Sentinel Spreadsheet data validation (running tally)

[Underfyre]

4 minutes ago, Sir Myshkin said:

What's worse: that I couldn't remember the Pylon formula off the top of my head, or the fact I realized I didn't need it to know what DPS 8:30 would produce? Also, you should be getting better than that without even considering Opportunity or Reactive/Degenerative -Res/-MaxHP. 

 

The real amusing part would be if you could get seven bodies to fill a team for the Gaussian proc, that thing would go off several times per minute and heavily skew results there as well.

 

I napkin mathed the gap versus the damage and there didn't look to be much difference in Gloom > Abysaal > Gloom > Antumberal versus Gloom > Abyssal > Dark Oblit > Antumberal until I remembered that the back-end formula on AoE for Procs will wreak havoc on Dark Oblit for ST on a Pylon alone which made me think of this:

 

If you haven't built it yet, I could suggest a minor tweak just for the kicks of this test and trade Engulfing Darkness for Smite and steal an extra slot from Netherworld Tentacles to give it a 5-piece suite of Hecatomb (exclude acc/rech) plus an extra damage proc. 

 

That'd give you Gloom > Abyssal > Smite > Antumberal and still keeps the bulk spirit of the build while incidentally giving it a rather heavy ST bump. Yes yes, Smite is a great ST attack but I had reasons for not including it originally.

I'm using an inverse formula where I put in predicted DPS to get Time, so that's where that number is coming from.  I've already build the character and I'm about the first start the test.

[Underfyre]

Alright, Dark/Rad test coming to you courtesy of @Sir Myshkin. Check his Post for the build.

 

So off the bat, Meltdown has a cooldown of ~118 seconds and Adrenal Booster has a cooldown of ~163 seconds, and I'm not going to expand the sheet out to cover that kind of time span. So the expected will probably be lower than the actual, as I've seen in almost every test so far. So this build can be End hungry, Ageless offsets it at first, but towards the end you have to hit Proton Therapy on cooldown to keep your endurance up, this ties up some cast time. Being pinned to Defensive Opportunity also helps with End use. Damage wise, it never went below half health, very sturdy with a near constant 44% (7.5% after resists) to hit debuff. Proton Therapy was purely to keep End up. Also, I haven't modeled for Tactics to hold Build Up. Whatever, it's fine.

 

Rotation: 3-6-4-7-19-(2)-(18)

Expected: 202.98, these tests are going to take forever. Roughly 8 and a half minutes per run. I'm not excited.

 

601 : 191.63

 

Switched rotation and build, dropped Engulfing Darkness to pick up Smite. Integrated Gloom into rotation, first run saw several procs. Retooled the Gaussian proc in the Aim slot to have Tactics numbers just to account for the damage increase. Make sure you make a duplicated page before you do this stuff kids.

Rotation: 2-4-6-7-20-(18)

Expected: 206.9

 

365 : 232.88

438 : 215.37

431 : 216.80

445 : 214.00

429 : 217.21

 

For the next runs I swapped the Ragnarok piece out of Dark Oblit for the Annihilation proc

Expected: 226.61

 

430 : 217.00

420 : 219.13

414 : 220.45

375 : 230.08

389 : 224.17

Edited May 20, 2021 by underfyre

[Underfyre]

2 hours ago, Sir Myshkin said:

That'd give you Gloom > Abyssal > Smite > Antumbral and still keeps the bulk spirit of the build while incidentally giving it a rather heavy ST bump. Yes yes, Smite is a great ST attack but I had reasons for not including it originally.

 

After putting in the numbers, I can say that adding Smite on its own isn't going to speed things up. The only saving grace here would be that Gloom is now an active part of the rotation and will give the Opportunity Strike proc a chance to shine, but at a 12% chance, I'm hesitant to call it shiny. Smite is also no Mind Probe, doing 3/4 of its damage, it's an alright attack. It could replace Netherworld Tentacles, but it would still give similar DPS. 

Edited May 20, 2021 by underfyre

[Underfyre]

Today's build is Psi/Regen/Ninja. This is the character I built after I challenged @ScarySai to a race between Sentinels and Blaster with the lowest end damage producers and he never acknowledged. I used Ninja instead of Psi to get a hold of some -res debuffs. The secondary does nothing to boost damage output, and Psi is filled with lackluster attacks. On top of that, the T1 ability is absolute garbage, so you're locked into Defensive Opportunity. Counterintuitively, I have Decimation slotted in the hardest hitting attack, now that I think about it, but whatever. Lotus Drop almost guarantees a Fury of the Gladiator proc, and Sting of the Wasp should give about a 50% uptime of Achilles' Heel.

