fewest amount of formulas?

I think someone mentioned that 50 was the original level cap... not sure if that would change anything.

Best fit, given the small amount of data. The precision of the fit of those formulas is too good to be random chance...given what we know to be true, I'd say there's a very high chance the formulas are generally the right form.


But yes, it takes 10 levels to halve, and then 40 more levels to halve again. Perhaps that's how they calculated it?

Bedtime, I'll check this topic tomorrow.

[Freysi]

This might be a stupid and insulting question but I didn't see it mentioned before.

Did you compensate for the 1CE that a 0hp cure gives? Did you confirm that a 1hp cure gives 2CE or does that 1CE maybe only happen when the cure is 0hp?

[Kaeko]

This might be a stupid and insulting question but I didn't see it mentioned before.

Did you compensate for the 1CE that a 0hp cure gives? Did you confirm that a 1hp cure gives 2CE or does that 1CE maybe only happen when the cure is 0hp?

At LVL75, Cure for 0, 1, or 2 HP all generates 1 CE. I verified that. This was definitely an important question.

The Cure for 0 HP is 1CE thing is just like a bottom cap. Everything has to generate at least 1 CE.

[Kaeko]

On a related question, it would be interesting to see the hate generated by a 345 HP Cure to a Lv75 player. If it is calculated as Cured*745/1024 it should yield 251 CE, whereas if it is Cured*0.7275 it would be 250.

I can do this for you tomorrow.

Turned out 250 CE for 345 HP Cured at LVL75...

Any ideas on this one? Appears 745/1024 isn't right unless I'm missing something in how the truncation works.

1489 / 2048 ? =/

[Suiram]

If it's a lookup table it could just be 0.7275, I guess? If you want to try a few other values on the same question 356 HP (258/259 CE), 367 HP (266/267), 378 (274/275), and 389 (282/283) also should cause differences between those two values. If it's the lower number in each case, that's more evidence for it just being 0.7275.

Edit: There are quite a few values for HP cured where 1489/2048 differs from 0.7275 in predicted CE generated, if you want one of those. What is a good range to find values to cure?

[Kaeko]

I can do anything up to 1292 HP on SCH.

Anything within 380 HP is the easiest though.

Ashi and I are out doing this right now if you have any other ideas.

[Suiram]

Try 385, 0.7275 predicts 280 CE, 1489/2048 predicts 279.

Edit unless by "within" you meant "strictly less than" in which case try 374 (272/271 CE).

[Suiram]

If all of those tests suggest that it's .7275, as one final thing try these last two:

389 HP cured: 0.7275 predicts 282 CE, 0.72751 (or higher) predicts 283
400 HP cured: 0.7275 predicts 291 CE, 0.72749 (or lower) predicts 290

All of this presupposes that the final stage in the calculation goes "HP cured" * "Function of level", though, although I guess that's as reasonable an assumption as any.

[Kaeko]

If all of those tests suggest that it's .7275, as one final thing try these last two:

389 HP cured: 0.7275 predicts 282 CE, 0.72751 (or higher) predicts 283
400 HP cured: 0.7275 predicts 291 CE, 0.72749 (or lower) predicts 290

All of this presupposes that the final stage in the calculation goes "HP cured" * "Function of level", though, although I guess that's as reasonable an assumption as any.

HP Cured = 385 | CE = 280 | It is not 1489 / 2024

HP Cured = 389 | CE = 282 | It is lower than 0.72751

HP Cured = 400 | CE = 290 | It is lower than 0.72749

Since it seems the # is below 0.72749, May be better to go from lower bound. In the 385 HP Cure, the lower bound is 0.72 repeating (0.727272727...).

Range is currently 0.72727 - 0.72749

0.72 repeating is 8 / 11 . Dunno if this matters.

[Suiram]

Interesting. Maybe at 75, CE = floor (HP cured / 1.375)?

Edit: If it is (Cured / F(Lv)) then for Lv70, HP/1.3 would also work from the numbers above.

[aurik]

that's an interesting thought: that the table is divisors instead of multipliers.

[Suiram]

It's working out really nicely if you assume it's divisors.

Assuming it's HP/F(Lv), then the following all work for the above data:

F(1) = 0.275
F(2) = 0.3
F(10) = 0.5
F(12) = 0.525
F(14) = 0.55
F(70) = 1.3
F(75) = 1.375


F(40) runs into a problem, as .875 should give 468 CE, so it'd be worth rechecking that, otherwise it could just be an order-of-calculations thing that could be ascertained later.

[Kaeko]

OK here is the current list of 100% sure correct mod ranges (couple of updates)...

Notice the values at 14-15 and 70-71...

LVL1 | HP Cured = 90 | CE = 327 | Mod Range = 3.6333 ~ 3.6444

LVL2 | HP Cured = 100 | CE = 333 | Mod Range = 3.3300 ~ 3.3400

LVL10 | HP Cured = 160 | CE = 320 | Mod Range = 2.0000 ~ 2.0063

LVL12 | HP Cured = 195 | CE = 371 | Mod Range = 1.9026 ~ 1.9077

LVL14 | HP Cured = 240 | CE = 436 | Mod Range = 1.8167 ~ 1.8208

LVL15 | HP Cured = 265 | CE = 481 | Mod Range = 1.8151 ~ 1.8189

LVL40 | HP Cured = 340 | CE = 388 | Mod Range = 1.1412 ~ 1.1441

LVL70 | HP Cured = 418 | CE = 321 | Mod Range = 0.7679 ~ 0.7703

LVL71 | HP Cured = 585 | CE = 450 | Mod Range = 0.7692 ~ 0.7709

LVL75 | HP Cured = 1029 | CE = 748 | Mod Range = 0.7269 ~ 0.7279

The values at 14 to 15 and 70 to 71 overlap. Equation change possibly? Seems like 14-15-16 , 69-70-71 might have the same CE mod... Trying to find those levels.