Closed filliph closed 7 years ago
Resetting the solver completely produced this output:
Monte Carlo Example
===================
# Action DUR CP QUA PRG IQ WWYW CTL QINC BPRG BQUA WAC Cond S/F
0 40 297 0 0 0 0 442 0 0 0 0 Normal 0
1 Innovation 40 279 0 0 0 0 442 0 0 0 0 Good 1
2 Great Strides 40 247 0 0 0 0 663 0 0 0 0 Normal 1
3 Innovation 40 229 0 0 0 0 663 0 0 0 0 Excellent 1
4 Advanced Touch 30 181 3251 0 0 0 663 3251 0 812 0 Poor 1
5 Great Strides 30 149 3251 0 0 0 663 0 0 0 0 Normal 1
6 Advanced Touch 20 101 4063 0 0 0 663 812 0 812 0 Good 1
7 Careful Synthesis II 10 101 4063 150 0 0 442 0 150 0 0 Normal 1
8 Master's Mend 40 9 4063 150 0 0 442 0 0 0 0 Excellent 1
9 Careful Synthesis II 30 9 4063 300 0 0 442 0 150 0 0 Poor 1
10 Careful Synthesis II 0 9 4063 450 0 0 442 0 150 0 0 Normal 1
Progress Check: true, Durability Check: true, CP Check: true, Tricks Check: true, Reliability Check: true, Cross Class Skills: 2, Wasted Actions: 0
Monte Carlo Result of 500 runs
==============================
DUR CP QUA PRG HQ%
## Expected Value: 0 9 1684.5 450.0 9.0
## Median Value: 0 9 1624.0 450.0 7.0
## Min Value: 0 9 0.0 450.0 1.0
## Success Rate: 100.0 %
It doesn't even do anything for the first Innovation buff before overwriting it!
What am I doing wrong?
Somebody else filed a bug about it using specialist actions when it shouldn't and eventually could only resolve it by switching browsers, which is quite mystifying to to me.
With regards to the solver putting in a wasted Steady Hand, that's due to how the optimization algorithm works. Essentially, it tries many random combinations of a sequence and finds the ones that have the best quality, mutates them by randomly swapping or changing actions and seeing which ones have improved. Rinse, repeat. It's know as a genetic algorithm, meant to mimic evolution. It doesn't truly understand what the actions mean and doesn't have any heuristic rules about not wasting Steady Hand effects etc.
You can try increasing the size of the population and the number of generations that the solver runs for in the options. You can also click the Continue button to let it run for another round of generations with the current settings to see if it can find something better with a later random mutation, but generally it will get stuck on a result that may be "good" but not necessarily the best (aka local minima).
I did find a better outcome after clicking the continue button a few times, thanks :)
Monte Carlo Example
===================
# Action DUR CP QUA PRG IQ WWYW CTL QINC BPRG BQUA WAC Cond S/F
0 40 297 0 0 0 0 442 0 0 0 0 Normal 0
1 Comfort Zone 40 231 0 0 0 0 442 0 0 0 0 Good 1
2 Great Strides 40 207 0 0 0 0 442 0 0 0 0 Normal 1
3 Innovation 40 197 0 0 0 0 442 0 0 0 0 Normal 1
4 Advanced Touch 30 157 812 0 0 0 663 812 0 812 0 Good 1
5 Great Strides 30 133 812 0 0 0 663 0 0 0 0 Normal 1
6 Advanced Touch 20 93 1624 0 0 0 663 812 0 812 0 Normal 1
7 Hasty Touch 10 101 1812 0 0 0 442 188 0 188 0 Good 1
8 Master's Mend 40 17 1812 0 0 0 442 0 0 0 0 Normal 1
9 Careful Synthesis II 30 25 1812 150 0 0 442 0 150 0 0 Normal 1
10 Steady Hand II 30 8 1812 150 0 0 442 0 0 0 0 Normal 1
11 Hasty Touch 20 16 2000 150 0 0 442 188 0 188 0 Normal 1
12 Hasty Touch 10 16 2188 150 0 0 442 188 0 188 0 Normal 1
13 Standard Synthesis 0 1 2188 338 0 0 442 0 188 0 0 Normal 1
Progress Check: true, Durability Check: true, CP Check: true, Tricks Check: true, Reliability Check: true, Cross Class Skills: 5, Wasted Actions: 0
Monte Carlo Result of 500 runs
==============================
DUR CP QUA PRG HQ%
## Expected Value: 0 1 2049.8 338.0 9.0
## Median Value: 0 1 2094.0 338.0 10.0
## Min Value: 0 1 188.0 338.0 2.0
## Success Rate: 100.0 %
Is there a way of inputting our own sequences and having it be locked in place, so that it doesn't try to improve upon the sequence, instead just running simulations on that exact sequence? Maybe it would be worth it for me to try to come up with my own rotation and using the simulator to try to predict the outcome (quality rating), if possible.
On the Simulayor page you can edit the sequence to put in exactly the actions you want and see the simulation results in the tabs below. You don't have to run the solver.
The solver is used to figure out a sequence from scratch or to try and improve on an existing sequence.
Yeah that works fine, thank you! Since it's not a valid bug report (other than the browser issue, but as you say it's been reported before) I'll close this.
Thanks!
Could you comment on #133 with the circumstances under which you saw the solver using specialist actions when it shouldn't? Crafter stats, recipe and the solver seed from the log if possible.
I'm simulating Dark Chestnut Lumber, as a lv56 CRP with 467 Craftsmanship, 442 Control and 297 CP.
Here's a screenshot of my crafter attributes for CRP: https://i.imgur.com/ziXfCwq.png
I am given the following output:
Why is Step 6 another Steady Hand II?
I understand that it wants to apply SH to the HT on Step 7, but would it not make more sense to make the HT step 6, and the SH2 step 7?
I then edited the solver to this: https://i.imgur.com/6Ehx6gX.png and hit solve, which produced this:
Am I doing something wrong for the initial solver to produce 500 less quality than it should with a slight amount of manual tinkering?
EDIT: The simulator is even more broken, by turning on specialist actions when that was never a part of the original configuration. I have no idea why it selected specialist actions when editing the solver, since I am not a CRP specialist and did not select it originally.