The Rotation: 2-4-8-20-22-(5).  At the tail end of Ageless a gap does start to form between SotW and LD.

The build:

Spoiler

| Copy & Paste this data into Mids Reborn : Hero Designer to view the build |
		|-------------------------------------------------------------------|
		|MxDz;1394;671;1342;HEX;|
		|78DA65934953134114C73BC9C410C222209B908424205B18092EA59616558A502A5|
		|831513C59A931B461A8D440110E72F41BB894DFC083B85DF40B7877BDB91EBCB9A1|
		|5E3DC537F3FF0BA962AA667E6F5EF7FFF57BAFBBE7AE4F353C99B931A97C4DA7CA5|
		|6A552C86B67CD767439386795ECA2729F90BCD1FFFE42CE724A7AC1CC56D68B8B76|
		|B17052546BDBA353FA9A762ADACCE99276F4AAB5662F3BAA31BBBC5C36A7EDD2A24|
		|C2A45F057767FEB3D3BBFA2F542B367CE6A6B41AF5616ED959ED32B76D1DC0A7CDE|
		|7696ACC29C2CA657D7BB24A394BCFE21F9F8DC1CAB4175DF502A6328FF06F900341|
		|E928FC05D8FC1DCB6D650BD01CFB73F4AC6C83898498079D1F8B89EEF2B7CBBBF81|
		|1DDFC1AE1FE44F30B6095E146D005A15E81364C2CA105F50215EF01DF37B0FD67D2|
		|03F82F59FC8CF6050B421C4F3854695E76BD8F07BBC246361C435C2ACAD81B535B1|
		|B6A6FFB5F5817ED9E608348148569094DC2F80891C98CA830199DBC8B51B3BC5884|
		|A1F2610675ED66EE658337B9361AF4232D6C27A5B26906BEB01F220B8E7307908EC|
		|F802ED65D1B6B17F6DFDA8D7903CDAB987EDAE21BED434E6A766C0FE33E459302C9|
		|A4E6A3AA9E9A766909A116AD2E7C03A59BB9BFBDE9D465E7BC74813EC1D2733602C|
		|0DEDD361A57AB85E4F4C8C41891B07C7926482FC1BF07A19911CA3E8938A56E5516|
		|1C920CE0CE22F7002FA5E92AFC8D760E20DF9167C261924192DC993724FE20D30DE|
		|C04DCCDB778BBC4DDE0107EF92CFA5B236431524DE10B54347FCDE491E3E4A1E238|
		|F832327D08D9149742329958DF2748C329734A93A943299A7C99D6995F9E3DCF5F1|
		|046ECDD2B07B73DD7B53ADF61A8A4FD5FD954ECDEEF0640D5E74F1F83CCF66FDD63|
		|DA6E7D70ECFEF5ACF3CB2F953EBE31DBE22BE0C7C55ABC6BE5A63676BEC7F420ED4|
		|76|
		|-------------------------------------------------------------------|

 

 

Expected: 225.93

 

384 : 227.69

390 : 226.15

399 : 223.93

363 : 233.46

332 : 243.33

 

Swapped the T1 for the T2, other wise everything is identical.

 

Expected: 212.33

 

417 : 219.78

428 : 217.42

458 : 211.55

425 : 218.05

417 : 219.78

Edited May 22, 2021 by underfyre

[Underfyre]

So my mathematician wife says that a 5% variance is statistically significant, but my sample size at minimum needs to be 33 runs to be valid. I'm not going to run these different builds 33 times a piece. So y'all choose which build, and I'll focus on that and get some statistically valid runs.

[Bopper]

1 hour ago, underfyre said:

So my mathematician wife says that a 5% variance is statistically significant, but my sample size at minimum needs to be 33 runs to be valid. I'm not going to run these different builds 33 times a piece. So y'all choose which build, and I'll focus on that and get some statistically valid runs.

Keep in mind, the pylon DPS formula is not very accurate. It assumes consistent HP regeneration when in fact chunks of HP come back in 15s intervals. This leads to fluctuations in DPS, for example it's possible to have a longer pylon time but a higher DPS. It's unintuitive, but it's completely true.

