Soultrigger
Dec 4, 2012, 11:35 PM
So I am currently in the process of reverse engineering grinding. A huge misconception about grinding is that each risk that is displayed is consistently the same, which is not true. So for instance, one Danger could be 45% success whereas the next Danger can be 40%. (These numbers are arbitrary, but the point still stands)
I've compiled a Google Doc that covers the Success Rates:
View-Only: https://docs.google.com/spreadsheet/ccc?key=0AltnsB1gh03-dEFlM2lyaVZqX2dKbmNoRWFiQm1KTGc#gid=0
Calculator: https://docs.google.com/spreadsheet/ccc?key=0AltnsB1gh03-dFBFRUw0d29RRDZYWmxRYWJraXFTYXc&hl=en#gid=0
How to read Success Rate Chart:
-Each value has 4 Success Rates: Base/+5%/+10%/+20%
-C is Certain
-S is Safe
-W is Warning
-D is Danger
How to use the Success Rate Calculator:
-Light grayed box mean the Success Rate may have an additional +5%
-Dark gray boxes means the data has yet to be collected
==================================================
Known Facts:
These are things that, based on collected data, MUST be true:
-Certain = 100%
-Safe begins at 80% *
-Warning begins at 60% *
*Note: These assume intervals of 5%. May have an additional percent of up to, but not including, 5%.
==================================================
To Do:
-Acquire data for 4~6* weapons.
-Acquire data for 10* weapons.
-Acquire data for 9* units.
-Verify 1~3* weapons have identical success rates.
-Verify 4~6* weapons have identical success rates.
-Verify 7~8* units have identical success rates.
-Verify success rates remain the same after leveling innate ability.
==================================================
Issues:
It should be noted that there are 2 major problems with testing the success rates:
1. I do not have access to a 30% Grinder booster
2. Due to 15% not existing, this means the results can never be foolproof as gaps exist in the data
==================================================
Logical Assumptions:
With that said, there are many logical assumptions I can make based on patterns:
1. SEGA most likely implemented the success rates based on intervals of 5
2. SEGA most likely made the risks labels at multiples of 10 (and not inbetween these multiples)
To support my claim, look at the values for grinding a 1~3*:
............. 0 ... 5 .. 10 .. 15 .. 20 ..
+1: ...... C ... C ... C ... C ... C ...
+2: ...... C ... C ... C ... C ... C ...
+3: ...... C ... C ... C ... C ... C ...
+4: ...... S ... S ... S ... ? ... C ...
+5: ...... S ... S ... S ... ? ... C ...
+6: ...... W .. S ... S ... ? ... S ...
+7: ...... W .. W .. S ... ? ... S ...
+8: ...... W .. W .. W .. ? ... S ...
+9: ...... W .. W .. W .. ? ... S ...
+10: .... D ... W .. W .. ? ... W ...
You may notice a diagonal pattern if you fill in the ? with the right success rates, with the success rates shifting right every level. Based on this, I can make these conclusions:
-Success rates exist in intervals of 20%s
-Certain = 100%
-Safe = 80%~95%
-Warning = 60%~75%
-Danger = 5%~55%
Of course, it is very possible that the success rates for +4 and +5 as well as +8 and +9 are exactly the same. In the case of 7* weapons for example, the pattern best fits when only changing success rates after increments of 2 grinds. In any case, even if these assumptions are wrong, the error is simply within 5%.
Note that, in the Success Rate Calculator, the grayed boxes show the worst case assumption, but have potential for an additional 5% (or additional <10% if the 5% interval assumption is wrong as well).
==================================================
Failure Rates:
Another thing to consider is how the percentages of penalties work out. There are two possible ways this could be done:
1. The different Failure Rates is mixed with the Success Rate, making a total of 100% that allocates between penalties.
2. The Failure Rate is simply the remainder from the Success Rate, and does a second roll to determine penalty.
The distinction is important because:
1. if the first case is true, then the Failure Rates can be reasonably guessed, if not proven using a 30% booster
2. Grind Boosters and Grind Protects would work vastly differently in both cases.
In the case of the first scenario, here's an example of how it may work (for 1~3* weapons):
.......... +1 .. 0 .... -1 .... -2 .... -3 ..
+1: ... 100 .. 0 ..... 0 ..... 0 ..... 0 ...
+2: ... 100 .. 0 ..... 0 ..... 0 ..... 0 ...
+3: ... 100 .. 0 ..... 0 ..... 0 ..... 0 ...
