Flip Probability

[Hooloovoo]

Guess who was bored and did a probability analysis for attack and damage flips

These are the base probabilities for a full deck for the first flip of a round.  The specific probability for any given flip will change as a round progresses and your deck is depleted but these values would provide a starting value to inform decisions.  These were calculated using brute force calculations with the number of duels calculated shown as "n duels".  The probabilities for a given set don't add up to exactly 100% due to the values being derived for a large set of averaged results instead of.

[Adran]

Cheers

One thing that I noticed that is probably an artifact of the brute force approach. The Black joker is always a more likely occurrence (or tied on a straight flip) than the red joker due to the nature of it being forced to be picked, but you have several instances where the Red joker is more likely.

 

I've tried doing the pure calculations a few times, but it gets time consuming to do so I never get around to finishing them.

[Trample]

Interesting. After a quick browse of the number I didn't see any numbers that were really surprising o far off of what I might expect. Maybe it is slightly more difficult to get a high enough delta to achieve a positive damage flip when you're +4 on a stat with a positive flip than I would have thought. What about you? You've obviously dug in quite a bit here (thank you). What stood out to you? 

[Hooloovoo]

@Adran:  Oops,  my bad I forgot to account for how the jokers interact when both are flipped at the same time.  I'll correct my code and re-run it although I don't think it will change the results by more than a percent and with how the absolute values are more dominated by the weak result convergence.

@Trample: I noticed that the neg_str and str_pos are the same hit percentage wise so now I have a better idea how valuable having a positive defense flip is.  Beyond that just having a better idea of how the hit statistics break down and how valuable focus is for hitting and damage.  I'm just starting to work on Misaki's crew so having an idea of how the math works out helps.

[Adran]

1 hour ago, Hooloovoo said:

@Adran:  Oops,  my bad I forgot to account for how the jokers interact when both are flipped at the same time.  I'll correct my code and re-run it although I don't think it will change the results by more than a percent and with how the absolute values are more dominated by the weak result convergence.

@Trample: I noticed that the neg_str and str_pos are the same hit percentage wise so now I have a better idea how valuable having a positive defense flip is.  Beyond that just having a better idea of how the hit statistics break down and how valuable focus is for hitting and damage.  I'm just starting to work on Misaki's crew so having an idea of how the math works out helps.

There are some peculiarities that you might not notice this way, like the mode of a positive flip ( result that is most common) is 13. And of course often the biggest affect of negative flips is the no cheating that it causes.   

You're right that there will probably be very little difference in percentages around the jokers and you might even have your program correct, it's a relatively rare occurrence that it matters. 

[Maniacal_cackle]

Very useful!

I like how it highlights you still have a decent chance to hit even with a weak stat.

Corpse candles spamming attacks is going to add up, for example.

[Harlekin]

On 7/17/2020 at 5:30 AM, Hooloovoo said:

Guess who was bored and did a probability analysis for attack and damage flips

These are the base probabilities for a full deck for the first flip of a round.  The specific probability for any given flip will change as a round progresses and your deck is depleted but these values would provide a starting value to inform decisions.  These were calculated using brute force calculations with the number of duels calculated shown as "n duels".  The probabilities for a given set don't add up to exactly 100% due to the values being derived for a large set of averaged results instead of.

Thanks a lot. 
I sadly am late to the party and your picture link is brocken. Would you mind to share it, again? 

[Hooloovoo]