 

Long story short, I wouldn't sweat the variance too much. Your numbers are coming in very close to the spreadsheet calculations, and that's all that matters. Keep in mind though, although this helps validate your spreadsheet you must understand that when comparing A to B, if it's close you can't say definitely which is better. In those situations, you may look to doing a large sample size test to help determine that answer.

Edited May 22, 2021 by Bopper

[Underfyre]

25 minutes ago, Bopper said:

... when comparing A to B, if it's close you can't say definitively which is better. In those situations, you may look to doing a large sample size test to help determine that answer.

 

I'll have to reserve that when comparing the different sheets against each other. Especially my voracious claims on Blasters. Thankfully, just about every other AT I have a sheet for does more damage than a Sentinel, so it won't take as long for each run.

[Bopper]

14 minutes ago, underfyre said:

 

I'll have to reserve that when comparing the different sheets against each other. Especially my voracious claims on Blasters. Thankfully, just about every other AT I have a sheet for does more damage than a Sentinel, so it won't take as long for each run.

I was mostly referring to comparing Sentinel A to Sentinel B. If you have a Blaster sheet, you'll want to validate that as well. I imagine defiance will be the major change between sheets.

[Bopper]

Btw, I noticed your expected Elec times changed with VS. What did you tweak to correct it?

[Underfyre]

1 minute ago, Bopper said:

Btw, I noticed your expected Elec times changed with VS. What did you tweak to correct it?

 

I think I copied the cells from an ability that has 2 pets off another sheet. Probably Haunt off of the Dominator sheet; I did a lot of work on that one since Doms are pet heavy. That was making the pet do double damage. After that I inserted your formula for the pets contribution to Reactive Radial, then lengthened out the rotation to take up around 56 seconds and put his cast time back in. That plus including the cast time for Aim, and changing Aim's formula to not include its cast time.

 

Now Aim is just Boost% * 10 / Recharge. It used to be Arcana time + Recharge

[Underfyre]

What bothers me there is the pets damage is already calculated down to DPS, then just DPS * rotation time. I guess at that point he would need some kind of variable cast time to compensate for shorter rotation times? I dunno.

[Underfyre]

Also bothersome is that @DarknessEternal looked at the sheet, looked at the damage for the VS was being doubled and only brought it up when I started doing this validation lol

[DarknessEternal]

7 hours ago, underfyre said:

Also bothersome is that @DarknessEternal looked at the sheet, looked at the damage for the VS was being doubled and only brought it up when I started doing this validation lol

Nah, I originally brought it up on your Blaster spreadsheet.

[Bopper]

Any updates for the OP after recent spreadsheet fixes?

[Underfyre]

9 hours ago, Bopper said:

Any updates for the OP after recent spreadsheet fixes?

 

I've been sprinkling stuff into the Sentinel sheet like fixing the PPM calcs, but I haven't added the big stuff from the Blaster sheet in yet. Usually I'm kinda lazy about updating with a separate sheet since it seemed like nobody was actually using the thing, so theoretically you could just use the link on the first post to grab new changes, but this time I'm actually using a separate sheet.

 

This is what's coming for the next round of updates. Myshkin's weird build forced me to reconsider my assumptions on what abilities people will take. DarknessEternal's prodding about Build Up is making me consider forcing abilities in rotation to have their effects. As it is I'm forcing Coordinated Targeting which is inconsistent to say the least.

 

- Minor fix to PPM calculations

- Added Reactive Radial/Core -Res debuff to Voltaic Sentinel Pet.

- Extended rotation bar to increase accuracy of the calculations. The longer your rotation accounts for, the more accurate your results will be.

- Numbers have been tested in game and found to have a variance between +/-5%. (before PPM fixes)

- Added a checkbox for Aim since apparently some mad men don't take it.

- Also added a check box for which mode you're under in Adaptation, and a check box for Molten Embrace since people want to be silly.

- Added pet calcs to Molten Embrace/Offensive Adaptation damage for Fiery Aura > Burn.

- Added pet damage cap to Fiery Aura > Burn.

[Bopper]

23 minutes ago, underfyre said:

Numbers have been tested in game and found to have a variance between +/-5%. (before PPM fixes)

This is the part I'm most interested in. Hopefully your testing was mostly with single target attacks and it won't impact much. But if you were using AoEs for anything, we'll see how much that lowers numbers and what the impact of variance may be.

[Underfyre]

1 minute ago, Bopper said:

This is the part I'm most interested in. Hopefully your testing was mostly with single target attacks and it won't impact much. But if you were using AoEs for anything, we'll see how much that lowers numbers and what the impact of variance may be.