+4: .... 85 .. 15 .... 0 ..... 0 ..... 0 ...
+5: .... 80 .. 20 .... 0 ..... 0 ..... 0 ...
+6: .... 75 .. 20 .... 5 ..... 0 ..... 0 ...
+7: .... 70 .. 20 ... 10 .... 0 ..... 0 ...
+8: .... 65 .. 20 ... 15 .... 0 ..... 0 ...
+9: .... 60 .. 20 ... 20 .... 0 ..... 0 ...
+10: .. 55 .. 20 ... 20 .... 5 ..... 0 ...
Of course, if you look at the jp wiki (http://pso2.swiki.jp/index.php?%E3%82%A2%E3%82%A4%E3%83%86%E3%83%A0%E5% BC%B7%E5%8C%96), the pattern doesn't match after 1~3* weapons (notably, the fact that Danger can still only have a penalty of -1 in some cases).
To adequately prove implementation #1, one would need a 30% booster and preview the results for a 1~3* weapon from +6 to +7. If the penalty changes from -1 to None, then this proves that Failure Rates are mixed in with the Success Rate. However, if the penalty does not change even after reading a Success Rate of Certain, then this does not prove anything.
==================================================
Practicality:
So why is this information useful? If the success rates are known, you can achieve the optimal success rates for the associated costs. For a simple albeit useful example, let's assume a 10*+10 has a success rate of 35%:
....................... Full ................. 20%+Full
1st try ............ 0.35% ............ 0.55%
2nd try ........... 0.5775% ........ 0.7975%
3rd try ............ 0.7254% ........ 0.9089%
4th try ............ 0.8215% ........ 0.9590%
Going by Ship 2 prices:
20% = 8 x Recyle = ~1.55mil
Full Protect = ~1mil
If you compare 1st try 20%+Full and 2nd try Full, Full not only beats the success rate of 20%+Full but comes out costing less. Even if the success rate was as low as 25%:
....................... Full ................. 20%+Full
1st try ............ 0.25% ............ 0.45%
2nd try ........... 0.4375% ........ 0.6975%
3rd try ............ 0.5781% ........ 0.9089%
4th try ............ 0.8336% ........ 0.9084%
You can see that 4th try Full (~4mil) has a higher chance than 2nd try 20%+Full (~5mil).
I'm sure all these costs will vastly change once innate abilities come out, but just a practical example of how to calculate success rate/cost.
I've compiled a Google Doc that covers the Success Rates:
View-Only: https://docs.google.com/spreadsheet/ccc?key=0AltnsB1gh03-dEFlM2lyaVZqX2dKbmNoRWFiQm1KTGc#gid=0
Calculator: https://docs.google.com/spreadsheet/ccc?key=0AltnsB1gh03-dFBFRUw0d29RRDZYWmxRYWJraXFTYXc&hl=en#gid=0
How to read Success Rate Chart:
-Each value has 4 Success Rates: Base/+5%/+10%/+20%
-C is Certain
-S is Safe
-W is Warning
-D is Danger
How to use the Success Rate Calculator:
-Light grayed box mean the Success Rate may have an additional +5%
-Dark gray boxes means the data has yet to be collected
==================================================
Known Facts:
These are things that, based on collected data, MUST be true:
-Certain = 100%
-Safe begins at 80% *
-Warning begins at 60% *
*Note: These assume intervals of 5%. May have an additional percent of up to, but not including, 5%.
==================================================
To Do:
-Acquire data for 4~6* weapons.
-Acquire data for 10* weapons.
-Acquire data for 9* units.
-Verify 1~3* weapons have identical success rates.
-Verify 4~6* weapons have identical success rates.
-Verify 7~8* units have identical success rates.
-Verify success rates remain the same after leveling innate ability.
==================================================
Issues:
It should be noted that there are 2 major problems with testing the success rates:
1. I do not have access to a 30% Grinder booster
2. Due to 15% not existing, this means the results can never be foolproof as gaps exist in the data
==================================================
Logical Assumptions:
With that said, there are many logical assumptions I can make based on patterns:
1. SEGA most likely implemented the success rates based on intervals of 5
2. SEGA most likely made the risks labels at multiples of 10 (and not inbetween these multiples)
To support my claim, look at the values for grinding a 1~3*:
............. 0 ... 5 .. 10 .. 15 .. 20 ..
+1: ...... C ... C ... C ... C ... C ...
+2: ...... C ... C ... C ... C ... C ...
