How good are Baldur's Gate 3 Advantages?

[@orcharddweller]

Hey, as many of you noticed, there's one tricky thing about the advantages, that's missing from the video. A test always fails when you roll 1 and always succeeds when you roll 20. That makes bonuses much weaker, in comparison to advantages at incredibly high DCs. I've implemented this in the app, so feel free to play around with high (over 20) and low (1 and less) DCs. This also has an interesting interaction with random bonuses! If there's anything else you feel is missing or are curious about, let me know! I'd love to include more stuff in the app/make a follow-up video.

[@kideo_h]

Clicking on this video, I didn't expect a 3Blue1Brown-style fully visualized statistical video essay, but the surprised is wholly welcomed. I tip my hat to you, one data analyst to another.

[@orcharddweller]

Thank you! It was soo much fun to make 🤓

[@cotojee]

cool video man! Interesting statistics.

[@Etiennefdp]

I've wanted to know the exact statistics behind advantage for a while now and couldn't find anything easy to understand on Google, so thanks a million times for your amazing video! I hope you have a nice playthrough :)

[@orcharddweller]

Thanks a lot!

[@tomnar3054]

Great video! But one thing is missing in the video (and the app): BG3 uses critical success/failure on 20/1 - so even on DC 99 you should have a 5% chance of success (or 9.75% with advantage). At that point +x boni are basically useless, but advantage still doubles you success rate. (Also: high level rogues are weird with probabilities)

[@orcharddweller]

Yup, that is a great point, I need to add that to the app at some point.

[@orcharddweller]

Hey, this should be fixed now!

[@dragoonsunite]

Exactly this. The key advantage to advantage in my view for a skill monkey is mitigating critical failures since most checks are low enough for then to pass. Its "doubling" the odds of critical success since you take the high number, but its dividing the odds of critical failure by 20 making the odds of failure now 0.2% or 1 in 400.

[@Giltintur]

Thank you! I was thinking about this and would have had a go at it tomorrow in excel, but this is waaaaay better!

[@orcharddweller]

Go for it! There’s still plenty of things to figure out 😉

[@YuumiGamer1243]

gotta say, I really enjoyed the way you slimmed the complicated looking code back into simpler and simpler code

[@jaredvp]

I had no idea how much i needed to see this until now 😮

[@orcharddweller]

Wow, thanks!

[@DrZaius3141]

First time I heard about 5e's concept of advantage, I immediately took to Excel and calculated the impact. And when I failed a roll that was virtually guaranteed to succeed, I made an Excel tool to calculate probabilities. And while you might be more professional, I feel that I have to point out that your tool has a crucial mistake. Rolling a 1 on your d20 is always a failure, regardless of the bonuses, while a 20 is always a success. Also, the mouse-over on your histogram reads "Total: x (failure)"/"Total: x (success)" not depending on the DC you set but on your baseline DC of 11.

[@orcharddweller]

Thank you! I think that Excel is a great tool for those visualizations ☺ I've fixed the nat 1 nat 20 problem, this also required me to rewrite the plot a little bit, so the problem you've found should no longer apply. But that is really impressive that you've noticed it! It's a very easy-to-miss thing, thank you for letting me know!

[@DrZaius3141]

@@orcharddweller That's some impressive patching speed. Found another bug, though. Set DC to 1 and give advantage. Obviously, that's 399/400 = 99.75%, yet it shows 99.25% (the dice rolls work). My best guess is that you manually inserted the clause for two nat 1s and mixed up the maths.

[@orcharddweller]

@@DrZaius3141 Ooops, yup! It was actually a typo, I picked the probability from a wrong distribution for the nat 1 case. I owe you one 😅

[@OzixiThrill]

I'm reasonably confident that, at least in 5e, critical success and failure only applies in attack rolls. In skill checks, a 1 is just the lowest possible result, but it could still allow you to succeed with enough skill bonuses.

[@dane4kapuse4ka]

nice work, thank you! 💋

[@orcharddweller]

Thanks!

[@Sordorack]

Great Video, but if I may, i would *please* implor you not to use red and green like at ca 3:00! ^^ I have a mild form of red-green-color blindness and could not discern at all where which color was. An easy way of fixing this for next time would be for example to use blue and red, as those dont have vision deficiencies between them (and if you use stronger colors, even better). For reference the histogramm at about 7:20 was very readable for me because of the stronger colors and bigger areas but people with a heavier vision definciency would maybe still not see well which color is which. Dont worry to much about it, its an easy mistake to make but i also think its just as easy to do it better next time maybe. As said before, thank you for the great video, I really like it and I'm gonna look at your other stuff too later if you have similar math-y content :D

[@orcharddweller]

Hey, sorry about that I’ll render a different version as I’ll get access to my computer and post a link to it in description!

[@Sordorack]

@@orcharddweller No problem, thank you for that. I just wanted to give an easy tip for better readability

[@orcharddweller]

@@Sordorack Thanks again for the feedback, in the future I'll try to make sure to never rely on the color alone. Here's a version with blue-red: kzbin.info/www/bejne/j5LSoHmhm6eJgrM

[@TheGrandable]

great video

[@orcharddweller]

Thank you!

[@pjoterem141]

a banger video

[@orcharddweller]

Thanks! 🙏

[@alexparra7979]

Love the video

[@orcharddweller]

Thanks!

[@lixnix2018]

There are a few cases where you need to pass a 30 and even 99 / few cases (a few options that are only meant to be passed with a critical success so advantage would double your chance more or less)

[@lixnix2018]

You will always succeed with a 20 so you need to change the app since no matter the dc you will always have a 5% chance on success at least

[@orcharddweller]

Great point! Fixed it!

[@pralinor]

Factually... When you roll with advantage critical failure and critical success (aka natural 1 and natural 20) completly change. On normal throw, you have a flat 5% chance rolling one or the other. with advantage you roll a critical failure (aka natural 1 or fumble) only if you roll a double 1, so 1/400. You roll critical success on any dice rolling a 20, so 39/400 (9.8%). HUGE difference. Especially the critical failure. Samy with disadvantage except this time your chances to crit are absymally low...

[@Sariiger]

I think you overcomplicated the advantage calculation a little. To see what the odds of rolling 16 or more are with advantage, we can calculate the odds of neither dice rolling 16 or more, and subtract that from 100%. The chances of rolling lower than 16 is 75%. Both die rolling lower than 16 is 75% * 75% = 56.25%. So the chance of at least one dice rolling at least 16 is 100% - 56.25% = 43.75%. Which is the exact result you also got, but it's just a multiplication followed by a subtraction. But other than that, the app is really neat and I love it!

[@orcharddweller]

Hey, I’m glad you’ve liked it! I’ve decided to calculate with the number of squares, because it makes for a prettier visualization, but yeah, the way you did it is cool as well ☺️

[@ricardocalderon9823]

This didn't really make me a better player but it was still interesting nonetheless lol

[@TeamCykelhold]

interesting, cool and well explained - but best with speed x1.75 :)

[@PythonPlusPlus]

Advantage = 1 - P(failure)^2 So if you have a really low probability of success, you will probably gain more from a +x bonus instead of advantage. But in most situations advantage is always better than a +x bonus.

[@shabbii296]

Now do disadvantage

[@orcharddweller]

Its the same just the other way 😅

[@shabbii296]

@orcharddweller so it just rolls 2 and you get the worse of the 2

[@orcharddweller]

Yes, and it is the worst in comparison to bonuses at mid dc range (e.g. disadvantage at dc 11 is as bad as -5)

[@Daniel-fq2bj]

Watching this video, I can now comfortably say that i know less than when i started