 

Well I listed the rotations used on each one, I usually brain dump things like that once I'm done with it so I can't remember much. I also don't remember stuff good in general. I'm fairly sure aoe abilities were used in most of the rotations simply because just adding in a -res proc is a big boost in damage and they're on a higher PPM than Decimation is. If you're concerned about the PPM calcs affecting aoes, you're right to be, but it also affected the single target abilities. So all the proc chances were off by upwards of 16% I think.

 

Thankfully I can just update the DPS numbers from the sheet and not have to run a hundred tests again lol

[Bopper]

12 minutes ago, underfyre said:

you're right to be, but it also affected the single target abilities. So all the proc chances were off by upwards of 16% I think.

Ah, I overlooked ST calcs were also wrong. Then yes, this will likely have a large effect on the numbers. If you were dividing by 0.25 + 0.75x60, instead of dividing by 60, that results in about 32.6% improvement to proc chances. I'll be curious to see how low this brings the calculated DPS down.

 

The other thing that makes me curious is the fact your numbers were close before so its fair to assume there is another bug not found yet. The first area I'll want to look is at your resistance debuff calculations to make sure you're not calculating smaller impacts than they should be. Your formulas jump around a lot so I can't quickly look to see if it all makes sense but I'll try to look this week.

[Underfyre]

2 minutes ago, Bopper said:

The other thing that makes me curious is the fact your numbers were close before so its fair to assume there is another bug not found yet.

 

Fair, I'll dump the numbers into the new sheet for a few builds and see what we get. What I was really hoping for was finding out that fixing the numbers would result in being able to use shorter rotations, but I doubt it.

[Underfyre]

Variance has climbed to 10% in 2 of the sets. One uses 2 -res procs, one uses 0 -res procs.

 

To clarify on the PPM calc, your guide says

"PPM * (MRT+ CastTime) / (60 * AreaMod)"

 

But your post on the blaster forum says

 

 

Dividing obviously drives those proc chances even lower. Right now I have it as multiplying.

Edited June 15, 2021 by underfyre

[Underfyre]

Never mind, I gave my wife a reason to talk down to me and she broke it down for me. Apparently A*B/C/D breaks down to AB/CD. So in this case,

 

"PPM * (MRT+ CastTime) / (60 * AreaMod)" and "PPM * (MRT+ CastTime) / 60 / AreaMod" are the same thing. 

 

For those curious:

A*B/C/D = ((A * B) / C) * (1 / D) = (A * B * 1) / (C * D) = (A * B)/ (C * D) = AB/CD

So multiplying 60 * AreaMod in the parenthesis was forcing the pairing where dividing them outside of the parenthesis was going to lead to their eventual pairing.

Edited June 15, 2021 by underfyre

[Bopper]

1 hour ago, underfyre said:

Never mind, I gave my wife a reason to talk down to me and she broke it down for me

You married a good one. Remember, she is always right. And in this case, she definitely is.

[Underfyre]

New test with Fire/Dark, because predicted numbers have fallen dramatically below their observed times. No Build Up procs, no Aim.

 

Predicted: 228.28

 

315 : 249.56

214 : 307.01

264 : 273.08

 

Still dramatically off.

 

 

Edited June 16, 2021 by underfyre

[Underfyre]

@Bopper So I see you replied to this post on Reddit: https://www.reddit.com/r/Cityofheroes/comments/cheum0/more_on_procs/

 

 

Should my area factor on damage procs (not PPM) be divided by 40000 instead of 30000? It shouldn't matter that much since I'm not using Assault Hybrid and only 1 cone ability in the rotation, but it's still something to fix.

Edited June 16, 2021 by underfyre

[Bopper]

15 minutes ago, underfyre said:

@Bopper So I see you replied to this post on Reddit: https://www.reddit.com/r/Cityofheroes/comments/cheum0/more_on_procs/

 

 

Should my area factor on damage procs (not PPM) be divided by 40000 instead of 30000? It shouldn't matter that much since I'm not using Assault Hybrid and only 1 cone ability in the rotation, but it's still something to fix.

Your formula should be correct now. There are many ways you can simplify the formula. What you have will work. This is what you have:

 

0.25 + 0.75 x (1 + R x (11 x A + 540) / 30,000)

 

I have:

 

1 + R x (11 x A + 540) / 40,000

 

They equal the same thing. 0.25 + 0.75 = 1 and 0.75 / 30,000 = 1 / 40,000