+3: ...... C ... C ... C ... C ... C ...
+4: ...... S ... S ... S ... ? ... C ...
+5: ...... S ... S ... S ... ? ... C ...
+6: ...... W .. S ... S ... ? ... S ...
+7: ...... W .. W .. S ... ? ... S ...
+8: ...... W .. W .. W .. ? ... S ...
+9: ...... W .. W .. W .. ? ... S ...
+10: .... D ... W .. W .. ? ... W ...
You may notice a diagonal pattern if you fill in the ? with the right success rates, with the success rates shifting right every level. Based on this, I can make these conclusions:
-Success rates exist in intervals of 20%s
-Certain = 100%
-Safe = 80%~95%
-Warning = 60%~75%
-Danger = 5%~55%
Of course, it is very possible that the success rates for +4 and +5 as well as +8 and +9 are exactly the same. In the case of 7* weapons for example, the pattern best fits when only changing success rates after increments of 2 grinds. In any case, even if these assumptions are wrong, the error is simply within 5%.
Note that, in the Success Rate Calculator, the grayed boxes show the worst case assumption, but have potential for an additional 5% (or additional <10% if the 5% interval assumption is wrong as well).
==================================================
Failure Rates:
Another thing to consider is how the percentages of penalties work out. There are two possible ways this could be done:
1. The different Failure Rates is mixed with the Success Rate, making a total of 100% that allocates between penalties.
2. The Failure Rate is simply the remainder from the Success Rate, and does a second roll to determine penalty.
The distinction is important because:
1. if the first case is true, then the Failure Rates can be reasonably guessed, if not proven using a 30% booster
2. Grind Boosters and Grind Protects would work vastly differently in both cases.
In the case of the first scenario, here's an example of how it may work (for 1~3* weapons):
.......... +1 .. 0 .... -1 .... -2 .... -3 ..
+1: ... 100 .. 0 ..... 0 ..... 0 ..... 0 ...
+2: ... 100 .. 0 ..... 0 ..... 0 ..... 0 ...
+3: ... 100 .. 0 ..... 0 ..... 0 ..... 0 ...
+4: .... 85 .. 15 .... 0 ..... 0 ..... 0 ...
+5: .... 80 .. 20 .... 0 ..... 0 ..... 0 ...
+6: .... 75 .. 20 .... 5 ..... 0 ..... 0 ...
+7: .... 70 .. 20 ... 10 .... 0 ..... 0 ...
+8: .... 65 .. 20 ... 15 .... 0 ..... 0 ...
+9: .... 60 .. 20 ... 20 .... 0 ..... 0 ...
+10: .. 55 .. 20 ... 20 .... 5 ..... 0 ...
Of course, if you look at the jp wiki (http://pso2.swiki.jp/index.php?%E3%82%A2%E3%82%A4%E3%83%86%E3%83%A0%E5% BC%B7%E5%8C%96), the pattern doesn't match after 1~3* weapons (notably, the fact that Danger can still only have a penalty of -1 in some cases).
To adequately prove implementation #1, one would need a 30% booster and preview the results for a 1~3* weapon from +6 to +7. If the penalty changes from -1 to None, then this proves that Failure Rates are mixed in with the Success Rate. However, if the penalty does not change even after reading a Success Rate of Certain, then this does not prove anything.
==================================================
Practicality:
So why is this information useful? If the success rates are known, you can achieve the optimal success rates for the associated costs. For a simple albeit useful example, let's assume a 10*+10 has a success rate of 35%:
....................... Full ................. 20%+Full
1st try ............ 0.35% ............ 0.55%
2nd try ........... 0.5775% ........ 0.7975%
3rd try ............ 0.7254% ........ 0.9089%
4th try ............ 0.8215% ........ 0.9590%
Going by Ship 2 prices:
20% = 8 x Recyle = ~1.55mil
Full Protect = ~1mil
If you compare 1st try 20%+Full and 2nd try Full, Full not only beats the success rate of 20%+Full but comes out costing less. Even if the success rate was as low as 25%:
....................... Full ................. 20%+Full
1st try ............ 0.25% ............ 0.45%
2nd try ........... 0.4375% ........ 0.6975%
3rd try ............ 0.5781% ........ 0.9089%
4th try ............ 0.8336% ........ 0.9084%
You can see that 4th try Full (~4mil) has a higher chance than 2nd try 20%+Full (~5mil).
I'm sure all these costs will vastly change once innate abilities come out, but just a practical example of how to calculate success rate/cost.